Medical Radiology · Diagnostic Imaging

Series Editors: Hans-Ulrich Kauczor · Paul M. Parizel · Wilfred C.G. Peh

Hatem Alkadhi
André Euler
David Maintz
Dushyant Sahani   Editors

Spectral
Imaging
Dual-Energy, Multi-Energy and
Photon-Counting CT
Medical Radiology

Diagnostic Imaging

Series Editors
Hans-Ulrich Kauczor
Paul M. Parizel
Wilfred C. G. Peh
The book series Medical Radiology  – Diagnostic Imaging provides accurate
and up-to-date overviews about the latest advances in the rapidly evolving
field of diagnostic imaging and interventional radiology. Each volume is
conceived as a practical and clinically useful reference book and is developed
under the direction of an experienced editor, who is a world-renowned
specialist in the field. Book chapters are written by expert authors in the field
and are richly illustrated with high quality figures, tables and graphs. Editors
and authors are committed to provide detailed and coherent information in a
readily accessible and easy-to-understand format, directly applicable to daily
practice.
Medical Radiology  – Diagnostic Imaging covers all organ systems and
addresses all modern imaging techniques and image-guided treatment
modalities, as well as hot topics in management, workflow, and quality and
safety issues in radiology and imaging. The judicious choice of relevant topics,
the careful selection of expert editors and authors, and the emphasis on
providing practically useful information, contribute to the wide appeal and
ongoing success of the series. The series is indexed in Scopus.

More information about this series at https://link.springer.com/bookseries/174

Hatem Alkadhi  •  André Euler
David Maintz  •  Dushyant Sahani
Editors

Spectral Imaging
Dual-Energy, Multi-Energy
and Photon-Counting CT
Editors
Hatem Alkadhi André Euler
Institute of Diagnostic Institute of Diagnostic
and Interventional Radiology and Interventional Radiology
University Hospital Zurich University Hospital Zurich
Zurich Zurich
Switzerland Switzerland

David Maintz Dushyant Sahani

Department of Radiology Department of Radiology
Universitätsklinikum Köln University of Washington
Köln, Nordrhein-Westfalen Seattle, WA
Germany USA

ISSN 0942-5373     ISSN 2197-4187 (electronic)

Medical Radiology
ISSN 2731-4677     ISSN 2731-4685 (electronic)
Diagnostic Imaging

ISBN 978-3-030-96284-5    ISBN 978-3-030-96285-2 (eBook)

https://doi.org/10.1007/978-3-030-96285-2

© The Editor(s) (if applicable) and The Author(s), under exclusive license to Springer Nature
Switzerland AG 2022
This work is subject to copyright. All rights are solely and exclusively licensed by the Publisher,
whether the whole or part of the material is concerned, specifically the rights of translation,
reprinting, reuse of illustrations, recitation, broadcasting, reproduction on microfilms or in any
other physical way, and transmission or information storage and retrieval, electronic adaptation,
computer software, or by similar or dissimilar methodology now known or hereafter developed.
The use of general descriptive names, registered names, trademarks, service marks, etc. in this
publication does not imply, even in the absence of a specific statement, that such names are
exempt from the relevant protective laws and regulations and therefore free for general use.
The publisher, the authors and the editors are safe to assume that the advice and information in
this book are believed to be true and accurate at the date of publication. Neither the publisher nor
the authors or the editors give a warranty, expressed or implied, with respect to the material
contained herein or for any errors or omissions that may have been made. The publisher remains
neutral with regard to jurisdictional claims in published maps and institutional affiliations.

This Springer imprint is published by the registered company Springer Nature Switzerland AG
The registered company address is: Gewerbestrasse 11, 6330 Cham, Switzerland
Foreword

CT was the first cross-sectional imaging technique to be widely introduced

over 50 years ago, and it was enthusiastically received by the medical com-
munity. After the advent of magnetic resonance imaging, however, CT began
to lose its appeal. Multiplanar slices, superior soft tissue contrast, and the
absence of ionizing radiation were considered substantial advantages. Not to
be forgotten, MRI was considered more intellectually demanding, because it
was necessary to select from the multitude of pulse sequences the one best
suited to the clinical problem at hand. Even new and specific contrast agents
were developed. Therefore, it was not surprising that CT was considered a
diligent and patient, but somewhat boring workhorse.
Tremendous developments have occurred since then: Spiral CT, multislice
CT, iterative image reconstruction, and finally Spectral Imaging and Photon
Counting CT.  It is to the great credit of the editors of this book, Spectral
Imaging: Dual-Energy, Multi-Energy and Photon-Counting CT, to provide a
competent and easily understandable overview of these fascinating
developments.
The international editorial team of Hatem Alkadhi, André Euler, David
Maintz, and Dushyant Sahani, all distinguished and highly recognized imag-
ing scientists, have recruited experienced and knowledgeable experts as
authors.
In the first part, the methodological and technical fundamentals of spectral
CT and photon counting CT as well as the application of contrast agents are
presented. These chapters impress by the clarity of the presentation, so that
they are understandable even for physicians without a more extensive back-
ground in science and engineering. The fundamentals of spectral CT and pho-
ton counting CT are covered, as are the emerging future developments. In
spectral CT, the major manufacturers pursue different strategies, which are
explained in individual chapters, so that the reader can make his or her own
objective assessment.
The chapter on contrast media impressively demonstrates that spectral and
photon counting CT also allow a differentiated and targeted use of contrast
media, which contributes to an improvement in the diagnostic accuracy. With
the new CT technologies, elements other than iodine, such as lanthanides,
hafnium, tantalum, tungsten, or even gold, could also be used as a basis for
new contrast media, which could result in fascinating and completely new
indications.

v
vi Foreword

The main focus of the book is taken by chapters on the clinical applica-
tions of spectral imaging in various organs and diseases. It becomes clear that
spectral imaging has great potential for a variety of clinical problems and that
extensive experience has been gained in this field in recent years. It becomes
clear that these are often very important, even therapy-decisive questions,
such as the detection or exclusion of an active hemorrhage or a fresh vertebral
fracture.
In the last two contributions, the question is raised as to what will happen
next with CT.  What can the new CT techniques in radiomics and artificial
intelligence contribute and what can be expected from photon counting CT.
Due to new developments, the limitations of CT compared to MRI mentioned
at the beginning of this preface no longer exist or exist only in part. Spectral
CT can provide multiplanar imaging, accurate tissue characterization, and is
associated with significantly lower radiation exposure. Needless to say, mod-
ern CT is anything but boring but an exciting and dynamically evolving tech-
nology that will keep us on our toes for a long time to come!
I would like to recommend this book to the readers, and I am sure that they
will find it as exciting and informative as I do. I would like to congratulate the
editors and the authors on their successful project and thank them for their
effort and hard work.

Prof. em. Dr. Dr. h.c. Maximilian Reiser, FACR, FRCR, ML

Deptartment of Radiology
Ludwig-Maximilians-University
Munich, Germany
Contents

Part I Technical Principles

Material Decomposition and Post-­processing:

History and Basic Principles ������������������������������������������������������������������   3
Jia Wang, Xinhui Duan, and Cynthia H. McCollough
Dual-Energy: The Siemens Approach����������������������������������������������������  15
Bernhard Schmidt and Thomas Flohr
Dual-Energy: The Philips Approach������������������������������������������������������  29
Ami Altman, Galit Kafri, and Sary Shenhav
Dual-Energy: The GE Approach������������������������������������������������������������  45
Scott Slavic and Mats Danielsson
 ual-Energy: The Canon Approach������������������������������������������������������  63
D
Kirsten Boedeker, Jay Vaishnav, Ruoqiao Zhang, Zhou Yu,
and Satoru Nakanishi

Basic Principles and Clinical Applications
of Photon-Counting CT ��������������������������������������������������������������������������  73
Thomas Flohr, Martin Petersilka, Stefan Ulzheimer,
Bernhard Schmidt, Klaus Erhard, Bernhard Brendel,
Marjorie Villien, Philippe Coulon, Salim Si-Mohamed,
and Sara Boccalini

Contrast Media for Modern Computed Tomography��������������������������  93
Hubertus Pietsch and Gregor Jost

Part II Clinical Applications

Neuroradiological Imaging��������������������������������������������������������������������� 109

Sebastian Winklhofer, Dominik Nakhostin,
and Mohammed Fahim Mohammed

Head and Neck Imaging�������������������������������������������������������������������������� 127
David Zopfs

Clinical Applications in Cardiac Imaging���������������������������������������������� 143
Basel Yacoub, Josua Decker, U. Joseph Schoepf,
Tilman Emrich, Jon F. Aldinger, and Akos Varga-Szemes

vii
viii Contents

Dual-Energy CT Angiography���������������������������������������������������������������� 163

Matthias Stefan May and Armin Muttke
Thoracic Imaging: Ventilation/Perfusion���������������������������������������������� 183
Hye Jeon Hwang, Sang Min Lee, and Joon Beom Seo
Thoracic Oncology ���������������������������������������������������������������������������������� 201
Philip Konietzke

Gastrointestinal Imaging: Oncology (Liver, Pancreas,
Bowel Cancer, and Treatment Response)���������������������������������������������� 219
Simon Lennartz and Nils Große Hokamp
Gastrointestinal Imaging: Liver Fat and Iron Quantification������������ 235
Malte Niklas Bongers
Bowel Imaging������������������������������������������������������������������������������������������ 245
Markus M. Obmann

Role of Dual-Energy Computed Tomography (DECT)
in Acute Abdomen������������������������������������������������������������������������������������ 255
Saira Hamid, Muhammad Umer Nasir, Aneta Kecler-Pietrzyk,
Adnan Sheikh, Nicolas Murray, Faisal Khosa,
and Savvas Nicolaou
Spectral Computed Tomography Imaging
of the Adrenal Glands������������������������������������������������������������������������������ 277
Matthias Benndorf, August Sigle, and Fabian Bamberg
Urogenital Imaging: Kidneys (Lesion Characterization)�������������������� 285
Ali Pourvaziri, Anushri Parakh, Avinash Kambadakone,
and Dushyant Sahani

Urogenital Imaging: Kidneys: Urinary Stones�������������������������������������� 295
Nils Große Hokamp
Skeletal Imaging: Bones�������������������������������������������������������������������������� 301
Christian Booz, Julian L. Wichmann, and Tommaso D’Angelo
Gout���������������������������������������������������������������������������������������������������������� 315
Torsten Diekhoff

Dual-Energy CT in Radiation Oncology������������������������������������������������ 333
Christian Richter and Patrick Wohlfahrt
The Future of Spectral CT: Radiomics and Beyond���������������������������� 347
Bettina Baessler and Davide Cester

Photon-Counting CT: Initial Clinical Experience�������������������������������� 363
Victor Mergen, André Euler, Kai Higashigaito, Matthias Eberhard,
and Hatem Alkadhi
 Part I
Technical Principles
 Material Decomposition
and Post-­processing: History
and Basic Principles

Jia Wang, Xinhui Duan,
and Cynthia H. McCollough

Contents
1 Introduction and History   3
2 Methods of Material Decomposition   4
3  ynthetic Images from Material Decomposition and Post-Processing 
S  6
3.1  Mixed CT Images   7
3.2  Material-Specific Images   7
3.3  Virtual Monoenergetic Images   8
3.4  Electron Density and Effective Atomic Number Images   9
4 Image Quality and Quantitative Accuracy of Synthetic DECT Images   10
5 Conclusion   11
References   11

1 Introduction and History lished in the 1970s and the 1980s established the
theoretical foundation of dual-energy CT imag-
The concept of using two-energy beams in CT ing (Alvarez and Macovski 1976; Macovski et al.
imaging dates back to Hounsfield’s early work 1976; Kelcz et  al. 1979; Lehmann et  al. 1981;
(Hounsfield 1973), where he explained that by Kalender et  al. 1986). Due to the limitations of
using two distinct tube potentials, one could CT technology at that time, dual-energy CT could
enhance the contrast of and differentiate between not be used routinely in clinical practice. A break-
specific materials. Several important papers pub- through happened in 2006, when a dual-source
CT scanner was made commercially available,
J. Wang enabling dual-­energy CT in routine patient exams
Department of Environmental Health and Safety, (Flohr et al. 2006; Johnson et al. 2007).
Stanford University, Stanford, CA, USA
e-mail: [email protected]
Currently, CT manufacturers offer several
solutions for dual-energy CT imaging. However,
X. Duan
UT Southwestern Medical Center, Dallas, TX, USA
they all share the similar basic physics principle.
e-mail: [email protected] The data acquisition for dual-energy CT requires
C. H. McCollough (*)
(1) the minimal time interval between the two CT
Department of Radiology, Mayo Clinic, acquisitions, which ideally occur simultaneously,
Rochester, MN, USA and (2) s­ufficient difference of photon energies
e-mail: [email protected]

© The Author(s), under exclusive license to Springer Nature Switzerland AG 2022 3

H. Alkadhi et al. (eds.), Spectral Imaging, Medical Radiology Diagnostic Imaging,
https://doi.org/10.1007/978-3-030-96285-2_1
4 J. Wang et al.

between the two CT acquisitions. The time inter- The three mass attenuation coefficients have
val requirement ensures that the images from the different dependencies on photon energy (Fig. 1).
two CT acquisitions have minimal registration The attenuation of the coherent scatter is much
errors, and the sufficient energy difference smaller than the other two in the energy range of
requirement controls noise level in the post-pro- diagnostic imaging, so its contribution to total
cessed images to ensure acceptable image quality attenuation may be ignored in the physical model,
for diagnosis. After two CT acquisitions are i.e.,
obtained, the post-processing performed in all the
dual-energy CT platforms perform the same æmö æmö æmö
ç ÷=ç ÷ +ç ÷ (2)
basic physics task i.e., material decomposition. è r ø è r ø Photoelectric è r øCompton

2 Methods of Material This mathematic formula for the mass attenu-

Decomposition ation due to the photoelectric effect and Compton
scatter have been long established through theo-
X-ray attenuation characteristics depend on pho- retical physics and experimental measurements
ton energy and the physical property of the inci- (Bushberg et  al. 2012). The mass attenuation
dent material, i.e., mass density (ρ) and atomic coefficient can be separated into a product of a
number (Z). In the photon energy range of diag- material-dependent term and a photon-energy-­
nostic imaging, there are three main mechanisms dependent term, so that the mass attenuation
in X-ray photon interacting with matter: photo- coefficient can be written as
electric effect, Compton scatter, and coherent
æmö
scatter (Rayleigh scatter). The mass attenuation ç ÷ ( E ) = a P fP ( E ) + a C fC ( E ) (3)
èrø
æmö
coefficient ç ÷ of a material is the summation
èrø where αp and αC are material-dependent factors
of the mass attenuation coefficients of the three for the photoelectric effect and Compton scatter,
mechanisms, and fp and fC are energy-dependent factors,
respectively. When ignoring the K-edge effect,
æmö æmö æmö æmö (1) the format of α and f are known through X-ray
ç ÷=ç ÷ +ç ÷ +ç ÷
èrø èrø Photoelectric èrø Compton èrø Coheerent
physics (Bushberg et al. 2012).

Coherent scatter Compton scatter Coherent scatter Compton scatter

Photoelectric Total Photoelectric Total

100.00 100.00
Mass Attenuation Coefficient

Mass Attenuation Coefficient

10.00 10.00
(cm^2/g)

(cm^2/g)

1.00 1.00

0.10 0.10

0.01 0.01
10.00 100.00 10.00 100.00
Photon Energy (keV) Photon Energy (keV)

Fig. 1  Mass attenuation coefficients as a function of pho- scatter, and coherent scatter have distinct dependences
ton energy for cortical bone (ICRU 44) (left) and soft tis- with photon energy. These unique behaviors are the foun-
sue (ICRU 44) (right). Photoelectric effect, Compton dation for dual-energy CT imaging
Material Decomposition and Post-processing: History and Basic Principles 5

To determine the αp and αC, we need attenua- a 3,Ca1, P - a 3, Pa1,C

tion measurements from two different energy lev- m2 =
a1, Pa 2,C - a 2, Pa1,C
els, ELow (low energy) and EHigh (high energy), .
Equation (7) shows the attenuation of any material
æmö
ç ÷ = a P fP ( ELow ) + a C fC ( ELow ) can be represented by a linear combination of two
è r ø Low basis materials. It is worth noting that although Eq.
(5) is deduced from the two physical effects model,
æmö
ç ÷ = a P fP ( EHigh ) + a C fC ( EHigh ) (4) it does not have the limitation that only two physi-
è r øHigh cal mechanisms are modeled since the attenuation

coefficients of the basis materials 1 and 2 contain
This process of determining material com- all the attenuation information. Neither the two
position through two measurements is often physical effects model nor the two-basis material
called material decomposition and Eq. (4) is a model work well when the K-edge effect is a sig-
physical effect model since it is based on the nificant source of attenuation.
photoelectric effect and Compton scatter. If a mixture contains three different materials
This  model is straightforward, but has limita- and we would like to determine the concentration
tions because it only models two mechanisms of each material, we will need another independent
for X-ray interactions with materials; other equation. Usually, we assume mass or volume con-
mechanisms which are ignored in the model servation which means the mass fraction or volume
will cause some error in the calculation fraction of the three materials must add up to 1,
(Williamson et  al. 2006). Another commonly
f1 + f2 + f3 = 1, f1 ³ 0, f2 ³ 0, f3 ³ 0, (8)
used model is the basis material model. Instead
of using the photoelectric and Compton effects where f1, f2, and f3 are mass or volume fractions of
as the base functions, we use the attenuation the materials. This is often called three-material
coefficient of two selected materials, namely decomposition.
basis materials or base materials. Given two The K-edge is a unique attenuation feature
basis materials 1 and 2 (e.g., iodine and bone), (Fig. 2). The abrupt change in attenuation around
the k-edge energy has a different behavior than
æmö
ç ÷ ( E ) = a1, P fP ( E ) + a1,C fC ( E ) the smooth attenuation curves of the photoelectric
è r ø1 effect and Compton scatter. Thus, a K-edge term

æmö
ç ÷ ( E ) = a 2, P fP ( E ) + a 2,C fC ( E ) (5)
è r ø2
Iodine Tissue Bone

1,000.0
Mass Attenuation Coefficient

a third material would have the same format,

æmö 100.0
ç ÷ ( E ) = a 3, P fP ( E ) + a 3,C fC ( E ) (6)
(cm^2/g)

èrø 10.0

Substituting fP(E) and fC(E) from (5) in to (6), 1.0

the third material can be represented as
0.1
æmö æmö æmö
ç ÷ ( E ) = m1 ç ÷ ( E ) + m2 ç ÷ ( E ) (7) 10 100
r
è ø r
è ø1 è r ø2 Photon Energy (keV)

Fig. 2  Mass attenuation coefficient of iodine, soft tissue
a 3 , P a 2 ,C - a 3 ,C a 2 , P (ICRU 44), and cortical bone (ICRU 44). Iodine curve has
where m1 = and an abrupt attenuation increase at 33.2 keV, which is called
a1, Pa 2,C - a 2, Pa1,C the absorption k-edge of iodine
6 J. Wang et al.

needs to be added into the decomposition model tion data. In dual-energy CT, the projection data
when the K-edge contributes substantially to the can be modeled as,
total attenuation. The mass attenuation coefficient
ì E Lm

with the K-edge term can be expressed as ïp

ï
Low
= ò S ( E ) exp ( - A f ( E ) - A f ( E ) ) dE
L p p C C

í
0
(11)
æmö
ç ÷(E) = a fP ( E ) + a C fC ( E ) + a K fK ( E ) (9) ïp
E Hm

èrø = ò S ( E ) exp ( - A f ( E ) - A f ( E ) ) dE
P

ïî High H p p C C
0

where aK and fK are the material- and energy-­
 
dependent terms representing the K-edge effect. where Ap = ò a A ( x ) ds, AC = ò a C ( x ) ds are the
If there are multiple K-edge materials in the mea- projections of the αA and αC, pL and pH are mea-
surements, each material needs to have a separate sured projection data, ELm and EHm are the maxi-
term in the attenuation formula. Similarly, basis mum photon energy for the low- and high-energy
material models can be updated to include K-edge X-ray beams, and SL(E) and SH(E) are the energy
materials, functions, which combine X-ray spectra and
detector response functions. Solving Eq. (11), αA
æmö æmö æmö
ç ÷ ( E ) = m1 ç ÷ ( E ) + m2 ç ÷ ( E ) and αC are obtained by performing CT recon-
r
è ø r
è ø1 è r ø2 struction using Ap and Ac. This is projection-­
æmö space decomposition using the photoelectric and
+ mk ç ÷ ( E ) (10)
è r øk
Compton effects and it works similarly as the
two-basis material model. In principle, the
projection-­space decomposition provides more
æmö
where mk and ç ÷ ( E ) are the terms for the accurate results than image-space decomposition
è r øk since it models the imaging chain more precisely.
K-edge material. However, in practice, the performance difference
When there are multiple K-edge materials in of image-space and projection-space decomposi-
the model, two measurements from dual-energy tion might not be as significant as expected
CT imaging become insufficient to solve the prob- (Jacobsen et al. 2018; Sellerer et al. 2018; Taylor
lem, even with mass or volume conservation. That et al. 2019).
is, the number of unknowns is higher than the
number of equations. Photon-counting CT can
measure the energy of each X-ray photon and store 3 Synthetic Images
the photon counts separately in multiple energy from Material
intervals, namely, energy bins. With the sufficient Decomposition
number of energy bins, this provides sufficient and Post-Processing
independent measurements for K-edge imaging.
Furthermore, photon-counting CT can adjust the In this section, we will discuss the creation,
positions of its energy bins to be just before and application, and pitfalls of the common types of
just after the K-edge energy level to boost the sig- synthetic images from spectral CT.  In conven-
nal from the K-edge attenuation (Roessl and tional single-energy CT, although CT images
Proksa 2007). This opens up the possibility for acquired at the chosen tube potential can be cre-
new CT contrast agents and imaging of multiple ated with multiple reconstruction kernels to serve
contrast agents simultaneously (Schlomka et  al. different clinical tasks, the reconstruction kernels
2008; Muenzel et al. 2017; Willemink et al. 2018). only impact the noise and spatial resolution of the
The material decomposition we discussed so CT images, and the tissue and material contrast
far is calculated using mass or linear attenuation of the CT images is approximately kept the same.
coefficients, which means the decomposition is On the other hand, spectral CT, including dual-­
performed using the CT images. The decomposi- energy CT (DECT), not only provides conven-
tion process can also be performed using projec- tional image sets with the desired noise and
Material Decomposition and Post-processing: History and Basic Principles 7

spatial resolution, but also provides images that image quality in a DECT angiography study
feature different tissue contrasts and additional (Behrendt et al. 2009). In clinical DECT applica-
material- specific information. These synthetic tions, the linear mix weighting factor (wLow) is
images can be generated either from projection commonly set between 0.5 and 0.7 to provide a
data (Alvarez and Macovski 1976; Lehmann balance between iodine contrast and image noise.
et al. 1981; Kalender et al. 1986; Wang and Pelc The mixed images can also be created in a
2011) or CT images of multiple energies (Yu non-linear fashion. Holmes et  al. showed in a
et  al. 2011; Liu et  al. 2009; Niu et  al. 2014; Li liver CT study that a non-linear blending method
et al. 2015) after a spectral CT scan. For clarity based on a modified sigmoid function provides
we will focus our discussion on DECT in the fol- higher iodine CNR and better subjective score
lowing sections, because CT systems with multi-­ compared to a linear mixed approach (Holmes
energy (more than two) capabilities have just 3rd et al. 2008).
been approved for clinical use. However, all
essential features of synthetic images discussed
below apply to both DECT and multi-energy CT 3.2 Material-Specific Images
systems.
The foremost advantage of DECT over conven-
tional single-energy CT is its ability to provide
3.1 Mixed CT Images material specific information for the anatomy of
interest. Through the material decomposition
Mixed images are generated by blending the low- steps discussed before, the density maps of basis
and high-energy CT images from a DECT scan. materials can be calculated to provide quantita-
The primary use of mixed images is for routine tive maps of their distribution.
diagnostic interpretation and they serve as a sur- Based on the two-basis material assumption,
rogate for the conventional single-energy CT solving Eq. (7) using data acquired at low- and
images that the clinical users are familiar with, high energy yields the mass densities of two basis
even though they may be new to DECT materials (Yu et al. 2011):
technology.
æmö æmö
One common way to create the mixed images m Low ç ÷ - m High
ç ÷
is to add the low- and high-energy images with èrø èrø
r1 =
High , 2 Low , 2

user-adjustable weighting factors: æmö æmö æmö æmö

-ç ÷
ç ÷ ç ÷ ç ÷
èrø èrø èrø èrø
I Mixed = wLow I Low + wHigh I High
Low ,1 High , 2 Low , 2 High ,1
(12)

where ILow and IHigh denote the low- and high-­ - m Low ç
æmö æmö
+ m High
÷ ç ÷
energy images respectively, and wLow and wHigh èrø èrø (13)
r2 =
High ,1 Low ,1

denote the corresponding weighting factor, and æmö æmö æ mö

-ç ÷
æmö
ç ÷ ç ÷ ç ÷
wLow + wHigh = 1. The optimal choice of weighting
èrø Low ,1 èrø High , 2 èrø èrø Low , 2 High ,1

factors depends on the clinical task, patient size,
and dose partition between low- and high-energy where μLow and μHigh denote the linear coefficients
scans. Yu et al. showed that linearly mixed images
æmö
from DECT provide similar or improved noise measured at low- and high-energy scan, ç ÷
and iodine contrast-to-noise ratio (CNR) relative è r ø Low,1
æmö
to 120 kV images using the same radiation dose and ç ÷ represent the mass attenuation
over a wide range of phantom sizes (from 30 cm è r øHigh,1
to 45  cm in lateral width), but inferior iodine coefficient of the basis material 1 at the low and
CNR compared to 80 kV images for 30-cm small high energy, respectively, and the similar denota-
phantom (Yu et al. 2009). Behrendt et al. found tion applies to basis material 2. If there is a third
weighting factor 0.5 provides the best subjective material in the mixture, either mass or volume
8 J. Wang et al.

conservation can be assumed to add another inde- and hemorrhage in unenhanced head CT exams
pendent condition to solve the three-material (Hu et al. 2016). Iron maps are used in DECT to
decomposition (Liu et al. 2009). In clinical appli- evaluate the liver iron accumulation for hemato-
cations, the commonly used unit of the mass den- logical patients (Luo et  al. 2015; Werner et  al.
sity maps is mg/ml, which indicates the 2019). Fat quantification in liver has been evalu-
concentration of the material. However, some CT ated in DECT exams using a three-material
manufacturers convert the concentration unit into decomposition including fat, liver tissues, and
CT numbers in Hounsfield units. The advantages iron (Fischer et al. 2011). Xenon maps acquired
of using CT-like density maps is that they are from DECT can be used to evaluate the distribu-
more familiar to clinical users and more compat- tion of morphologic and functional changes in
ible with PACS systems. However, if the users pulmonary diseases such as chronic obstructive
are looking for the absolute concentration infor- pulmonary disease (COPD) and asthma (Kong
mation of certain materials, the density maps et al. 2014).
with the unit of mg/ml are preferred.
DECT provides a variety of material specific 3.2.3 Material Negative Images
images depending on the clinical tasks. With material density maps calculated from
DECT material decomposition, certain material
3.2.1 Material Differentiation Images can be virtually removed from the CT images to
Material differentiation images, sometimes improve diagnostic accuracy and confidence. In
referred as material maps, display a material’s iodine contrast enhanced DECT exams, a set of
distribution, typically in a color-coded fashion, virtual non-contrast (VNC) images are created
but does not provide the concentration informa- to potentially eliminate the need to acquire the
tion. For example, in DECT scan of patient with true non-contrast images, therefore reducing
kidney stones, color maps are generated for dif- radiation dose to patients (Ferda et  al. 2009;
ferentiation of uric acid from non-uric-acid kid- Graser et al. 2009). In DECT of bone marrow,
ney stones based on their difference in effective calcium signal is removed from bone images to
atomic numbers (Primak et al. 2007; Graser et al. evaluate bone marrow edema (Wang et  al.
2008). Similarly, based on DECT differentiation 2013).
of uric acid from calcium, diagnosis of gout was
established by using DECT to identify monoso-
dium urate crystals in synovial fluid or tissue 3.3 Virtual Monoenergetic Images
aspirates (Bongartz et al. 2015). DECT has also
been used to provide color-coded maps of silicon In addition to mixed images and material-specific
for evaluation of the integrity of silicone breast images, DECT datasets can be used to create vir-
implants (Johnson et al. 2013). tual monoenergetic images (VMI) which are
gray-scale images mimicking the appearance of
3.2.2 Material Quantification Images CT images acquired with a true monochromatic
Material quantification images provide both the X-ray source. In theory, monoenergetic images
distribution and concentration information of tar- can be created at any individual X-ray energy, but
get materials. Iodine quantification is the most in clinical practice, the typical range of selected
common use of DECT and has been adopted in a energy is from 40 to 200 keV, depending on the
variety of clinical applications for characteriza- diagnostic tasks.
tion of liver (Lee et  al. 2011) and renal lesions In principle, the creation of VMIs can be con-
(Mileto et  al. 2014), hemorrhage in stroke sidered as the by-product of the mass density maps
patients (Gupta et al. 2010), thyroid nodules (Li from material decomposition in Sect. 3.2. Once
et al. 2012), and bowel disease (Fulwadhva et al. the mass density images are generated from either
2016). Calcium quantification with DECT has projection-space or image-space material decom-
been applied to differentiate between calcium position, the linear attenuation coefficient μ(E) can
Material Decomposition and Post-processing: History and Basic Principles 9

be calculated based on basis material mass density than conventional 100 kV or 120 kV CT images
maps and each material’s mass attenuation coeffi- (Albrecht et al. 2019). However, the image noise
cient at a chosen energy, as shown below: of VMIs also increases at lower energies. Alvarez
et al. showed that the noise of VMIs has a non-
æmö æmö
m ( E ) = ç ÷ ( E ) · r1 + ç ÷ ( E ) · r2 (14) monotonic relationship with the monochromatic
è r ø1 è r ø2 energy and a minimum noise exists (Alvarez and
Seppi 1979). Yu et  al. showed that a maximum
æmö æmö
where ç ÷ ( E ) and ç ÷ ( E ) denote the mass CNR of iodine exists at certain monochromatic
è r ø1 è r ø2 energy, but the optimal energy level depends on
attenuation coefficient of the two basis materials the phantom size and dose partition between the
at the chosen energy E, and ρ1 and ρ2 are their low- and high-energy scans. Third, when the
mass density. The VMIs can then be obtained by monoenergetic images are created from projec-
scaling the μ(E) from Eq. (14) with μ(E) of water tion-space material decomposition, the beam
to create the CT-like images. Alternatively, Yu hardening artifacts can be more effectively
et al. showed that monoenergetic images can be reduced because the shape of the polychromatic
calculated by the linear combination of CT X-ray spectrum and spectral response of CT
images from the low and high energy (Yu et al. detectors are properly considered in the material
2011): decomposition process. Fourth, because beam
CT ( E ) = w ( E ) · CT + (1 - w ( E ) ) · CT (15) hardening is a significant cause of metal-induced
Low High
artifacts in CT scan, monoenergetic images cre-
where CTLow and CTHigh denote the low- and high-­ ated at higher energy levels can potentially reduce
energy CT images, and w(E) is the weighting fac- artifacts caused by metal objects (Cha et al. 2017;
tor. Two things are worth mentioning here. First, Guggenberger et al. 2012). Fifth, with VMIs cre-
the w(E) here should not be confused with the ated over a range of monochromatic energies,
weighting factor wLow and wHigh used for creating spectral attenuation curves can be created to char-
mixed images in Eq. (12). Second, although not acterize tissue types, such as differentiation of
given explicitly in Eq. (15), the approach of lin- contrast enhancing solid renal mass from hyper-
ear combination of CT images is also based on dense cyst (Silva et al. 2011).
the assumption that there are only two basis
materials in the material decomposition process
(Yu et al. 2011). 3.4 Electron Density and Effective
Because of the freedom to create VMIs at any Atomic Number Images
energy, they play multiple roles in DECT clinical
applications. First, on DECT scanners using fast Electron density ρe can be calculated from the
kV switching technology, no mixed images from mass density as following:
low- and high-energy scans are created. VMIs are
r
used as a substitute for the conventional single- re = ZN A (16)
energy images for routine diagnosis purpose. The A m
monoenergetic images can be created at an energy
level that approximately matches the mean energy where ρe is the electron density, ρ is mass density,
of either a 100  kV or 120  kV polychromatic Z is the atomic number, Am is the atom mass, and
energy beam, depending on user preference. NA is Avogadro’s number. The effective atomic
Second, VMIs are often created to provide opti- number of a composite material is defined as,
mal iodine signal in contrast enhanced DECT
exams. Because the X-ray attenuation of iodine Z eff = åwie Z in , i = 1,¼ N (17)
i
increases faster than tissue with the decrease of e
X-ray beam energy, VMIs at low energy, typically where wi is the function of the total number of
below 60 keV, show higher iodine contrast signal electrons of the ithmaterial, Zi is the atomic
10 J. Wang et al.

number of the ith material, N is the total number taking advantage of the fact that when using all
of materials in the composite, and n is a constant the radiation dose in the DECT acquisition, a min-
in the range of 3–4 (Heismann et  al. 2003). imum noise level can be reached either at certain
Effective atomic number of materials can be esti- monochromatic energy or by optimally mixing
mated with either material decomposition images from low- and high energies. Grant et al.
(Alvarez and Macovski 1976) or empirical cali- used a frequency-split technique to reduce the
bration based methods (Liu et al. 2009; Heismann image noise in VMIs (Grant et  al. 2014). By
et al. 2003; Landry et al. 2013). Attention must be decomposing the VMIs into low- and high-­
paid to the unit of the created images. For exam- frequency image sets, they created composite
ple, on one CT manufacturer’s DECT platform, low-energy (e.g., 40 keV) monoenergetic images
the electron density values are converted to the that maintain the high iodine contrast but have the
Hounsfield Unit scale, which means water has a image noise reduced to the minimal level typi-
value of 0 HU and air has a value of −1000 cally found at 70 keV. This approach is currently
HU.  The effective atom number is often pre- implemented on a commercial DECT platform
sented in units of 1 and shown as colored overlay (Mono+, syngo Dual-Energy, Siemens
image. Using electronic density and effective Healthcare). Leng et  al. developed a method to
atomic number information, DECT has been reduce the image noise in multi-energy CT by
evaluated to improve the accuracy of stopping exploiting information redundancies in the energy
power ratio calculation over single-energy CT for domain (Leng et al. 2011). Using this approach,
proton therapy treatment (Yang et  al. 2010). they showed a noise reduction of up to 59% and
Atomic number Zeff is converted to mean excita- CNR increase of up to 64% in VMIs in clinical
tion potential, which was used with electron den- DECT exams (Leng et al. 2015). Tao et al. used a
sity to calculate stopping power ratio (Bourque prior-knowledge-aware iterative denoising
et al. 2014). The DECT-­based approach is found approach to reduce noise and improve iodine
to be more accurate and less susceptible to small CNR of VMIs from DECT (Tao et al. 2019). By
perturbations of human tissues compositions in introducing total-variation regularization of the
SPR calculation than using single-energy CT difference image between the original VMIs and
images (Yang et al. 2010). the low-noise mixed images of low and high
energy, the spatial and spectral data redundancy
are both exploited. With this approach, besides
4 Image Quality preserving the image resolution and noise texture
and Quantitative Accuracy of original VMIs, the denoised images showed a
of Synthetic DECT Images 1.8-fold increase in iodine CNR compared to the
VMIs produced by the commercial dual-energy
Material-specific and virtual monoenergetic processing application (Mono+).
images created from DECT are susceptible to Recently, deep learning has been increasingly
noise because of the noise amplification in the used in medical imaging reconstruction (Wang
material decomposition process (Alvarez and et al. 2020), image noise reduction in CT (Chen
Seppi 1979). The increase of noise in virtual et al. 2017; Yang et al. 2018; Solomon et al. 2020;
monoenergetic images with decreasing energy Missert et  al. 2020), and CT dose optimization
may diminish the advantage of enhanced iodine (McCollough and Leng 2020). In DECT applica-
contrast at low monochromatic energy (Alvarez tions, Zhang et al. developed a butterfly convolu-
and Seppi 1979). On most DECT platforms, con- tional neural network (CNN) to perform image
ventional iterative reconstruction, which is rou- domain DECT material decomposition and
tinely used on single-energy CT images, can also showed superior noise reduction in deep learning
be applied to synthetic DECT images. generated material images over images from
A lot of effort has been dedicated to reducing direct matrix inversion and iterative decomposi-
image noise in DECT material decomposition by tion method (Zhang et  al. 2019). Poirot et  al.
Material Decomposition and Post-processing: History and Basic Principles 11

incorporated the conventional VNC images as a nostic images with better image quality, and
function of the corresponding CT numbers at low potentially reducing radiation dose to patients. A
and high energy into a CNN to reduce image good understanding on the advantages and limi-
noise of VNC images. The deep learning tations of the synthetic images is essential for a
approach showed lower noise and higher similar- successful adoption of DECT into clinical
ity to true non-contrast images compared to the practice.
conventional VNC images (Poirot et  al. 2019).
Gong et  al. proposed an Incept-net CNN archi- Compliance with Ethical Standards
tecture that utilizes multiresolution features of
local image structure and improves the robust- Disclosure of Interests  CHM is the principal investiga-
ness against local noise and artifacts while pre- tor of a research grant to Mayo Clinic from Siemens
Healthcare.
serving the structural details. They showed
improved accuracy of material quantification and
reduced image noise and artifact in comparison
to conventional least-square-based, total-­References
variation regularized, and U-net-based material Albrecht MH, Vogl TJ, Martin SS et al (2019) Review of
decomposition methods (Gong et al. 2020). clinical applications for virtual Monoenergetic dual-­
When quantitative information is required energy CT.  Radiology 293(2):260–271. https://doi.
from material-specific images, it is important to org/10.1148/radiol.2019182297
Alvarez RE, Macovski A (1976) Energy-selective
understand the impact of the assumption used in reconstructions in X-ray computerized tomogra-
material decomposition (McCollough et  al. phy. Phys Med Biol 21(5):733–744. https://doi.
2020). For example, bone will appear in iodine org/10.1088/0031-­9155/21/5/002
maps from the iodine-water based material Alvarez R, Seppi E (1979) A comparison of noise and
dose in conventional and energy selective computed
decomposition because the attenuation of bone is tomography. IEEE Trans Nucl Sci 26(2):2853–2856.
decomposed as a combination of iodine and https://doi.org/10.1109/TNS.1979.4330549
water and the portion that is assigned to iodine Behrendt FF, Schmidt B, Plumhans C et al (2009) Image
will be shown as positive signal on the iodine fusion in dual energy computed tomography: effect
on contrast enhancement, signal-to-noise ratio and
map. Similarly, the pixel values of fat on a VNC image quality in computed tomography angiography.
image might be different from the fat on a true Investig Radiol 44(1):1–6. https://doi.org/10.1097/
non-contrast image because the iodine-water pair RLI.0b013e31818c3d4b
is assumed in creation of the VNC images. Even Bongartz T, Glazebrook KN, Kavros SJ et  al (2015)
Dual-energy CT for the diagnosis of gout: an
with accurate models for material decomposi- accuracy and diagnostic yield study. Ann Rheum
tion, material quantification accuracy and repro- Dis 74(6):1072–1077. https://doi.org/10.1136/
ducibility in DECT is also impacted by other annrheumdis-­2013-­205095
factors, such as radiation dose, patient size, and Bourque AE, Carrier JF, Bouchard H (2014) A stoichio-
metric calibration method for dual energy computed
manufacturer-­dependent DECT acquisition tech- tomography. Phys Med Biol 59(8):2059–2088. https://
niques (Jacobsen et al. 2018; Goodsitt et al. 2011; doi.org/10.1088/0031-­9155/59/8/2059
Foley et al. 2016; Pelgrim et al. 2017; Nute et al. Bushberg JT, Seibert AJ, Leidholdt EM Jr, Boone JM
2018). (2012) The essential physics of the medical imaging.
Lippincott Williams & Wilkins
Cha J, Kim HJ, Kim ST, Kim YK, Kim HY, Park
GM (2017) Dual-energy CT with virtual mono-
5 Conclusion chromatic images and metal artifact reduction
software for reducing metallic dental artifacts.
Acta Radiol 58(11):1312–1319. https://doi.
With the advancement in emerging detector org/10.1177/0284185117692174
technology and novel material decomposition Chen H, Zhang Y, Kalra MK et al (2017) Low-dose CT
methods, DECT and multi-energy CT are play- with a residual encoder-decoder convolutional neu-
ing an increasing role in clinical CT imaging by ral network. IEEE Trans Med Imaging 36(12):2524–
2535. https://doi.org/10.1109/TMI.2017.2715284
adding functional information, providing diag-
12 J. Wang et al.

Ferda J, Novák M, Mírka H et al (2009) The assessment Radiology 257(1):205–211. https://doi.org/10.1148/
of intracranial bleeding with virtual unenhanced imag- radiol.10091806
ing by means of dual-energy CT angiography. Eur Heismann BJ, Leppert J, Stierstorfer K (2003) Density
Radiol 19(10):2518–2522. https://doi.org/10.1007/ and atomic number measurements with spectral x-ray
s00330-­009-­1495-­2 attenuation method. J Appl Phys 94(3):2073–2079.
Fischer MA, Gnannt R, Raptis D et al (2011) Quantification https://doi.org/10.1063/1.1586963
of liver fat in the presence of iron and iodine: an ex-­ Holmes DR 3rd, Fletcher JG, Apel A et  al (2008)
vivo dual-energy CT study. Investig Radiol 46(6):351– Evaluation of non-linear blending in dual-energy
358. https://doi.org/10.1097/RLI.0b013e31820e1486 computed tomography. Eur J Radiol 68(3):409–413.
Flohr TG, McCollough CH, Bruder H et al (2006) First https://doi.org/10.1016/j.ejrad.2008.09.017
performance evaluation of a dual-source CT (DSCT) Hounsfield GN (1973) Computerized transverse axial
system. Eur Radiol 16(2):256–268. https://doi. scanning (tomography). 1. Description of sys-
org/10.1007/s00330-­005-­2919-­2 tem. Br J Radiol 46(552):1016–1022. https://doi.
Foley WD, Shuman WP, Siegel MJ et  al (2016) White org/10.1259/0007-­1285-­46-­552-­1016
paper of the society of computed body tomography Hu R, Besheli LD, Young J et al (2016) Dual-energy head
and magnetic resonance on dual-energy CT, part 2: CT enables accurate distinction of Intraparenchymal
radiation dose and iodine sensitivity. J Comput Assist hemorrhage from calcification in emergency depart-
Tomogr 40(6):846–850. https://doi.org/10.1097/ ment patients. Radiology 280(1):177–183. https://doi.
rct.0000000000000539 org/10.1148/radiol.2015150877
Fulwadhva UP, Wortman JR, Sodickson AD (2016) Use Jacobsen MC, Schellingerhout D, Wood CA et al (2018)
of dual-energy CT and iodine maps in evaluation of Intermanufacturer comparison of dual-energy CT
bowel disease. Radiographics 36(2):393–406. https:// iodine quantification and monochromatic attenuation:
doi.org/10.1148/rg.2016150151 a phantom study. Radiology 287(1):224–234. https://
Gong H, Tao S, Rajendran K, Zhou W, McCollough CH, doi.org/10.1148/radiol.2017170896
Leng S (2020) Deep-learning-based direct inversion Johnson TR, Krauss B, Sedlmair M et al (2007) Material
for material decomposition. Med Phys. https://doi. differentiation by dual energy CT: initial experience.
org/10.1002/mp.14523 Eur Radiol 17(6):1510–1517. https://doi.org/10.1007/
Goodsitt MM, Christodoulou EG, Larson SC (2011) s00330-­006-­0517-­6
Accuracies of the synthesized monochromatic CT Johnson TR, Himsl I, Hellerhoff K et  al (2013) Dual-­
numbers and effective atomic numbers obtained energy CT for the evaluation of silicone breast
with a rapid kVp switching dual energy CT scan- implants. Eur Radiol 23(4):991–996. https://doi.
ner. Med Phys 38(4):2222–2232. https://doi. org/10.1007/s00330-­012-­2667-­z
org/10.1118/1.3567509 Kalender WA, Perman WH, Vetter JR, Klotz E (1986)
Grant KL, Flohr TG, Krauss B, Sedlmair M, Thomas C, Evaluation of a prototype dual-energy computed
Schmidt B (2014) Assessment of an advanced image-­ tomographic apparatus. I. Phantom studies. Med Phys
based technique to calculate virtual monoenergetic 13(3):334–339. https://doi.org/10.1118/1.595958
computed tomographic images from a dual-energy Kelcz F, Joseph PM, Hilal SK (1979) Noise considerations
examination to improve contrast-to-noise ratio in in dual energy CT scanning. Med Phys 6(5):418–425.
examinations using iodinated contrast media. Investig https://doi.org/10.1118/1.594520
Radiol 49(9):586–592. https://doi.org/10.1097/ Kong X, Sheng HX, Lu GM et  al (2014) Xenon-­
rli.0000000000000060 enhanced dual-energy CT lung ventilation imag-
Graser A, Johnson TR, Bader M et  al (2008) Dual ing: techniques and clinical applications. Am J
energy CT characterization of urinary calculi: Roentgenol 202(2):309–317. https://doi.org/10.2214/
initial in  vitro and clinical experience. Investig AJR.13.11191
Radiol 43(2):112–119. https://doi.org/10.1097/ Landry G, Seco J, Gaudreault M, Verhaegen F
RLI.0b013e318157a144 (2013) Deriving effective atomic numbers from
Graser A, Johnson TR, Hecht EM et al (2009) Dual-energy DECT based on a parameterization of the ratio
CT in patients suspected of having renal masses: can of high and low linear attenuation coefficients.
virtual nonenhanced images replace true nonenhanced Phys Med Biol 58(19):6851–6866. https://doi.
images? Radiology 252(2):433–440. https://doi. org/10.1088/0031-­9155/58/19/6851
org/10.1148/radiol.2522080557 Lee SH, Lee JM, Kim KW et  al (2011) Dual-energy
Guggenberger R, Winklhofer S, Osterhoff G et al (2012) computed tomography to assess tumor response to
Metallic artefact reduction with monoenergetic dual-­ hepatic radiofrequency ablation: potential diagnostic
energy CT: systematic ex vivo evaluation of posterior value of virtual noncontrast images and iodine maps.
spinal fusion implants from various vendors and dif- Investig Radiol 46(2):77–84. https://doi.org/10.1097/
ferent spine levels. Eur Radiol 22(11):2357–2364. RLI.0b013e3181f23fcd
https://doi.org/10.1007/s00330-­012-­2501-­7 Lehmann LA, Alvarez RE, Macovski A et  al (1981)
Gupta R, Phan CM, Leidecker C et al (2010) Evaluation Generalized image combinations in dual KVP digi-
of dual-energy CT for differentiating intracerebral tal radiography. Med Phys 8(5):659–667. https://doi.
hemorrhage from iodinated contrast material staining. org/10.1118/1.595025
Material Decomposition and Post-processing: History and Basic Principles 13

Leng S, Yu L, Wang J, Fletcher JG, Mistretta CA, Niu T, Dong X, Petrongolo M, Zhu L (2014)
McCollough CH (2011) Noise reduction in spectral Iterative image-­ domain decomposition for dual-
CT: reducing dose and breaking the trade-off between energy CT.  Med Phys 41(4):041901. https://doi.
image noise and energy bin selection. Med Phys org/10.1118/1.4866386
38(9):4946–4957. https://doi.org/10.1118/1.3609097 Nute JL, Jacobsen MC, Stefan W, Wei W, Cody DD (2018)
Leng S, Yu L, Fletcher JG, McCollough CH (2015) Development of a dual-energy computed tomography
Maximizing iodine contrast-to-noise ratios in abdomi- quality control program: characterization of scanner
nal CT imaging through use of energy domain response and definition of relevant parameters for a
noise reduction and virtual Monoenergetic dual-­ fast-kVp switching dual-energy computed tomogra-
energy CT.  Radiology 276(2):562–570. https://doi. phy system. Med Phys 45(4):1444–1458. https://doi.
org/10.1148/radiol.2015140857 org/10.1002/mp.12812
Li M, Zheng X, Li J et  al (2012) Dual-energy com- Pelgrim GJ, van Hamersvelt RW, Willemink MJ et  al
puted tomography imaging of thyroid nodule (2017) Accuracy of iodine quantification using dual
specimens: comparison with pathologic findings. energy CT in latest generation dual source and dual
Investig Radiol 47(1):58–64. https://doi.org/10.1097/ layer CT.  Eur Radiol 27(9):3904–3912. https://doi.
RLI.0b013e318229fef3 org/10.1007/s00330-­017-­4752-­9
Li Z, Leng S, Yu L, Yu Z, McCollough CH (2015) Poirot MG, Bergmans RHJ, Thomson BR et  al
Image-­ based material decomposition with a (2019) Physics-informed deep learning for dual-­
general volume constraint for photon-counting energy computed tomography image process-
CT.  Proc SPIE Int Soc Opt Eng 9412. https://doi. ing. Sci Rep 9(1):17709. https://doi.org/10.1038/
org/10.1117/12.2082069 s41598-­019-­54176-­0
Liu X, Yu L, Primak AN, McCollough CH (2009) Primak AN, Fletcher JG, Vrtiska TJ et  al (2007)
Quantitative imaging of element composition and Noninvasive differentiation of uric acid versus non-­
mass fraction using dual-energy CT: three-material uric acid kidney stones using dual-energy CT.  Acad
decomposition. Med Phys 36(5):1602–1609. https:// Radiol 14(12):1441–1447. https://doi.org/10.1016/j.
doi.org/10.1118/1.3097632 acra.2007.09.016
Luo XF, Xie XQ, Cheng S et al (2015) Dual-energy CT Roessl E, Proksa R (2007) K-edge imaging in x-ray com-
for patients suspected of having liver iron overload: puted tomography using multi-bin photon counting
can virtual iron content imaging accurately quantify detectors. Phys Med Biol 52(15):4679–4696. https://
liver iron content? Radiology 277(1):95–103. https:// doi.org/10.1088/0031-­9155/52/15/020
doi.org/10.1148/radiol.2015141856 Schlomka JP, Roessl E, Dorscheid R et  al (2008)
Macovski A, Alvarez RE, Chan JL, Stonestrom Experimental feasibility of multi-energy photon-­
JP, Zatz LM (1976) Energy dependent recon- counting K-edge imaging in pre-clinical computed
struction in X-ray computerized tomography. tomography. Phys Med Biol 53(15):4031–4047.
Comput Biol Med 6(4):325–336. https://doi. https://doi.org/10.1088/0031-­9155/53/15/002
org/10.1016/0010-­4825(76)90069-­x Sellerer T, Noël PB, Patino M et  al (2018) Dual-energy
McCollough CH, Leng S (2020) Use of artificial intel- CT: a phantom comparison of different platforms for
ligence in computed tomography dose optimisa- abdominal imaging. Eur Radiol 28(7):2745–2755.
tion. Ann ICRP 49(1_suppl):113–125. https://doi. https://doi.org/10.1007/s00330-­017-­5238-­5
org/10.1177/0146645320940827 Silva AC, Morse BG, Hara AK, Paden RG, Hongo
McCollough CH, Boedeker K, Cody D et  al (2020) N, Pavlicek W (2011) Dual-energy (spectral) CT:
Principles and applications of multienergy CT: report applications in abdominal imaging. Radiographics
of AAPM task group 291. Med Phys 47(7):e881– 31(4):1031–1046. https://doi.org/10.1148/
e912. https://doi.org/10.1002/mp.14157 rg.314105159. discussion 1047-50
Mileto A, Marin D, Alfaro-Cordoba M et al (2014) Iodine Solomon J, Lyu P, Marin D, Samei E (2020) Noise and
quantification to distinguish clear cell from papil- spatial resolution properties of a commercially avail-
lary renal cell carcinoma at dual-energy multidetec- able deep learning-based CT reconstruction algorithm.
tor CT: a multireader diagnostic performance study. Med Phys 47(9):3961–3971. https://doi.org/10.1002/
Radiology 273(3):813–820. https://doi.org/10.1148/ mp.14319
radiol.14140171 Tao S, Rajendran K, Zhou W, Fletcher JG, McCollough
Missert AD, Yu L, Leng S, Fletcher JG, McCollough CH, Leng S (2019) Improving iodine contrast to
CH (2020) Synthesizing images from multiple ker- noise ratio using virtual monoenergetic imaging and
nels using a deep convolutional neural network. prior-knowledge-aware iterative denoising (mono-­
Med Phys 47(2):422–430. https://doi.org/10.1002/ PKAID). Phys Med Biol 64(10):105014. https://doi.
mp.13918 org/10.1088/1361-­6560/ab17fa
Muenzel D, Bar-Ness D, Roessl E et  al (2017) Taylor RE, Mager P, Yu NC, Katz DP, Brady JR, Gupta N
Spectral photon-­ counting CT: initial experience (2019) Iodine quantification and detectability thresh-
with dual-contrast agent K-edge colonography. olds among major dual-energy CT platforms. Br J
Radiology 283(3):723–728. https://doi.org/10.1148/ Radiol 92(1104):20190530. https://doi.org/10.1259/
radiol.2016160890 bjr.20190530
14 J. Wang et al.

Wang AS, Pelc NJ (2011) Synthetic CT: simulating low ing. Med Phys 33(11):4115–4129. https://doi.
dose single and dual energy protocols from a dual org/10.1118/1.2349688
energy scan. Med Phys 38(10):5551–5562. https://doi. Yang M, Virshup G, Clayton J, Zhu XR, Mohan R, Dong
org/10.1118/1.3633895 L (2010) Theoretical variance analysis of single- and
Wang CK, Tsai JM, Chuang MT, Wang MT, Huang KY, dual-energy computed tomography methods for cal-
Lin RM (2013) Bone marrow edema in vertebral com- culating proton stopping power ratios of biological
pression fractures: detection with d­ual-­energy CT. tissues. Phys Med Biol 55(5):1343–1362. https://doi.
Radiology 269(2):525–533. https://doi.org/10.1148/ org/10.1088/0031-­9155/55/5/006
radiol.13122577 Yang Q, Yan P, Zhang Y et al (2018) Low-dose CT image
Wang G, Ye JC, De Man B (2020) Deep learning for Denoising using a generative adversarial network
tomographic image reconstruction. Nat Mach with Wasserstein distance and perceptual loss. IEEE
Intell 2(12):737–748. https://doi.org/10.1038/ Trans Med Imaging 37(6):1348–1357. https://doi.
s42256-­020-­00273-­z org/10.1109/TMI.2018.2827462
Werner S, Krauss B, Haberland U et  al (2019) Yu L, Primak AN, Liu X, McCollough CH (2009) Image
Dual-­ energy CT for liver iron quantification quality optimization and evaluation of linearly mixed
in patients with haematological disorders. Eur images in dual-source, dual-energy CT.  Med Phys
Radiol 29(6):2868–2877. https://doi.org/10.1007/ 36(3):1019–1024. https://doi.org/10.1118/1.3077921
s00330-­018-­5785-­4 Yu L, Christner JA, Leng S, Wang J, Fletcher JG,
Willemink MJ, Persson M, Pourmorteza A, Pelc McCollough CH (2011) Virtual monochromatic imag-
NJ, Fleischmann D (2018) Photon-counting ing in dual-source dual-energy CT: radiation dose and
CT: technical principles and clinical prospects. image quality. Med Phys 38(12):6371–6379. https://
Radiology 289(2):293–312. https://doi.org/10.1148/ doi.org/10.1118/1.3658568
radiol.2018172656 Zhang W, Zhang H, Wang L et al (2019) Image domain
Williamson JF, Li S, Devic S, Whiting BR, Lerma FA dual material decomposition for dual-energy CT
(2006) On two-parameter models of photon cross using butterfly network. Med Phys 46(5):2037–2051.
sections: application to dual-energy CT imag- https://doi.org/10.1002/mp.13489
 Dual-Energy: The Siemens
Approach

Bernhard Schmidt and Thomas Flohr

Contents
1 General Aspects and Workflow   15
2 Dual Source-Based Dual-Energy   19
3 Twin Spiral Dual-Energy   22
4 TwinBeam Dual-Energy   23
5 Conclusions   26
References   26

1 General Aspects the main clinical benefit of this new technical capa-
and Workflow bility was a significant improvement in the accu-
racy of bone mineral density quantification.
The first medical CT scanner capable of DE scan- Although several clinical studies demonstrated the
ning by means of fast kV-switching was the advantage of DE CT over single energy CT espe-
SOMATOM DRH introduced by Siemens in 1987 cially in the presence of fat—elimination of the so-
(Kalender 1987). It allowed for dual-energy acqui- called “fat error”—(Laval-Jeantet 1986; Genant
sitions in topogram and axial scan mode; however, 1977; Vetter 1986), the fast kV-switching technique
to acquire DE data was abandoned in subsequent
CT systems. Technical limitations led to compro-
B. Schmidt (*) mised scan data acquisition, degradation in image
Siemens Healthcare GmbH, Computed Tomography,
Forchheim, Germany quality, and increased radiation dose to the patient
since attenuation-based tube current modulation
Friedrich-Alexander-Universität Erlangen, Institute
for Medical Physics, Erlangen, Germany and necessary adaptations of the tube current
e-mail: [email protected] between the kV-switches was not feasible due to
T. Flohr basic physical limitations of the technology. On top
Siemens Healthcare GmbH, Computed Tomography, of this new technical opportunities such as DEXA
Forchheim, Germany (dual-emission X-ray absorptiometry) became
Eberhard-Karls-Universität Tübingen, Institute of available and limited the need for DE CT bone den-
Diagnostic and Interventional Radiology, sitometry. Other c­linically relevant applications,
Tübingen, Germany however, were not within reach at that time.
e-mail: [email protected]

© The Author(s), under exclusive license to Springer Nature Switzerland AG 2022 15

H. Alkadhi et al. (eds.), Spectral Imaging, Medical Radiology Diagnostic Imaging,
https://doi.org/10.1007/978-3-030-96285-2_2
16 B. Schmidt and T. Flohr

In 2006 DE CT was reintroduced with the 2008; Remy-Jardin et  al. 2014; Apfaltrer et  al.
advent of dual source CT systems (Flohr et  al. 2014), as well as post-traumatic bone bruises or
2006) which enabled the acquisition of DE CT bone-marrow infiltration of the spine in patients
data without the significant limitations of the pre- with multiple myeloma (Pache et  al. 2010;
vious technology, rather a much wider clinical Thomas et al. 2015). DE CT has been applied in
application spectrum. Since then, the utilization the characterization of perfusion defects in the
of DE CT has been growing steadily, prompted in myocardium (Ruzsics et  al. 2009; Vliegenthart
part by the introduction of newer generations of et al. 2012), as well as the iron uptake of the liver
dual source CT systems with further improved in patients with liver iron overload (Luo et  al.
dual-energy performance. A few years later, the 2015; Werner et  al. 2019). Maps of effective
portfolio for DE data acquisition techniques was atomic numbers obtained with DE CT may be
extended to non-dual source CT systems with the used to improve radiation treatment planning,
introduction of subsequent spiral scanning at dif- particularly in proton therapy. The applications
ferent X-ray tube voltages (Twin Spiral; Siemens listed above are an incomplete list of clinically
Healthineers, Germany). This—in a second relevant DE applications which are available on
development step—was further improved upon Siemens CT machines, and are either already
by adding just to the high voltage spiral portion being applied clinically or currently evaluated.
of the Twin Spiral acquisition an additional pre-­ Besides dedicated clinical DE applications
filtration, already being well established for dual like the ones mentioned above, acquired data also
source CT systems. Hereby a dual-energy separa- can be used for calculation of virtual mono-­
tion comparable to second and third generation energetic images. Similar to raw data-based
of dual source CT was accomplished. To further approaches back in 1986 (Kalender et al. 1986),
substantially reduce the time delay between the high- and low-energy images can be used for
high and low voltage data acquisition, a few years decomposition into materials differing in photo-
later TwinBeam (Siemens Healthineers, electric and Compton characteristics, such as
Germany) was introduced to single source CT water and iodine (Yu et al. 2012). The concentra-
systems. In this system a split filter is added to tions of both materials in each image pixel are
the X-ray tube collimator box, allowing examina- calculated by means of an image-based material
tions even in case of high contrast dynamics as in decomposition algorithm that requires calibra-
a contrast-enhanced scan in the arterial phase. tion measurements for the attenuation of iodine
From a clinical perspective, DE CT imaging in phantoms of different diameter for the differ-
proves far more valuable than standard CT scans ent X-ray spectra. The concentrations are subse-
because, unlike them, it provides functional quently multiplied with predicted CT numbers
information on top of the mere anatomical visu- per concentration at the desired pseudo mono-
alization (Schneider et al. 2014). Kidney stones chromatic energy (derived from NIST tables) and
can be differentiated (Primak et al. 2007; Scheffel summed up to form the final mono-energetic
et al. 2007; Stolzmann et al. 2010), the differen- images. Pseudo mono-energetic images derived
tial diagnosis of gout is possible (Mallinson et al. from image-based material decomposition have
2016), and lesions can be characterized by quan- similar applications as raw data-based images.
tifying their iodine uptake (Graser et al. 2010). In Metal artifacts are reduced at higher energies/
oncology, structures more responsive to treat- keV (Mangold et al. 2014), the contrast of iodine
ment may potentially be identified early during and bone increases at lower energies. Like in raw
treatment by reduced iodine uptake rather than by data-based techniques, image noise is amplified
mere lesion size measurements (Apfaltrer et  al. at energy levels far away from the mean energy
2012; Uhrig et  al. 2013; Agrawal et  al. 2014; of the mixed images (~ 70  keV). To avoid this
Knobloch et  al. 2014). Perfusion defects in the undesirable increase in noise, which not only
lung parenchyma in patients with pulmonary limits the ultimate clinical benefit of mono-­
embolism are visible in DE scans (Pontana et al. energetic images but also limits the gains from
Dual-Energy: The Siemens Approach 17

improved iodine contrast-to-noise ratio at lower than to perform low kV-scans, which is today the
keV, a novel algorithm for the computation of recommended method to improve iodine
pseudo mono-energetic images was recently CNR.  The image-based Mono+ approach is
introduced (Mono+, Siemens Healthcare, available for all DE data from Siemens CT sys-
Forchheim, Germany) that efficiently reduces tems, and levels of the pseudo mono-energetic
image noise in pseudo mono-energetic images at images can be selected between 40  keV and
low and high keV (Grant et al. 2014). Using this 190 keV in steps of 1 keV. Figure 2 shows a clini-
approach, images at the target keV and images at cal example for illustration.
optimal keV from a noise perspective (typically, Although clinical DE applications and the
minimum image noise is obtained at approxi- possibility for improved image quality and better
mately 70  keV) are computed. By means of a contrast-to-noise ratio are well perceived in the
frequency-split technique, both the images at the medical community, one of the remaining chal-
target keV and the images with minimum image lenges of dual-energy imaging is workflow and
noise level are broken down into two sets of sub-­ workflow optimizations. In general, DE process-
images. The first set contains only lower spatial ing on Siemens CT systems is mainly based on
frequencies and thus most of the object informa- the high and low voltage image stacks, which can
tion, the second one contains the remaining high be flexibly stored in PACS, or loaded and pro-
spatial frequencies and subsequently mostly cessed retrospectively by the various DE applica-
image noise. Finally, the lower spatial frequency tion classes provided in the Siemens syngo
stack at the target keV is combined with the high Dual-Energy application—irrespective of which
spatial frequency stack at optimal keV from a scanner of the Siemens Healthineers CT scanner
noise perspective to combine the benefits of both portfolio is used for data acquisition. In addition,
image stacks, see Fig. 1. automated dual-energy processing is offered to
The Mono+ technique can be used to signifi- users who prefer PACS-focused reading over
cantly increase the iodine CNR in CT angio- interactive processing and expert viewing in the
graphic studies by computing pseudo syngo Dual-Energy application. User selected
mono-energetic images at low keV (Albrecht combinations of DE results are processed and
et al. 2019). It has been shown (Grant et al. 2014) transferred to the user-owned PACS fully auto-
that it may be more efficient to perform DE scans matically, thus being available whenever and
and compute pseudo mono-energetic images at wherever the user may need them, even for retro-
40  keV using Mono+ to optimize iodine CNR spective analysis. The most established mecha-

40 keV
70 keV
high contrast medium contrast
high noise low noise

40 keV +

Fig. 1  Schematic illustration of the concept used to cal- images showing lower iodine contrast and lower noise
culate Mono+ images. In this example, the CNR of iodine (70  keV) to obtain improved CNR at low keVs (Grant
versus soft tissue is enhanced. Images with high iodine et al. 2014)
contrast and high image noise (40  keV) are mixed with
18 B. Schmidt and T. Flohr

a c d

Fig. 2  Left: DE images acquired on a third-generation acquisition (a). Right: Pseudo mono-energetic image
DSCT scanner through the upper abdomen using the kV-­ using the Mono+ technique at 50  keV (d) compared to
combination 80 kV/150 kV with tin filter, and only 30 cc 80 kV (c). Note the significantly increased iodine CNR of
IV contrast media. Improved contrast and better lesion vascular structures with Mono+. (Courtesy of NYU
discernability in case of Mono+ at low keVs (b) compared Medical Center, Department of Radiology, New  York,
to a mixed image, corresponding to a standard 120  kV USA)

nisms for this workflow are “Rapid Results It can be shown that the thin absorber model
Technology” or “Recon&Go Inline Results.” breaks down for iodine samples with more than
Here the user has the ability to define at the scan- 5000  HU  cm in water based on 120  kV, which
ner which processing steps should happen auto- corresponds to the clinical situation of an object
matically, such as the generation of oblique with 200  HU iodine enhancement and 25  cm
Mono+ images or classified VRT images from a thickness. In almost all clinically relevant situa-
DE renal calculi application. Respective data are tions, the thin absorber model remains valid.
then processed automatically, and the results sent Exceptions are scan scenarios where extremely
to the PACS system for further reading allowing high iodine concentrations may be present, such
for an efficient and integrated DE workflow (see as CT urographic scans.
Fig. 3). In addition, the thin absorber model is based
One basic assumption for image-based mate- on the concept of an effective spectrum: the mea-
rial decomposition—used in the Siemens sured absorption with a polychromatic X-ray
approach—is the validity of the thin absorber spectrum is assumed to be independent of the
model. If we use water and iodine, for example, spatial distribution of the traversed materials
as the basis materials for image-based dual-­ along the beam. In practice, this means that nei-
energy evaluation, the maximum X-ray attenua- ther the CT-value of water nor the CT-value of a
tion coefficient μΙ(E) and the maximum thickness small iodine sample depends on its position
dI of the iodine along any measured ray path are within the scanned object. The scanner must
expected to be so small that it is valid to assume therefore be equipped with a bowtie filter of
a linear contribution of the additional non-water-­ ­sufficient beam hardening and the approximately
like attenuation μΙ(E)· dI to the total attenuation. cylindrical patient cross-section has to be cen-
Dual-Energy: The Siemens Approach 19

a CT scanner syngo.via PACS System

Tech performs syngo.via Radiologist uses results for

standard thorax scan pre-processes data diagnosis and distribution in PACS

Fig. 3  Optimized workflow on Siemens CT systems: further reading. Examples for respective results are shown
Acquired DE image data are automatically labeled, pro- in (b), where in addition to the visualization of DE results
cessed by syngo via in the background and sent to PACS the gout volume is automatically calculated (in this exam-
using Rapid Results Technology without any additional ple 5.02 cm3). Beside material and Mono+ images, ana-
user interactions (a). Alternatively, on selected CT models tomically oriented DE VRTs with classified results (left
a similar automatic advanced processing is possible and middle) and fused images (right: overlay of mixed
directly at the scanner, where yet again results are pro- and iodine) can be generated
cessed self-operationally and sent directly to PACS for

tered within the SFOV.  In a real-world setting, 2 Dual Source-Based

electronics noise, scanner calibration, stability of Dual-Energy
emitted spectra, cone beam effects, and scattered
radiation can have a larger impact on the obtained A dual source CT is a CT system with two mea-
results than the analysis method. surements systems, such as two X-ray tubes and
As stated in the beginning of this article, the corresponding detectors. Both measurement
Siemens is processing—irrespectively of the CT systems acquire scan data simultaneously at the
system type used for acquisition—based on low same anatomical level of the patient (same
and high voltage image stacks. This makes stor- z-position). In 2006, the first dual source CT
ing of images, transferring and retrospective (DSCT) was commercially introduced by
application of various DE classes easy. On the Siemens, the SOMATOM Definition (Siemens
data acquisition side Siemens offers various solu- Healthcare GmbH, Forchheim, Germany), see
tions—depending on the technical capabilities of Fig. 4.
the respective CT system. In the following para- The two acquisition systems A and B are
graphs, the different concepts and technical real- mounted onto the rotating gantry at an angular
izations are introduced and discussed. offset of 90° for the first-generation DSCT (Flohr
20 B. Schmidt and T. Flohr

a b c

Fig. 4  DSCT with two independent measurement sys- surement systems is 90°. (b) Second generation: To
tems. Left: Open gantry of a dual source CT system—red increase the SFOV of detector B, a larger system angle of
and green fans indicate the beams of the two X-ray tubes. 95° was chosen. With the third-generation DSCT (c), the
(a) First generation: The system angle between both mea- SFOV of detector B was further increased to 35.5 cm

et al. 2006), and at an angular offset of 95° for the restrictions in the choice of scan parameters such
second- and third-generation DSCT. Detector A as gantry rotation time. Use of anatomical tube
covers the full SFOV of 50  cm diameter, while current modulation allows for adaptation of the
detector B is restricted to a smaller FOV of 26 cm radiation dose to the patient’s anatomy. Mixed
(first generation), 33 cm (second generation), or images (a weighted average of low- and high-­
35.6  cm (third generation) as a consequence of energy images) and Mono+ are routinely avail-
space limitations on the gantry. The shortest gan- able, allowing dual-energy CT scans to be
try rotation times are 0.33  s (first generation), performed in routine clinical practice similar to
0.28 s (second generation), and 0.25 s (third gen- conventional imaging protocols, with dual-­
eration). DSCT systems provide significantly energy information available when needed. It
improved temporal resolution for cardio-thoracic should be noted that dual-energy imaging and the
imaging. The shortest data acquisition time for an ability for Mono+ as well as iodine quantification
image corresponds to a quarter of the gantry rota- being helpful, e.g., for the assessment of the per-
tion time (Flohr et al. 2006). Meanwhile, several fusion of the myocardium is also possible for
clinical studies have demonstrated the potential gated cardiac scans on dual source systems. An
of DSCT to accurately assess coronary artery ste- example is shown in Fig. 5.
nosis in patients with high and irregular heart Spectral separation, a key measure for DE per-
rates (Achenbach et al. 2006, Johnson et al. 2006, formance, can be improved by introducing addi-
Scheffel et al. 2006, Matt et al. 2007, Leber et al. tional pre-filtration into the high kV beam, e.g.,
2007, Ropers et al. 2007). by means of a filter that can be moved into the
Moreover, with a DSCT system, dual-energy beam when needed and moved out for standard
data can be acquired by simultaneously operating applications. The quality of DE CT examinations
both X-ray tubes at different kV settings, e.g., generally relies on the separation of the energy
80  kV and 140  kV (Flohr et  al. 2006; Johnson spectra. High spectral overlap and bad energy
et al. 2007). Scan parameters (e.g., tube current separation result in increased image noise in the
and potential) can be adjusted individually for base-material decomposition which in turn
both measurement systems, resulting in a bal- requires compensation by increased radiation
anced radiation dose distribution between the dose. The second-generation DSCT makes use of
low- and the high-energy scans. A wide range of an additional tin filter (Sn) with a thickness of
routine scan protocols is available, with no 0.4 mm to shift the mean energy of the 140 kV
Dual-Energy: The Siemens Approach 21

a b c

Fig. 5  ECG-triggered “step-and-shoot” DE cardiac CT acquired simultaneously, DE information can be used to

acquisition using a third-generation DSCT at optimize iodine contrast in coronary vessels based on
90  kV/150  Sn kV, rotation time 0.25  s: Based on the Mono+ imaging: 55 keV (a), 70 keV (b), and 110 keV (c).
acquired data, images with 66 ms temporal resolution can Courtesy of Medical University of South Carolina,
be reconstructed. In addition, since data form A and B are Charleston, USA

spectrum from 86  keV to 97  keV (after 20  cm 2015). The larger DE ratio results in better condi-
water), see Fig. 6. The mean energy of the 80 kV tioned equations for base-material differentiation
spectrum is 60 keV. The third-generation DSCT into, for example, water and iodine as base mate-
provides 150  kV X-ray tube voltage with more rials, and leads to less image noise in the material-­
aggressive tin pre-filtration (0.6 mm), shifting the specific images. Consequently, this enables DE
mean energy of the 150 kV spectrum to 107 keV, data acquisition at the radiation dose of typical
see also Fig.  6. The tin filter improves spectral medical CT examinations, without additional
separation between the low- and high-energy dose penalty, see, e.g., (Schenzle et  al. 2010;
spectra, narrows the high kV spectrum (which Bauer et al. 2011; Henzler et al. 2012).
results in better dose efficiency and less beam As a downside, DE evaluation with dual
hardening artifacts), and reduces the influence of source CT is restricted to the smaller central
cross-scattering. A relevant parameter to quantify SFOV of detector B. Raw data-based dual-energy
the performance of a DE CT acquisition tech- algorithms cannot be realized because high-­
nique with regard to energy separation and mate- energy and low-energy projections are not simul-
rial differentiation capability is the Dual-Energy taneously acquired at the same z-position.
(DE) ratio. The DE ratio of a material is defined Dual-energy algorithms are therefore image-­
as its CT number (in HU) at low kV divided by its based. Another challenge of dual source DE CT
CT number (in HU) at high kV (Krauss et  al. is cross-scattered radiation, i.e., scattered radia-
2015). Water has a DE ratio of 1, meaning its CT tion originating from tube A and detected by
number does not change in CT scans at different detector B, and vice versa, which has to be care-
kV settings. The DE ratio for iodine, a commonly fully corrected for to avoid distortions of CT
used base material for material decomposition in numbers by cupping or streaking artifacts. This
contrast-enhanced CT scans, increases from 1.9 can be done either by measurement of cross-­
to 2 at the standard 80  kV/140  kV X-ray tube scattered radiation or by model-based approaches
voltage combination to about 3.4 for (Petersilka et al. 2010). Figure 7 shows a clinical
80 kV/150 kV and 0.6 mm tin pre-filtration (mea- example of a DE CT scan acquired with a third-­
sured in a 20  cm water phantom, Krauss et  al. generation DSCT system.
22 B. Schmidt and T. Flohr

Fig. 6  Typical 80 kV

and 140 kV spectra
(after 20 cm water),
normalized to equal
areas under the curves
(top). 80 kV spectrum
and 140 kV spectrum
with additional 0.4 mm
tin pre-filtration (center),
and 80 kV spectrum and
150 kV spectrum with
additional 0.6 mm tin
pre-filtration (bottom).
Note the shift of the
mean energy of the
high-energy spectrum to
higher values (arrow)
and the reduced spectral
overlap. The yellow
marked areas indicate
the portion of the spectra
that measured redundant
information

3 Twin Spiral Dual-Energy with a 50  cm diameter full scan field of view
(SFOV). However, the disadvantage is that the
For non-dual source CT systems, the most time delay between the two scans presents a chal-
straightforward approach for acquiring dual-­ lenge for the evaluation of fast moving organs.
energy CT data are two subsequent CT scans of Furthermore, examinations with administration of
the same anatomical structure, one with low X-ray contrast agent are challenging, at least in early
tube voltage (80 kV), the other with high X-ray arterial phases when the contrast density changes
tube voltage (140  kV). The spectral separation rapidly between the two scans. The resulting
achieved with this approach is reasonably effec- CT-number changes will be misinterpreted by DE
tive. DE CT is feasible with standard CT systems material decomposition techniques.
Dual-Energy: The Siemens Approach 23

Fig. 7  Contrast-enhanced DE CT scan of a patient with occluding clot. Right: In addition to parenchymal infor-
acute pulmonary embolism acquired with a third-­ mation, DE-based color-coding of the vessel lumen. Non-­
generation DSCT scanner at 90 kV/150 kV with 0.6 mm iodinated vessels—matching with the perfusion
tin pre-filtration. Left: Mixed image overlayed by the defect—are highlighted in red. Courtesy of University
quantitative perfusion information of the lung paren- Hospital of Frankfurt, Germany
chyma, showing a v-like perfusion defect caused by an

A technical realization was first introduced tial improvement of spectral separation and better
with the SOMATOM Definition AS, SOMATOM DE performance (see Fig. 8). A clinical example
Edge (Siemens Healthcare GmbH, Forchheim, is shown in Fig. 9, where a voltage combination
Germany), where the DE acquisition relies on of 80 kV and 150 Sn kV had been used.
two automatically, workflow-wise completely
coupled spiral (helical) scans of the same body
region, the first performed at 80 kV and the sec- 4 TwinBeam Dual-Energy
ond at 140 kV. The approach to go for volumetric
acquisitions instead of a single axial scan is Recently, a new method was introduced to acquire
advantageous since mismatch due to motion can DE CT data with a single source CT system with-
be corrected by a respective volumetric out kV-switching, but with better temporal registra-
­registration approach. As in standard CT exami- tion than by performing two separate consecutive
nations, radiation dose to the patient can be opti- axial or spiral scans of the examination volume of
mized by anatomical tube current modulation interest. Two different pre-filters in the tube colli-
(Marin et  al. 2014), and iterative reconstruction mator housing are used to split the X-ray beam in
can be applied. Because of the small time delay the scan direction, called “TwinBeam” (e.g., on
between the two spiral scans, the use of this tech- SOMATOM Definition Edge, Siemens Healthcare
nique is indicated for non-dynamic examinations GmbH, Forchheim, Germany), see Fig. 10.
that do not require the administration of contrast The X-ray tube is operated at 120 kV tube volt-
agent, such as characterization of kidney stones, age. One half of the multi-slice detector in the scan
or the examination of tophaceous lesions in direction is illuminated by an X-ray beam pre-fil-
patients with gout, or for the calculation of tered with 0.6  mm tin; compared to the standard
pseudo mono-energetic images to reduce metal 120 kV spectrum, the mean energy of this pre-fil-
artifacts at a metal-specific high energy. tered spectrum is increased, see Fig. 10, right. The
With the introduction of the SOMATOM X. other half of the detector in the scan direction is pre-
Cite in 2019 and the SOMATOM X.ceed 2021, filtered with a thin gold filter; as a consequence of
Siemens introduced tin filtration that had already the K-edge of gold at 80.7 keV, the mean energy of
been well established in Dual Source CT systems this spectrum is decreased, see Fig. 10, right. The
to Twin Spiral scanning, allowing for a substan- total attenuation of the pre-filters is adjusted to bal-
24 B. Schmidt and T. Flohr

Fig. 8  DE performance can be measured quantitatively performance. Additional tin filtration of 0.7  mm intro-
in terms of iodine ratio (HU at low energy divided by HU duced with the SOMATOM X.cite and X.ceed, increases
at high energy): For the high kV beam, additional tin fil- dual-energy separation for Twin Spiral into the range of
tration was introduced for dual systems to improve DE second- and third-generation dual source CTs

a b

Fig. 9  Clinical example for dual-energy imaging tech- image (a). Iodine overlay image (b) derived from spectral
nique relying on two consecutive spiral scans (Twin information reveals that area of hyper-attenuation corre-
Spiral) at 80 kV and 150 Sn kV on SOMATOM X.cite. sponds to an area of diffused contrast material extravasa-
Post-thrombectomy follow-up CT: Hyperattenuating tion. Virtual non-contrast image excludes active bleeding.
intraparenchymal area visible in the conventional mixed Courtesy of University Hospital Zurich, Switzerland

ance the radiation dose of the low-energy and the X.cite and X.ceed, even allow for adaptation of tube
high-­energy beam. The CT system is operated in a voltage. Depending on the clinical need and the size
spiral (helical) scan mode at fast gantry rotation of the patient (attention information is derived from
speed (0.28 s) with a maximum spiral pitch of 0.5 the topogram), either 120 kV or 140 kV is selected
(referring to the full z-width of the detector). for the use with the split filter. The possibility for a
Moreover, recently introduced CT systems higher voltage is of advantage for two reasons:
equipped with a split filter, the SOMATOM go.Top, First, spectral separation is increased. Secondly,
Dual-Energy: The Siemens Approach 25

Fig. 10  Left: Principle of a DE acquisition technique that Right: The standard 120 kV spectrum is split into a low-­
uses a split filter, called “TwinBeam” (Siemens Healthcare energy spectrum after filtering with gold (Au, top), and a
GmbH, Forchheim, Germany), in the tube collimator high-energy spectrum after filtering with tin (Sn, bottom)
housing to split the X-ray beam in the scan direction.

a b c d

Fig. 11  Contrast-enhanced thorax-abdominal CT exami- spicuity of lesions is improved compared to 70 keV images
nation on a SOMATOM X.cite using Twin Spiral tech- (a). In addition, virtual non-contrast images (c) and
nique. From the acquired DE data, Mono+ images can be iodine—as overlay—images (d) can be extracted. The lat-
generated to increase iodine contrast. At lower energies of ter in particular allows for quantitative analysis of iodine
45 keV (b), delineation of anatomical structures and con- distribution. Courtesy of Erlangen University, Germany

140 kV allows for a higher tube output and enables reconstruction techniques. The data sets are tempo-
scanning of larger patients. Then, for the recon- rally registered, enabling DE CT scans with con-
struction of the TwinBeam data, each half of the trast agent even in the arterial phase. The downside
detector acquires a complete spiral data set because of this approach is that spectral separation is worse
of the lower spiral pitch, allowing for low- and high- than with approaches using two different kV-set-
energy images to be reconstructed at any z-position tings of the X-ray tube. Furthermore, a powerful
as an input into Mono+ and/ or image-based mate- X-ray tube is required because the pre-filtration
rial decomposition techniques. absorbs a considerable portion of the X-ray flux, in
This technology provides DE data in the full turn limiting the use of this technique to non-obese
SFOV of 50 cm diameter. The radiation dose to the patients. Because of the maximum spiral pitch of
patient can be optimized by means of anatomical 0.5, maximum volume coverage speed is limited.
tube current modulation or risk organ-­dependent Figure 11 shows a clinical example acquired with
tube current modulation, in addition to iterative the split filter technique.
26 B. Schmidt and T. Flohr

5 Conclusions Apfaltrer P, Meyer M, Meier C, Henzler T, Barraza JM

Jr, Dinter DJ, Hohenberger P, Schoepf UJ, Schoenberg
SO, Fink C (2012) Contrast-enhanced dual-energy CT
Since 2006, with the introduction of DE on Dual of gastrointestinal stromal tumors: is iodine-related
Source CT systems, DE imaging has experienced attenuation a potential indicator of tumor response?
a renaissance. Its clinical application spectrum is Invest Radiol 47(1):65–70
Apfaltrer P, Sudarski S, Schneider D et  al (2014) Value
growing, and different technical solutions for of monoenergetic low-kV dual energy CT datasets for
data acquisition are available. With ongoing tech- improved image quality of CT pulmonary angiogra-
nological progress, such as improved spectral phy. Eur J Radiol 83:322–328
separation by means of dedicated pre-filtration of Bauer RW, Kramer S, Renker M et  al (2011) Dose and
image quality at CT pulmonary angiography: compar-
the high-energy beam, successful DE applica- ison of first and second generation dual energy CT and
tions have been demonstrated without increased 64-slice CT. Eur Radiol 21:2139–2147
radiation dose to the patient compared to stan- Flohr TG, McCollough CH, Bruder H, Petersilka M,
dard single energy CT scans. The dual source Gruber K, Süß C, Grasruck M, Stierstorfer K, Krauss
B, Raupach R, Primak AN, Küttner A, Achenbach S,
dual-energy CT solution provides technical and Becker C, Kopp A, Ohnesorge BM (2006) First per-
clinical advantages, such as flexible selection of formance evaluation of a dual-source CT (DSCT) sys-
the kVs of the low-energy beam allows for opti- tem. Eur Radiol 16(2):256–268
mized scanning, adapted to clinical needs. The Genant HK, Boyd D (1977) Quantitative bone mineral
analysis using dual energy computed tomography.
wide detector coverage combined with the high Invest Radiol 12(6):545–551
rotation time allows scanning of even challeng- Grant KL, Flohr TG, Krauss B, Sedlmair M, Thomas C,
ing patients, for example those short of breath. Schmidt B (2014) Assessment of an advanced image-­
Twin Spiral and TwinBeam are alternatives avail- based technique to calculate virtual monoenergetic
computed tomographic images from a dual-energy
able for single source CTs. Despite the different examination to improve contrast-to-noise ratio in
technical realizations, the same image-based examinations using iodinated contrast media. Invest
automatic processing techniques and processing Radiol 49(9):586–592
applications are applied, allowing the use of DE Graser A, Becker CR, Staehler M et  al (2010) Single-­
phase dual-energy CT allows for characterization of
in a routine clinical workflow. renal masses as benign or malignant. Invest Radiol
45(7):399–405
Compliance with Ethical Standards Henzler T, Fink C, Schoenberg SO, Schoepf UJ (2012)
Dual energy CT: radiation dose aspects. Am J
Roentgenol 199:S16–S25
Disclosure of Interests  Bernhard Schmidt and Thomas
Johnson TRC, Nikolaou K, Wintersperger BJ, Leber AW,
Flohr are employees of Siemens Healthineers.
von Ziegler F, Rist C, Buhmann S, Knez A, Reiser MF,
Becker CR (2006) Dual source cardiac CT imaging:
initial experience. Eur Radiol 16:1409–1415
Johnson TRC, Krauß B, Sedlmair M, Grasruck M, Bruder
References H, Morhard D, Fink C, Weckbach S, Lenhard M,
Schmidt B, Flohr T, Reiser MF, Becker CR (2007)
Achenbach S, Ropers D, Kuettner A, Flohr T, Ohnesorge Material differentiation by dual energy CT: initial
B, Bruder H, Theessen H, Karakaya M, Daniel WG, experience. Eur Radiol 17(6):1510–1517
Bautz W, Kalender WA, Anders K (2006) Contrast-­ Kalender W (1987) Vertebral bone mineral analysis.
enhanced coronary artery visualization by dual-source Radiology 164:419–423
computed tomography  – initial experience. Eur J Kalender WA, Perman WH, Vetter JR, Klotz E (1986)
Radiol 57(3):331–335 Evaluation of a prototype dual-energy computed
Agrawal MD, Pinho DF, Kulkarni NM, Hahn PF, tomographic apparatus. I. Phantom studies. Med Phys
Guimaraes AR, Sahani DV (2014) Oncologic applica- 13(3):334–339
tions of dual-energy CT in the abdomen. Radiographics Knobloch G, Jost G, Huppertz A, Hamm B, Pietsch H
34(3):589–612 (2014) Dual-energy computed tomography for the
Albrecht MH, Vogl TJ, Martin SS, Nance JW, Duguay assessment of early treatment effects of regorafenib
TM, Wichmann JL, De Cecco CN, Varga-Szemes A, in a preclinical tumor model: comparison with
van Assen M, Tesche C, Schoepf UJ (2019) Review dynamic contrast-enhanced CT and conventional
of clinical applications for virtual monoenergetic dual-­ contrast-enhanced single-energy CT.  Eur Radiol
energy CT. Radiology 293(2):260–271 24(8):1896–1905
Dual-Energy: The Siemens Approach 27

Krauss B, Grant KL, Schmidt BT, Flohr TG (2015) The puted tomography coronary angiography. J Am Coll
importance of spectral separation: an assessment of Cardiol 50(25):2393–2398
dual-energy spectral separation for quantitative ability Ruzsics B, Schwarz F, Schoepf UJ et al (2009) Comparison
and dose efficiency. Invest Radiol 50(2):114–118 of dual-energy computed tomography of the heart
Laval-Jeantet AM, Roger B, Bouysee S, Bergot C, with single photon emission computed tomography
Mazess RB (1986) Radiology 159(2):463–6. for assessment of coronary artery stenosis and of
https://doi.org/10.1148/radiology.159.2.3961178. the myocardial blood supply. Am J Cardiol 104(3):
Leber AW, Johnson T, Becker A, von Ziegler F, Tittus J, 318–326
Nikolaou K, Reiser M, Steinbeck G, Becker CR, Knez Scheffel H, Alkadhi H, Plass A, Vachenauer R, Desbiolles
A (2007) Diagnostic accuracy of dual-source multi-­ L, Gaemperli O, Schepis T, Frauenfelder T, Schertler
slice CT-coronary angiography in patients with an T, Husmann L, Grunenfelder J, Genoni M, Kaufmann
intermediate pretest likelihood for coronary artery dis- PA, Marincek B, Leschka S (2006) Accuracy of dual-­
ease. Eur Heart J 28(19):2354–2360 source CT coronary angiography: first experience in a
Luo XF, Xie XQ, Cheng S, Yang Y, Yan J, Zhang H, Chai high pre-test probability population without heart rate
WM, Schmidt B, Yan FH (2015) Dual-energy CT for control. Eur Radiol 16(12):2739–2747
patients suspected of having liver Iron overload: can Scheffel H, Stolzmann P, Frauenfelder T, Schertler T,
virtual iron content imaging accurately quantify liver Desbiolles L, Leschka S, Marincek B, Alkadhi H
iron content? Radiology 277(1):95–103 (2007) Dual-energy contrast-enhanced computed
Mallinson PI, Coupal TM, McLaughlin PD, Nicolaou S, tomography for the detection of urinary stone disease.
Munk PL, Ouellette HA (2016) Dual-energy CT for the Invest Radiol 42(12):823–829
musculoskeletal system. Radiology 281(3):690–707 Schenzle JC, Sommer WH, Neumaier K et al (2010) Dual
Mangold S, Gatidis S, Luz O, König B, Schabel C, energy CT of the chest: how about the dose? Invest
Bongers MN, Flohr TG, Claussen CD, Thomas C Radiol 45:347–353
(2014) Single-source dual-energy computed tomogra- Schneider D, Apfaltrer P, Sudarski S et  al (2014)
phy: use of monoenergetic extrapolation for a reduc- Optimization of kiloelectron volt settings in cerebral
tion of metal artifacts. Invest Radiol 49(12):788–793 and cervical dual-energy CT angiography determined
Marin D, Boll DT, Mileto A, Nelson RC (2014) State of with virtual monoenergetic imaging. Acad Radiol
the art: dual-energy CT of the abdomen. Radiology 21:431–436
271(2):327–342 Stolzmann P et al (2010) In vivo identification of uric acid
Matt D, Scheffel H, Leschka S, Flohr TG, Marincek B, stones with dual-energy CT: diagnostic performance
Kaufmann PA, Alkadhi H (2007) Dual-source CT coro- evaluation in patients. Abdom Imaging 35(5):629–635
nary angiography: image quality, mean heart rate, and Thomas C, Schabel C, Krauss B, Weisel K, Bongers M,
heart rate variability. Am J Roentgenol 189(3):567–573 Claussen CD, Horger M (2015) Dual-energy CT: vir-
Pache G, Krauss B, Strohm P, Saueressig U, Blanke P, tual calcium subtraction for assessment of bone mar-
Bulla S, Schäfer O, Helwig P, Kotter E, Langer M, row involvement of the spine in multiple myeloma.
Baumann T (2010) Dual-energy CT virtual noncal- Am J Roentgenol 204(3):W324–W331
cium technique: detecting posttraumatic bone marrow Uhrig M, Sedlmair M, Schlemmer HP, Hassel JC, Ganten M
lesions – feasibility study. Radiology 256(2):617–624 (2013) Monitoring targeted therapy using dual-­energy
Petersilka M, Stierstorfer K, Bruder H, Flohr T (2010) CT: semi-automatic RECIST plus supplementary func-
Strategies for scatter correction in dual source tional information by quantifying iodine uptake of
CT. Med Phys 37:5971–5992 melanoma metastases. Cancer Imaging 13(3):306–313
Pontana F, Faivre JB, Remy-Jardin M et al (2008) Lung Vetter JR, Perman WH, Kalender WA, Mazess RB, Holden
perfusion with dual-energy multidetector-row CT JE (1986) Evaluation of a prototype dual-energy com-
(MDCT): feasibility for the evaluation of acute pul- puted tomographic apparatus. II. Determination of ver-
monary embolism in 117 consecutive patients. Acad tebral bone mineral content. Med Phys 13(3):340–3.
Radiol 15(12):1494–1504 https://doi.org/10.1118/1.595951.
Primak AN, Fletcher JG, Vrtiska TJ, Dzyubak OP, Lieske Vliegenthart R, Pilgrim GJ, Ebersberger U, Rowe GW,
JC, Jackson ME, Williams JC Jr, McCollough CH Oudkerk M, Schoepf U (2012) Dual-energy CT of the
(2007) Noninvasive differentiation of uric acid ver- heart. Am J Roentgenol 199:S54–S63
sus non-uric acid kidney stones using dual-energy Werner A, Krauss B, Haberland U, Bongers M, Starke
CT. Acad Radiol 14(12):1441–1447 U, Bakchoul T, Enkel S, Nikolaou K, Horger M
Remy-Jardin M, Faivre JB, Pontana F, Molinari F, Tacelli (2019) Dual-energy CT for liver iron quantification
N, Remy J (2014) Thoracic applications of dual in patients with haematological disorders. Eur Radiol
energy. Semin Respir Crit Care Med 35(1):64–73 29(6):2868–2877
Ropers U, Ropers D, Pflederer T, Anders K, Kuettner Yu L, Leng S, McCollough C et  al (2012) Dual-energy
A, Stilianakis NI, Komatsu S, Kalender W, Bautz W, CT-based monochromatic imaging. Am J Roentgenol
Daniel WG, Achenbach S (2007) Influence of heart 199(5 Suppl):S9–S15
rate on the diagnostic accuracy of dual-source com-
 Dual-Energy: The Philips Approach

Ami Altman, Galit Kafri, and Sary Shenhav

Contents
1 Spectral Detection Through a Dual-Layer Detector  30
2 Spectral Material Decomposition and Reconstruction  31
3 Spectral Results  34
3.1  MonoE: Monoenergetic Images [HU]  34
3.2  VNC: Virtual Non-Contrast  36
3.3  Iodine No Water [mg/ml]  37
3.4  Calcium Suppression [HU]  37
3.5  Iodine Density [mg/ml]  38
3.6  Contrast-Enhanced Structures [HU]  38
3.7  Iodine Removed [HU]  39
3.8  Uric Acid and Uric Acid Removed Pair [HU]  39
3.9  Z Effective  40
3.10  Electron Density [%EDW]  40
References  42

Abstract single-source approach to spectral imaging

has some clear advantages: energies acquisi-
The Philips dual-layer detector approach for tion is perfectly aligned by design; it does not
spectral imaging was introduced as early as require a special dual-energy acquisition mode
2005  in a prototype installed in Hadassah thus making all scans spectral and enables
University Medical Center, Israel. Since then projection-based material decomposition. It
both the detector design and the material also allows for advanced clinical application,
decomposition and reconstruction techniques such as cardiac and perfusion, to utilize spec-
were further developed and improved to allow tral imaging. In addition, it allows to easily
better SNR and CNR in spectral results. This overcome limitations of the source-based
techniques in the sense that it does not have
dose penalties, field of view restriction, it does
A. Altman · G. Kafri · S. Shenhav (*) not require to slow rotation time, and it is not
Philips Medical Systems Technologies Ltd., limited by patient habitus.
Advanced Technologies Center, Haifa, Israel
e-mail: [email protected]

© The Author(s), under exclusive license to Springer Nature Switzerland AG 2022 29

H. Alkadhi et al. (eds.), Spectral Imaging, Medical Radiology Diagnostic Imaging,
https://doi.org/10.1007/978-3-030-96285-2_3
30 A. Altman et al.

In this chapter, we will discuss the dual-­enough to maintain the same detector pitch and
layer detector architecture, projection-based geometrical efficiency as a conventional CT
material decomposition and image reconstruc- detector (Figs. 1 and 2). The top scintillator lay-
tion as well as the different spectral results and
er’s atomic number and thickness have been
their clinical use. optimized to maximize energy separation at 120
and 140  kVp, while maintaining high enough
signal statistics for the low-­energy raw data even
for a large patient. A low Z (atomic number)
1 Spectral Detection Through Garnet scintillator material, with a high light-
a Dual-Layer Detector output (~15%–20% better than GOS), has been
developed to meet these requirements. This con-
While material decomposition in energy-­ tributes to a high SNR in the top (low-energy)
selective CT was proposed by Alvarez and layer detector, enabling it to function at a very
Macovski in 1976, a dual-layer detector for a low dose without causing artifacts, typical to
simultaneous acquisition of two energies in CT electronic-noise dominant signals.
was first proposed by Brooks and Di Chiro in The mean energy separation of the dual-layer
1978. A Philips Healthcare team proposed a dif- detector, at 120 kVp, with and without a 30-cm
ferent configuration and implementation of that water absorber, is shown in Fig. 3. The decrease
idea (Carmi et al. 2005; Altman et al. 2006) and in energy separation with increasing patient size
in October 2005 Philips has installed the world is compensated due to the complete consistency
first clinical prototype utilizing a single-source in sampling of the two energies.
spectral detection technique at the Hadassah The Philips approach to spectral imaging is
Medical Center in Jerusalem, Israel. This imple- unique in the sense that it is based on a single
mentation included two attached scintillator lay- source and spectral detection. Compared to
ers, optically separated, and read by a source-based dual-energy techniques, the
side-looking, edge-on, silicon photodiode, thin detector-­based technique has some clear advan-

A thin side looking photodiode array

shielded by a 0.5 mm Tungsten layer
~50%
X-Rays E1 image
optical glue Low Energy Raw data
Low Energy

1-mmn Low Z,
high yield Garnet
+
reflecting paint
Bottom ~50%
Scintillator: High Energy Raw data E2 image
Y
2-mm GOS High Energy

=
Weighted combined Raw data CT image
Full CT Image

Fig. 1  A schematic view of the Philips Healthcare dual-layer detector

Dual-Energy: The Philips Approach 31

tages: energies acquisition is perfectly aligned by

design as they are not only simultaneously mea-
sured, but also sampled exactly at the same angle
at the same pixel, unlike the other methods of
dual-energy CT; it does not require a special
dual-energy acquisition mode thus making all
scans spectral and enables projection-based
material decomposition (see next section) which
in turns allows better SNR and CNR in spectral
results, like iodine quantification, effective
atomic number image, electron densities, and vir-
tual MonoE images (Sellerer et  al. 2018; Ehn
et al. 2018). It also allows for advanced clinical
Fig. 2  A vertical implementation of a 16  ×  16 pixels
dual-layer detector application, such as cardiac and perfusion, to uti-
lize spectral imaging. In addition, it allows to
easily overcome limitations of the source-based
x 104 LowE and HighE, Photon Spectra, in air techniques in the sense that it does not have dose
a
penalties, field of view restriction, it does not
10
require to slow rotation time and it is not limited
dN/DE (# / [keV•mm2•mAs])

8
by patient habitus.

∆ E = 22keV

2 Spectral Material
4
Decomposition
2
and Reconstruction

0
Dual-energy spectral decomposition in the projec-
10 20 30 40 50 60 70 80 90 100 110 120 130
X-Ray Energy (keV)
tion domain, using a two-base model, was first pro-
b LowE and HighE, Photon Spectra, 30•cm Water posed by Alvarez and Macovski (1976). In this
approach, the two projections data sets of the low
300 and high energies are transformed into a new pair
dN/DE (# / [keV•mm2•mAs])

250
of projections data sets, from which all the various
spectral results can be derived. The resulting two
∆ E = 17keV
200 spectral-projections data sets are independent of
150
the incident spectrum, hence clear from beam hard-
ening effect, while even metal beam hardening and
100 artifacts can be suppressed quite easily. This is a
very important result of the projection-domain
50
spectral decomposition, unlike image-domain
0
10 20 30 40 50 60 70 80 90 100 110 120 130
spectral decomposition, where the resulting spec-
X-Ray Energy (keV) tral images (e.g., virtual monochromatic images)
suffer from beam hardening effects, and require
Fig. 3  Top and bottom layers’ spectra and mean energy
difference in air (a) and with 30-cm water absorber (b), special beam hardening correction, which is, often,
for 120 kVp insufficient (see e.g., Carmi et al. 2005).
32 A. Altman et al.

SBI
Photo-electric image
Photo-Electric Sinograms
Low-Dose-Noise and Noise-
induced-Bias correction

Spectral decomposition
Image
µ Compton
Low & High E reconstruction
Image-Domain
Sinograms Compton Scatt. Compton-Scatt image Structure
Sinogram Synchronization filter,
and frequency
X-Ray Energy
dependent denoiser
ACF (Anti Correlation Filter)

Noise image
Conventional image
Weighted Multiplexing HighE & LowE

Image
HighE reconstruction
⊗
LowE

Fig. 4  A schematic description of the Philips spectral decomposition and image reconstruction

The Philips implementation of projection-­ The above transformation utilizes a two-base

domain spectral decomposition (Altman et al. model:
2015) assumes an accurate forward model of the
projection’s formation, as an input. This includes
 M  E    scM  E    ph
M
 E   fsc  E   scM  Z , 
the precise source X-ray spectrum, the beam fil-  f ph  E    ph
M
 Z , 
tration, its variations in the directions to all the
detection pixels, and the resulting different spec- where M represents a material mix, optimized for
trum incident on each detection pixel in air. This energy-dependence “universality” (~71% soft
implementation is simpler than the empirical tissue, ~28% bone, ~1% iodine), while assuming
approach proposed in (Alvarez and Macovski the following:
1976), or the maximum-likelihood method pro-
posed by Alvarez (Alvarez et al. 2011). 1. Variable separation between the energy
The basic Philips approach to the transforma- dependence and the Z dependence of the
tion from the low and high energies projections attenuation coefficient components (cross
data sets to the Photo-Electric and Compton-­ sections) is possible within the X-rays energy
scattering projections data sets is described below range used in CT.
in Fig. 4 and the following equations: 2. The energy dependence of Compton
Scattering, and the Photo-Electric cross sec-
P  , P 
high low

P  , P 
M
scatt .
M
photo
tions, can be considered to be the same for all
patients and test phantoms, based on the
Spectrum dependent Spectrum independent
above material mix (this excludes phantoms


 0  ascatt
M
 
. ,  0  aphoto
M
 and objects with a very large content of heavy
metals/elements).
where μ0 is an arbitrary quantity with a dimen- 3. Coherent scattering can be either neglected or
sion of an attenuation coefficient, which the end included in the photo-electric component
results are divided by at the end of the process. (Energy dependence is ~1/E2).
Dual-Energy: The Philips Approach 33

Assuming the above, the line integrals can be

expressed as:

Ph,l   log  0
  
  exp     ,E  d  F  E   EdE 
h ,l 
and the projections value for each line integral
(detection pixel) can be obtained:
  
  0 Fh,l  E   EdE 

 
  exp  f  E   a M  f  E   a M  F  E   EdE 

Ph,l   log  0 
scat . scat phot phot h ,l

  
  0 Fh,l  E   EdE 

M M
From which, the values of ascat and aphot , the ing Photo-Electric signal, and vice versa.
spectral line integrals are derived. This would result in a correlated noise
Where, between the two, hence would need a spe-
cial consideration of how to reduce this
1. The functions fphot(E) and fscat(E) are derived noise through an Anti-Correlation Filter
from the E dependence of the chosen material (ACF) in the projection domain.
mix M, calculated from NIST tables. (b) Noise Induced Bias: It occurs in many
2. Fl(E) and Fh(E) are the energy distributions of cases, since the original radiation inten-
the X-ray beam in air, incident on the top sity per detection pixel is divided between
layer pixels and the bottom layer pixels, two separate projection signals (Scatter
respectively (after all filtration and without and Photo projections sets in Philips
any scanned object/body). Note that both approach). In this case, taking the loga-
Fl(E) and Fh(E) are matrices of distribution rithm of the raw signal, the statistical
functions that must be pre-calculated and uncertainty of which is at the tail of the
require full input/knowledge of the filtration Poisson distribution, is causing a DC
along the rays to each detector pixel. bias. This bias is corrected by a special
3. The material decomposition procedure filter at the input to the material decom-
described above has two important “side position process.
effects” that are addressed: 4. Following the reconstruction of the Photo-­
(a) Anti-correlated Noise: The transforma- Electric and the Scatter images, special filters/
tion, mentioned above, determines the algorithms are used, in the image domain, to
specific portions of LowE signal and conserve image structures, edges, and fea-
HighE signal, assigned to the Scatter and tures, adapted from the conventional image,
to the Photo projection signals, respec- while reducing the image noise in a frequency-­
tively. As a result any portion of a radia- dependent form.
tion signal that is assigned to the Scatter
projection signal, while it should have It is important to emphasize that projection-­
been assigned to the Photo signal, would domain spectral decomposition can be used only
cause a correlated error in the correspond- if the signals from the two X-ray energies are
34 A. Altman et al.

sampled both simultaneously (“Equi-time”) and and are utilized for routine diagnostic purposes.
at the same angle (“Equi-angle”). Only the For every scan, the pre reconstructed sum of the
spectral-­
detector and the fast-kVp-switching signals from the simultaneous acquisition of the
methods meet this condition. high- and the low-energy data is combined to
obtain the total amount of absorbed energy.
Filtered back projection or iterative reconstruc-
3 Spectral Results tion algorithms are then used to reconstruct the
combined raw data and create true conventional
The Philips spectral CT system can provide a images. It has been shown that image quality of
variety of spectral image types, on top of the con- these conventional images from the Philips spec-
ventional images, which are reconstructed tral CT system are comparable to images obtained
through advanced spectral algorithms. Those from a single-energy scanner (Hojjati et al. 2017;
spectral results can be generated either prospec- Van Ommen et al. 2018).
tively, on the scanner and sent to PACS, or retro- Spectral results can have units of attenuation
spectively, directly on PACS utilizing Spectral (HU) and other voxel values that represents phys-
Based Images known as SBI (as described above) ical quantities, for example, density in mg/ml.
and have the potential for additional clinical The clinical use of the different spectral results
information to conventional CT imaging. Each generated from the spectral-detector DECT have
spectral result is designed for a different clinical been demonstrated in several papers for several
usage as will be explained in this chapter. clinical applications in body, cardiac and neuro
As described above, data from the low and imaging (Brun Andersen et al. 2020; Fulton et al.
high energy layers of the spectral detector 2017; Rajiah et al. 2017a; Neuhaus et al. 2017a)
undergo spectral decomposition in the projection as well as for emergency imaging (Demirler
space to generate optimal photo-electric absorp- Simsir et  al. 2020). The high quantitative accu-
tion and Compton scattering images that are used racy of the various spectral results was recently
as a two-base model. These optimal photo-elec- demonstrated in several studies (Ehn et al. 2017;
tric absorption and Compton scattering images Hua et al. 2018).
do not represent the pure Compton and photo- In the following, the algorithm and the poten-
electric effects. They are slightly modified ver- tial clinical usage of the different spectral results
sion of them. This is because different materials will be discussed.
have slightly different energy dependencies,
mainly at low energies, from the expected theo-
retical models of the two basic physical interac- 3.1 MonoE: Monoenergetic
tions. Moreover, coherent scattering which is the Images [HU]
third component that contributes to the total
attenuation of X-ray in matter need to be included, MonoE images are virtual mono-energy images
despite its relatively small contribution. which simulate images as if they are obtained
The optimal Photo and Scatter sets of images using a pure monochromatic X-ray beam at a
are stored together with the derived noise-images specific keV value. Virtual monochromatic
set in a special SBI (Spectral Based Images) for- images are generated between 40  keV and
mat. All the relevant spectral results and images 200 keV, in increment of 1 keV, and the voxels in
can be derived from the SBI series, using various these images represent Hounsfield values (HU).
algorithms. Hence, the spectral results can be This is illustrated in Fig. 5. The MonoE images
created on demand and reviewed as needed on are created by a linear combination of the two-­
PACS and are not required to be sent to PACS base model, namely the photo-electric effect (PE)
prospectively. and Compton scattering (Sc) components, where
The conventional images are analogous to the a different weight is used for each KeV.  In the
images obtained from a single-energy scanner low energy range the proportion of the photo-
Dual-Energy: The Philips Approach 35

Fig. 5  Virtual mono-energetic images from 40 to 200  keV displaying tissue attenuation properties similar to those
resulting from imaging with a mono-energetic beam at a single keV level (Fulton et al. 2017)

electric effect is high and the Compton scattering corresponding conventional images due to spe-
is low, whereas in the high energy range it is cial noise reduction techniques that are used in
opposite. In the 70 keV the proportions of these the spectral reconstruction process to minimize
two components are very similar. the noise and to optimize the signal to noise ratio
It should be noted that it is possible to gener- (Kaltsz et  al. 2017). See also Spectral Material
ate monoE results also for keV values higher than Decomposition and Reconstruction section
the used kVp value since the behavior of the uni- above.
versal energy dependent basis is known also for A specific result named mono E-equivalent to
high energies. Also, in order to stay above the conventional CT can be generated as well. This
K-edge of Iodine (33.2 keV), the lowest monoE result has almost the same HU value as a conven-
was selected to be 40 keV. tional image generated from 120  kVp voltage
In the Philips spectral CT system, the noise in (regardless of the actual tube voltage used during
the spectral results is typically lower than in the the scan) but with lower artifacts and noise.
36 A. Altman et al.

The equivalent monochromatic images are hemorrhage and brain lesions (Lennartz et  al.
70 keV, 66 keV, and 64 keV for body, head, and 2018). In addition, it was demonstrated that
extremities, respectively. improved gray-white matter differentiation in
Monochromatic images may overcome some cranial CT by using virtual mono-energetic
of the limitations of a polychromatic X-ray beam. images enables a radiation dose reduction com-
A conventional CT image is created from a poly-­ pared to conventional images (Reimer et  al.
energetic X-ray tube with a certain voltage (e.g., 2019). The reduction of blooming artifacts for
120  kVp, 140  kVp) and therefore it is recon- coronary stent assessment and calcium blooming
structed from multiple energies. Since the mono-­ reduction in cardiac imaging was recently dem-
energetic series (MonoE) represents a single onstrated as well (Hickethier et  al. 2017; Van
energy, it minimizes some of the known limita- Hedent et al. 2018).
tions of a polychromatic X-ray beam. For exam- The next spectral results that are described are
ple, beam-hardening, metallic, and calcium the virtual non-contrast (VNC), iodine no water,
blooming artifacts. and calcium suppressed. These three spectral
The potential benefits of the high virtual types belong to the same category of two-­material
mono-energetic images derived from dual-layer decomposition. A two-material decomposition
CT scans for reduction of artifacts caused by algorithm assumes that each voxel consists of
orthopedic metal implants, were demonstrated two types of materials only. In order to achieve
both in a phantom study (Wellenberg et al. 2017) this, a simple linear basis transformation from the
and clinical studies for implants in the spine, pel- basis of the photo-electric effect and Compton
vis, and extremities (Neuhaus et al. 2017b; Große scattering components into a new basis represen-
Hokamp et al. 2017a) as well as dental implants tation of two selected materials is performed. By
(Große Hokamp et al. 2018). selecting the two materials to be iodine and water,
Low-energy mono-energetic images are influ- it is possible to generate the virtual non-contrast
enced by photo-electric data, resulted in increas- (VNC) and the iodine no water results. By select-
ing attenuation and signal to noise ratio (SNR) ing the materials to be a calcium-based material
and are therefore useful for all vascular imaging. and water, a calcium suppressed image can be
The increased attenuation at low keVs can be created.
used for contrast reduction, which is especially
important in patients with renal insufficiency
(Oda et  al. 2018; Tsang et  al. 2017; Nagayama 3.2 VNC: Virtual Non-Contrast
et  al. 2017; Hickethier et  al. 2020) and for
improvement in image quality when the contrast This is a spectral result that mimics the attenua-
enhancement is suboptimal, salvaging angio- tions values of a non-contrast CT scan from a
graphic studies and reducing the need for addi- data that is acquired with a contrast injection. It
tional contrast or radiation dose. Also, low thus has the potential of replacing a true non-­
MonoE reconstructions allow the user to create contrast series. By applying the two-martials
angiography studies from a routine contrast-­ decomposition to a NIST-based iodine and water
enhanced exam, adding additional diagnostic basis, the iodine attenuation contribution of each
information to the exam. In addition, low MonoE voxel is removed and only the water attenuation
images are used for better lesion conspicuity (Liu contribution of each voxel in HU is displayed
et  al. 2019; Yoon et  al. 2020; Große Hokamp according to the mono 70 keV image. Due to the
et al. 2017b). nature of this algorithm, the HU of the bony
The advantage of the monochromatic images structures and calcium pixels are decreased by
was also demonstrated in neuroimaging by reduc- about 50% of their value in the 70 keV image.
ing beam hardening artifacts for optimized gray-­ A recent study compared the VNC images
white matter contrast (Neuhaus et al. 2017c) and derived from the dual-layer spectral detector to a
for visualization improvement of intracranial true non-contrast (TNC), found a good agree-
Dual-Energy: The Philips Approach 37

ment of the attenuation measurements between Iodine no water [mg/ml] images have the
the two images in most abdominal tissues with an potential to allow for improved visualization of
overestimation in fatty tissues (Jamali et  al. iodine-enhanced tissues. This result can also be
2019). This is in agreement with prior studies used for iodine quantification, but only in areas
(Ananthakrishnan et al. 2017; Sauter et al. 2018) where iodine is present.
where the quality of iodine removal in VNC The accuracy of the iodine concentrations was
images was not influenced by the original con- tested in a phantom study for a range of concen-
trast enhancement and thus has a potential advan- trations between 2 and 20 mg/ml and found to be
tage in reducing the radiation dose delivered to within a 0.3 mg/ml accuracy (Hua et al. 2018).
the patient in biphasic and triphasic In Fig. 6, we show an example where a con-
examinations. ventional, VNC, and iodine no water images are
It was also shown that with an appropriately compared for a renal lesion.
chosen proportionality factor as a correction
coefficient, spectral CT VNC can reliably esti-
mate the calcium score from a contrast-enhanced 3.4 Calcium Suppression [HU]
coronary CTA and shows good agreement with
the conventional technique (Nadjiri et al. 2018). In this image type, voxels containing calcium are
suppressed and replaced by virtual HU values as
similar as possible to the expected HU without
3.3 Iodine No Water [mg/ml] calcium contribution to the attenuation. In a simi-
lar way to the VNC algorithm, the contribution of
This is a spectral result type in which the voxel the calcium-based material attenuation of each
values represent the iodine concentration of the voxel is removed and only the water attenuation
displayed tissue in mg/ml as calculated from the contribution of each voxel is displayed in HU
iodine-water two-material decomposition algo- according to the mono 70  keV.  In contrast to
rithm. The quantification of the iodine density is VNC algorithm, where the iodine material is rep-
calculated by scaling the iodine projection on the resented by a specific slope, the bony structures
water basis. Non-enhanced soft tissues are set to cannot be represented by one slope but of a range
approximately 0 mg/ml of iodine. of calcium-based materials slopes. According to

a b c

Fig. 6 (a) Axial CT scan at the level of left kidney shows image at the same level shows absence of significant
a cystic lesion, which has attenuation higher than a simple iodine in the lesion thus confirming that there was no con-
cyst (arrow); (b) virtual non-contrast CT at the same level trast uptake in the lesion, but the high attenuation is con-
shows that there is higher attenuation in the VNC image sistent with hemorrhage, thus a complicated cyst (Fulton
indicating that this is a hemorrhage; and (c) iodine-only et al. 2017)
38 A. Altman et al.

the calcium composition weight in each bony iodine-containing structures as well as direct
structure, an appropriate index value can be quantification of iodine.
selected. A low index value selection targets tis- The accuracy of the iodine concentrations for
sues with a low calcium composition weight; a iodine density was tested in a phantom study for
high index value selection targets tissues with a a range of concentrations between 2 and 20 mg/
high calcium composition weight. ml and found to be within a 0.3 mg/ml accuracy
In a recent study, the calcium suppressed (Hua et al. 2018).
images from dual-layer CT have been used to It was shown that iodine density allows one
visualize the bone marrow edema in traumatic to detect occult bone lesions that cannot be
vertebral compression fractures (Neuhaus et  al. detected with conventional CT due to the high
2018). In a different study (Abdullayev et  al. contrast of the bone and the lack of bone destruc-
2019), it was found that calcium suppressed tions. Iodine density yields high sensitivity and
images are capable of improving differentiation adequate specificity for the differentiation of
between a metastatic and a normal bone. Also, it vertebral trabecular metastases and healthy tra-
was demonstrated that calcium suppressed becular bone (Borggrefe et al. 2019). In another
images could clearly present the temporoman- study (Kikano et  al. 2020), it was shown that
dibular joint displacement. This suggested that iodine density images can help elucidate and
calcium suppressed images could be used to differentiate between various cardiothoracic
diagnose the displacement of the temporoman- pulmonary perfusion anomalies and may
dibular joint disc (Zhang et al. 2020). enhance a radiologist’s diagnostic confidence.
In the above described spectral types, where Also, Lennartz et al. showed that iodine overlay
the two-material decomposition is applied, the images obtained with spectral-detector CT
algorithm didn’t attempt to separate between two improve visual and quantitative diagnostic accu-
selected materials. In the spectral results that are racy in assessing skeletal muscle metastases
explained below, a classification method between compared to conventional images (Lennartz
two materials is performed. The spectral results et  al. 2019). In a study case, iodine density
that are included in this category are the iodine images allowed for differentiation of benign and
density, contrast-enhanced structures, iodine malignant pulmonary nodules (Große Hokamp
removed, and the pair uric acid and non-uric acid et al. 2019).
removed. In Fig.  7, we show hypodense lesion in the
head of the pancreas using conventional CT,
mono-energetic 40  KeV, iodine density, and Z
3.5 Iodine Density [mg/ml] effective. The Z effective result will be discussed
shortly.
This is an image type in which, similarly to the
iodine no water result, the voxels values repre-
sent the iodine concentration of the displayed tis- 3.6 Contrast-Enhanced
sue in mg/ml. In this image, the non-negative Structures [HU]
iodine quantification for all voxels which are
classified as including Iodine is calculated by In this result all the voxels, which are classified
scaling the iodine projection on the water basis as as including iodine, remain identical to MonoE
explained for the iodine no water result. Voxels 70 keV. All bone classified voxels are displayed
which are classified as not included iodine are set as black voxels. Contrast-enhanced structures
to 0 mg/ml and are visualized as black pixels. images have the potential of providing bone-free
Similar to the iodine no water result, iodine images which can help in visualizing vascular
density images enable the identification of structures without bone or calcifications.
Dual-Energy: The Philips Approach 39

Fig. 7  Patient with a slight increase in volume in the head well as mono 40 keV and Zeff images, the lesion stands
of the pancreas and stranding in the peripancreatic fat. In out against the normal pancreatic parenchyma and the
the pancreatic head, a slight hypodense lesion (arrow- lesion was correctly diagnosed as a pancreatic adenocarci-
heads outline the lesion) is visible on the conventional noma (Brun Andersen et al. 2020)
image; however, it is easily missed. On iodine density as

3.7 Iodine Removed [HU] 3.8.1 Uric Acid [HU]

In the uric acid result, all the voxels which are clas-
An image type in which all the voxels which are sified as including uric acid remain identical to
classified as not including iodine above a certain MonoE 70 keV. All other voxels classified as not
threshold are displayed as black voxels. All other including uric acid are displayed as black voxels.
voxels remain identical to MonoE 70 keV.
3.8.2 Uric Acid Removed [HU]
In the uric acid removed result, all the voxels
3.8  ric Acid and Uric Acid
U which are classified as not including uric acid
Removed Pair [HU] remain identical to MonoE 70 keV. All other vox-
els classified as including uric acid are displayed
In these two spectral results, the classification is as black voxels.
done between uric acid and calcium voxels and the The uric acid and uric acid removed image pair
two results are complementary with each other. are intended for uric acid and calcium classifica-
40 A. Altman et al.

tion and therefore have a potential use in gout dis- rations. It was found that the accuracy of the Z
ease diagnosis and stone characterization. effective is on the order of +/− 2% for both soft
In gout disease, the attenuation of low-atomic-­ tissue and bone-equivalent materials, with some-
weight monosodium urate (MSU) crystals differs what larger percentage deviations for lung-­
from that of high-atomic-weight calcium such as mimicking materials. The accuracies were found
calcium pyrophosphate dihydrate (CPPD) crys- to be similar in different scans and reconstruction
tals, that exist in pseudogout and can have a clini- parameters.
cal presentation like that of gout. Z Effective images have the potential to dif-
Uric acid-calcium pairs are also useful in ferentiate tissues based on their atomic number
characterizing urinary calculus composition as values, for example, in stone characterization
demonstrated by Rajiah et al. (2017b). (Fulton et al. 2017). This is shown in Fig. 8 in a
The last two spectral results that are described different study. In addition, it was shown that the
below, Z effective and electron density can be use of contrast-enhanced spectral CT including Z
used for material characterization. effective images increases the confidence of the
radiologists in correctly characterizing various
lesions and minimizes the need for supplemen-
3.9 Z Effective tary examinations (Brun Andersen et  al. 2020).
See Fig. 8.
This is an image in which the voxel values repre- The benefit of the dual-layer spectral CT was
sent the effective atomic number of the displayed recently demonstrated in the emergency depart-
tissue. While the atomic number is characteristic ment (Demirler Simsir et al. 2020). It was shown
of an element, the effective atomic number is that by using iodine density and Zeff maps, the
characteristic of materials consisting of more detection of subtle filling defects and demonstra-
than one element such as water, soft tissue, and tion of the presence or lack of lung perfusion
bone. deficits in pulmonary embolism is better. This is
A Z effective calculation is based on the ratio shown in Fig. 9.
of attenuations at two different energies and is
therefore independent of the material density.
The expected Zeff values for known different tis- 3.10 Electron Density [%EDW]
sues from the literature were calculated and a
conversion curve between the monoEs ratio and This is a spectral result that displays the electron
the calculated Z effective was created. density of each voxel relative to the electron den-
The Z effective images can be displayed in sity of water (3.34 × 10^29 electrons × m-3) in
color or gray scales. The dynamic range is set to units of percent where the expected value for
be between 5 and 30, where water is 7.4 and air is water in these units is 100 [%EDW].
set to be zero. Non-enhanced soft tissues have Z The electron density (ED) estimation is based
effective value of approximately seven. Fatty tis- on a linear combination of the photo-electric
sues have lower Z effective than water, while effect and Compton scattering where the
bone and contrast-enhanced tissues have higher Compton scattering component dominates. The
values compared with water Z effective. Metal two parameters of the linear combination coeffi-
implants have a high Z effective possibly higher cients were determined by finding the best fit to
than 30. the expected electron densities of known litera-
In a phantom study (Hua et al. 2018), different ture tissues. The normalization was chosen so
tissues with different Z effective values from that the relative ED of water is 100.
approximately 6 (adipose tissue) to about 13 For several decades, in order to calculate the
(cortical bone) were scanned in different configu- radiation dose distribution, HU were converted to
Dual-Energy: The Philips Approach 41

a b

Fig. 8  Urinary calculus composition. (a) Coronal 120-­ number-based reconstruction at the same level shows that
kVp routine diagnostic image in a patient with acute the calculus has high atomic number (arrow) consistent
abdominal pain shows a 7-mm calculus in the inferior with a calcium calculus (Rassouli et al. 2017)
pole of the right kidney (arrow). (b) Effective atomic

a b c

d e f

Fig. 9  A 36-year-old woman presented with right-sided Iodine density map demonstrated a wedge-shaped area of
acute chest pain (upper row). (a) On conventional CT, a decreased perfusion (iodine density; 0.08  mg/ml; blue
small filling defect was present in right lower lobe sub-­ arrow) compared to the adjacent lung parenchyma (iodine
segmental pulmonary artery (arrow). (b) Iodine density density; 1.27  mg/ml; white arrow) indicating a sub-­
and (c) Zeff maps showed no perfusion deficit distal to the segmental pulmonary embolism with perfusion deficit. (f)
filling defect (arrows). A 90-year-old woman presented The Zeff map demonstrated the wedge-shaped area with
with dyspnea, right-sided chest pain (lower row), (d) decreased perfusion color-coded in light blue, yellow, and
Conventional CT, small filling defect was present in right red (arrow) (Demirler Simsir et al. 2020)
lower lobe sub-segmental pulmonary artery (arrow). (e)
42 A. Altman et al.

the electron or physical density via a calibration References

curve, in the case of photon therapy or to the pro-
ton stopping power, in the case of proton therapy. Abdullayev N et  al (2019) Improvements of diagnostic
Dual-layer CT allows a direct calculation of accuracy and visualization of vertebral metastasis
using multi-level virtual non-calcium reconstructions
the electron density on a pixel-by-pixel basis. from dual-layer spectral detector computed tomogra-
This approach can potentially enhance the work- phy. Eur Radiol 29(11):5941–5949
flow of radiation therapy planning by eliminating Altman A, Shapiro O, Levene S, et  al. (2006) Double
the need for a calibration curve mapping for pho- decker detector for spectral CT. US States Patent US
7,968,853
ton therapy and improve the range accuracy for Altman A, Goshen L, Shechter G, Brown K, Yagil Y,
proton therapy. Katchalski T, Carmi R, Proksa R, Thran A, Brendel
The accuracy of the electron density was B, Daerr H, Romman Z (2015) From a dual-energy
assessed recently in phantoms studies (Hua et al. spectral detector to photon-counting, and the ben-
efit of multi energy-bins detectors. CERN SpecsXray
2018; Rassouli et al. 2017). Hua et al. concluded Workshop April 2015
that the electron density accuracy is of the order Alvarez RE (2011) Estimator for photon counting energy
of ±1% for both soft tissue and bone-equivalent selective x-ray imaging with multi-bin pulse height
materials and did not appear to be sensitive to dif- analysis. Med Phy 38(5):2324–2334
Alvarez RE, Macovski A (1976) Energy-selective recon-
ferent scans and reconstruction parameters. Ohira struction in X-ray computerized tomography. Phys
et  al. (2019) found that percentage error of the Med Biol 21(5):733–744
electron density for each reference material var- Ananthakrishnan L et  al (2017) Spectral detector
ied from −2.0% to 1.2% and therefore the use of CT-derived virtual noncontrast images: comparison
of attenuation values with unenhanced CT.  Abdom
dual-layer spectral CT can enhance radiotherapy Radiol 42(3):702–709
treatment planning. In another study, it was dem- Borggrefe J et  al (2019) Accuracy of iodine density
onstrated that in homogeneous and heteroge- thresholds for the separation of vertebral bone metas-
neous phantoms, the use of electron density and tases from healthy-appearing trabecular bone in
spectral detector computed tomography. Eur Radiol
Z effective can improve the accuracy of the stop- 29(6):3253–3261
ping power ratio (SPR) predictions for particle Brooks RA, Di Chiro G (1978) Split detector com-
therapy resulting in remaining discrepancies in puted tomography: a preliminary report. Radiology
the order of 1%. This can minimize the beam 126(1):255–257
Brun Andersen M et  al (2020) Impact of spectral body
range uncertainty and allow for reduced safety imaging in patients suspected for occult cancer:
margins in the patient (Faller et al. 2020). a prospective study of 503 patients. Eur Radiol
Recently, it was demonstrated on few case 30(10):5539–5550
reports that ED imaging improves GGO visual- Carmi R, Naveh G, Altman A (2005) Material separa-
tion with dual-layer CT.  Presented at the IEEE-MIC
ization compared with conventional CT.  It was Conference. Puerto Rico: Wyndham El Conquistador;
suggested that ED imaging may improve the 23–29 Oct 2005. First, and original presentation of the
detection of early-stage COVID-19, a stage for 2D scatter-plot based dual-energy material decompo-
which conventional CT has shown limited sensi- sition method in image plane. See also US PATENT:
US7,778,380 B2, 2010, by the same authors
tivity for detection (Daoud et al. 2020). Daoud B et  al (2020) Could spectral CT have a poten-
tial benefit in coronavirus disease (COVID-19)? Am J
Compliance with Ethical Standards Roentgenol 216(2):349–354
Demirler Simsir B, Danse E, Coche E (2020) Benefit of
dual-layer spectral CT in emergency imaging of dif-
Disclosure of Interests 
All authors are Philips ferent organ systems. Clin Radiol
employees. Ehn S et al (2017) Assessment of quantification accuracy
and image quality of a full-body dual-layer spectral
Ethical Approval  This article does not contain any stud- CT system. J Appl Clin Med Phys 19(1):204–217
ies with human participants performed by any of the Faller FK et  al (2020) Pre-clinical evaluation of dual-­
authors. layer spectral computed tomography-based stopping
power prediction for particle therapy planning at the
Heidelberg Ion Beam Therapy Center. Phys Med Biol
Ethical Approval  This article does not contain any stud- 65(9):095007
ies with animals performed by any of the authors.
Dual-Energy: The Philips Approach 43

Fulton N et al (2017) Abdominal applications of a novel Nadjiri J et  al (2018) Accuracy of calcium scoring
detector-based spectral CT.  Curr Probl Diagn Radiol calculated from contrast-enhanced coronary com-
47(2):110–118 puted tomography angiography using a dual-layer
Große Hokamp N et  al (2017a) Reduction of artifacts spectral CT: a comparison of calcium scoring
caused by orthopedic hardware in the spine in spectral from real and virtual non-contrast data. PLoS One
detector CT examinations using virtual monoenergetic 13(12):e0208588
image reconstructions and metal-artifact-reduction Nagayama Y et al (2017) Dual-layer DECT for multipha-
algorithms. Skeletal Radiol 47(2):195–201 sic hepatic CT with 50 percent iodine load: a matched-­
Große Hokamp N et al (2017b) Assessment of arterially pair comparison with a 120 kVp protocol. Eur Radiol
hyper-enhancing liver lesions using virtual monoener- 28(4):1719–1730
getic images from spectral detector CT: phantom and Neuhaus V et al (2017a) Improvement of image qual-
patient experience. Abdom Radiol 43(8):2066–2074 ity in unenhanced dual-layer CT of the head using
Große Hokamp N et  al (2018) Artifact reduction from virtual monoenergetic images compared with
dental implants using virtual monoenergetic recon- polyenergetic single-energy CT.  Invest Radiol
structions from novel spectral detector CT.  Eur J 52(8):470–476
Radiol 104:136–142 Neuhaus V et al (2017b) Metal artifact reduction by dual-­
Große Hokamp N et al (2019) Stratification of pulmonary layer computed tomography using virtual monoener-
nodules using quantitative iodine maps from dual-­ getic images. Eur J Radiol 93:143–148
energy computed tomography. Am J Respir Crit Care Neuhaus V et  al (2017c) Improvement of image quality
Med 199(2):e3–e4 in unenhanced dual-layer CT of the head using virtual
Hickethier T et al (2017) Fourth update on CT angiogra- monoenergetic images compared with poly energetic
phy of coronary stents: in vitro evaluation of 24 novel single-energy CT. Invest Radiol 52(8):470–476
stent types. Acta Radiol 59(9):1060–1065 Neuhaus V et al (2018) Bone marrow edema in traumatic
Hickethier T et  al (2020) Venous-phase chest CT with vertebral compression fractures: Diagnostic accuracy
reduced contrast medium dose: Utilization of spectral of dual-layer detector CT using calcium suppressed
low keV monoenergetic images improves image qual- images. Eur J Radiol 105:216–220
ity. Eur J Radiol 122:108756 Oda S et al (2018) Low contrast material dose coronary
Hojjati M et  al (2017) Quality of routine diagnostic computed tomographic angiography using a dual-
abdominal images generated from a novel detec- layer spectral detector system in patients at risk for
tor-based spectral CT scanner: a technical report contrast induced nephropathy. Br J Radiol 92(1094):
on a phantom and clinical study. Abdom Radiol 20180215
42(11):2752–2759 Ohira S et al (2019) Estimation of electron density, effec-
Hua C-H et al (2018) Accuracy of electron density, effec- tive atomic number and stopping power ratio using
tive atomic number, and iodine concentration deter- dual-layer computed tomography for radiotherapy
mination with a dual-layer dual-energy computed treatment planning. Phys Med 56:34–40
tomography system. Med Phys 45(6):2486–2497 Rajiah P et  al (2017a) Spectral detector CT for car-
Jamali S et al (2019) Virtual unenhanced phase with spec- diovascular applications. Diagn Interv Radiol
tral dual-energy CT: is it an alternative to conventional 23(3):187–193
true unenhanced phase for abdominal tissues? Diagn Rajiah P et  al (2017b) Benefit and clinical significance
Interv Imaging 100(9):503–511 of retrospectively obtained spectral data with a novel
Kaltsz K et  al (2017) Noise characteristics of virtual detector-based spectral computed tomography – initial
monoenergetic images from a novel detector based experiences and results. Clin Imaging 49:65–72
spectral CT scanner. Eur J Radiol 98:118–125 Rassouli N et al (2017) Detector-based spectral CT with
Kikano EG et  al (2020) Utility of iodine density perfu- a novel dual-layer technology: principles and applica-
sion maps from dual-energy spectral detector CT in tions. Insights Imaging 8(6):589–598
evaluating cardiothoracic conditions: a primer for the Reimer RP et  al (2019) Virtual monoenergetic images
radiologist. Am J Roentgenol 214(4):775–785 from spectral detector CT enable radiation dose
Lennartz S et  al (2018) Dual-layer detector CT of the reduction in unenhanced cranial CT.  AJNR Am J
head: initial experience in visualization of intracranial Neuroradiol 40(10):1617–1623
hemorrhage and hypodense brain lesions using virtual Sauter AP et  al (2018) Dual-layer spectral computed
monoenergetic images. Eur J Radiol 108:177–183 tomography: virtual non-contrast in comparison to
Lennartz S et  al (2019) Diagnostic value of spectral true non-contrast images. Eur J Radiol 104:108–114
reconstructions in detecting incidental skeletal mus- Sellerer T et al (2018) Dual-energy CT: a phantom com-
cle metastases in CT staging examinations. Cancer parison of different platforms for abdominal imag-
Imaging 19(1):50 ing. Euro Radiology 28(7):2745–2755. https://doi.
Liu J-J et al (2019) Improved visualization of gastric can- org/10.1007/s00330-­017-­5238-­5
cer and increased diagnostic performance in lesion Tsang DS et  al (2017) Quantifying potential reduction
depiction and depth identification using monoener- in contrast dose with monoenergetic images synthe-
getic reconstructions from a novel dual-layer spectral sized from dual-layer detector spectral CT. Br J Radiol
detector CT. Acad Radiol 27(6):e140–e147 90(1078):20170290
44 A. Altman et al.

Van Hedent S et al (2018) Effect of virtual monoenergetic detector spectral CT imaging in unilateral and bilat-
images from spectral detector computed tomogra- eral total hip prostheses. Eur J Radiol 88:61–70
phy on coronary calcium blooming. J Comput Assist Yoon JH et al (2020) Double low-dose dual-energy liver
Tomogr 42(6):912–918 CT in patients at high-risk of0020HCC. Invest Radiol
Van Ommen F et al (2018) Image quality of conventional 55(6):340–348
images of dual-layer SPECTRAL CT: a phantom Zhang X et al (2020) Evaluating the temporomandibular
study. Med Phys 45(7):3031–3042 joint disc using calcium-suppressed technique in dual-­
Wellenberg RHH et al (2017) Quantifying metal artefact layer detector computed tomography. J Int Med Res
reduction using virtual monochromaticdual-layer 48(3):300060519891332
 Dual-Energy: The GE Approach

Scott Slavic and Mats Danielsson

Contents
1 Background   46
2 Balanced System Design   46
3 Image Reconstruction   48
4 Projection-Based Material Decomposition   48
5 I mage Generation and Post-processing   49
5.1  Image Types   49
5.2  Noise Suppression   52
6 Workflow and Clinical Processing   53
7 Spectral Applications and Future   55
8  hoton Counting: The Next Leap in Spectral Imaging 
P  56
8.1  Photon Counting with Deep Silicon   57
8.2  X-Ray Detection Efficiency for Deep Silicon   58
8.3  X-Ray Scatter in the Detector   58
8.4  Count Rate Performance   59
8.5  Deep Silicon Summary   61
References   61

Abstract

This chapter provides an overview of the GE

Dual-Energy Approach to Spectral Imaging
S. Slavic (*)
known as Gemstone Spectral Imaging (GSI).
GE Healthcare, Waukesha, WI, USA Since 2010 when GE launched GSI on
e-mail: [email protected] Discovery™ CT750 HD, GE has provided
M. Danielsson continued improvements to the technology,
Department of Physics, KTH Royal Institute of capabilities, and workflow, as well as spectral
Technology, AlbaNova University Center, clinical tools to support advances in diagnosis
Stockholm, Sweden
e-mail: [email protected]
and solutions across the Discovery™ GSI,

© The Author(s), under exclusive license to Springer Nature Switzerland AG 2022 45

H. Alkadhi et al. (eds.), Spectral Imaging, Medical Radiology Diagnostic Imaging,
https://doi.org/10.1007/978-3-030-96285-2_4
46 S. Slavic and M. Danielsson

Revolution™ CT Frontier, and Revolution™ acquisitions, and full 50 cm field of view. It also
CT/Revolution™ Apex platforms. presents some implementation challenges.
The GSI acquisition method enables pre- A balanced system design must enable rapid
cise temporal registration of the dual-energy kVp rise and fall1 while achieving sufficient
sinograms, projection-based material decom- energy separation and angular view sampling
position, and delivers a full 50  cm material within the constraints of medical diagnostic rota-
decomposition scan field of view. tion speeds. Detector primary decay and after-
The technologies employed to achieve the glow performance are critical to avoiding spectral
dual-energy acquisitions are detailed in the blurring between views. Concurrently, there
discussion of balanced system design below. needs to be a strategy for dose efficiency through
Calibration of fast kVp switching data, mate- balancing the flux between the two spectrums
rial decomposition, and visualization of the and noise reduction processing, and advanced
resulting images are covered in the image reconstruction techniques to ensure image qual-
reconstruction, projection-based material ity across resolution, spectral performance, noise,
decomposition, and post-processing/clinical texture, and other image quality metrics.
applications sections. The chapter closes with Gemstone spectral imaging (GSI) is based on
GSI implementation in the context of work- projection-based material decomposition. The
flow, imaging improvements, solutions to mixed kVp sinogram is transformed into view-­
challenging diagnostic applications, and the aligned low and high kVp sinograms. Material
path to the future of photon counting. basis decomposition is performed on the paired
Academic research and clinical exploration kVp measurements accounting for the tube spec-
of GSI have grown rapidly and resulted in rich trum, bowtie filter, and beam hardening proper-
scientific publications and extensive clinical ties of the basis materials. The noise correlation
adoption. GSI has shown benefits in: of the resulting material density images is well
understood and employed in noise reduction pro-
• Better lesion characterization by providing cessing (Alvarez and Seppi 1979) and advanced
information about the chemical composition deep learning techniques.
and material characteristics While the technology to enable dual-energy is
• Improving lesion detection with enhanced important, image quality, clinical workflow, and
contrast-to-noise ratio clinical applications to provide the right images
• Reducing beam hardening and metal artifacts and clinical answers are equally critical to ensure
• Optimizing iodine load in contrast enhanced adoption and clinical acceptance.
CT studies Future research and development into spectral
imaging continues to expand and new technolo-
gies are in development in the dual-energy space
and photon counting space (Chandra and Langan
1 Background 2011).

Dual-energy in CT has grown since vendors

introduced it in the early 2010s. Relative to con- 2 Balanced System Design
ventional single kVp imaging, dual-energy imag-
ing has the capability to enhance material The fundamental principle behind fast kVp
differentiation and reduce beam hardening arti- switching is the acquisition of two different
facts. Fast kVp switching, where the kVp alter-
nates between low and high kVp for every view
has several benefits including fine temporal view Since the kVp rise and fall are incorporated into the view
1 

measurements, a fast switching generator is required, and

registration (high temporal consistency between any remaining low level nonidealities are accounted for in
high and low energy views), helical and axial the spectral calibration of the data.
Dual-Energy: The GE Approach 47

kVps, alternating on a view-by-view basis

between low and high kVp in a single rotation.
This enables precise temporal registration of
views, thereby freezing motion as the alternating
spectrums penetrate the patient. This signifi-
cantly reduces image artifact. For decades, the
barrier preventing successful clinical introduc-
tion has been the limitations in the acquisition
and detection systems.
The advancements in system components
comprising GE scanners make GSI possible.
Both high and low energy data sets are acquired
simultaneously for axial and helical acquisitions
at the full 50 cm field of view.
The generator and tube are capable of reliably
switching between 80 and 140  kVp settings
within microseconds to achieve 0.25  ms cycle
time (Fig. 1). This capability has been achieved
by a patented high frequency generator and tube
with novel generator control hardware and firm- Fig. 1  Illustration of ultra-fast kV switching. The views
ware, a low capacitance interface, and objected at high and low kVp levels are indicated in red
microsecond-­level tube electrostatic focal spot and green, respectively
size and position control.
The Gemstone detector is a key contributor to
fast kVp switching acquisitions through its scin-
tillator and data acquisition system (DAS). GE mA
Healthcare’s (GEHC) scintillator material, Switching
referred to as Gemstone, is a complex rare earth-­
based oxide, which has a chemically replicated kV
garnet crystal structure. This lends itself to imag- Switching
ing that requires high light output, fast primary
speed, very low afterglow,2 and almost undetect-
Fig. 2  Simultaneous kVp and mA switching. In this
able radiation damage. example, the mA value used for the high kVp views is
Gemstone™ has a primary decay time of lower than for the low kVp views
30  ns, making it 100 times faster than GOS
(Gd2O2S), while also having afterglow levels that In order to combat the traditional flux issues
are lower than 25% of GOS levels, making it that have challenged fast kVp switching, low
ideal for fast sampling. The capabilities of the kVp and high kVp acquisitions are flux balanced
scintillator are matched with a fast-sampling through advances in the DAS and generator via
DAS, enabling simultaneous acquisition of low two methods: dynamically changing view inte-
and high kVp sinograms at customary rotation gration times, and more recently, near instanta-
speeds. neous modulation of both kVp and mA (Fig. 2).
Dynamic view integration times are accom-
Afterglow refers to a secondary decay of light emitting
2  plished by allocating additional time to the low
from the scintillator for several milliseconds after the kVp acquisition relative to the high kVp acquisi-
X-ray source is turned off. It carries a part of the signal tion in order to reduce the possibility of photon
from one view to the next during a scan, thereby smearing
the information, and potentially causing unwanted spec- starvation conditions. Coupled with the appropri-
tral decomposition artifact. ate rotation speed, a more balanced flux condi-
48 S. Slavic and M. Danielsson

tion between the two kVp scans is achieved and with

this serves to minimize patient dose. Near instan-  b  E ,d  Ib  d 
Gk  d   Sk  E  E 1  e
 d  E  td
taneous kVp and mA switching is accomplished e b
dE
 
via advanced cathode technologies that allow for E

very fast control of both kVp and mA. where μd(E) is the linear attenuation coefficient of
By means of flux balance methodologies and the detectors, td is detector thickness, μb(E,d) and
noise reduction techniques, GSI is designed to Ib(d) are respectively the linear attenuation coeffi-
minimize additional dose relative to single-­ cient and the thickness of bowtie material b corre-
energy scans (Chandra and Langan 2011). sponding to detector channel d. R(d) can be
Studies performed between 2012 and 2014 measured through a fast-switching air scan, and
demonstrated abdominal GSI doses of between Gk(d) can be calculated based on the system geom-
12.8 and 21.8 mGy. These values are below the etry. The problem now becomes one of solving for
ACR’s 25 mGy reference dose level for a single-­ R(d). This is an overdetermined problem and can
phase abdominal CT. The noise levels with GSI be easily solved by least squares fitting. Once the
were comparable to those with SECT (Dubourg calibration corrections have been applied to the
et al. 2014; Lin et al. 2012; Schuman et al. 2014). low and high kVp data sets, they are aligned in
projection space, transformed into a material basis
pair projection (such as water and iodine), and
3 Image Reconstruction then reconstructed. These material density images3
may be combined to create a monochromatic
Following the acquisition, calibration corrections image at any specific keV level. A pictorial flow is
are applied to the data. Spectral calibration is presented in Fig. 3 (Chandra and Langan 2011).
complicated by the nonideal kVp rise and fall
making it difficult to find a fixed kVp having pre-
cisely the same spectral response as an actual fast 4 Projection-Based Material
kVp switching energy spectrum. As a result, the Decomposition
spectrum is fitted to a linear combination of single
kVp spectra. The overall spectrum (Sp) is decom- Dual-energy material decomposition is based
posed into a superposition of several known kVp upon the mass attenuation coefficient across the
(p) spectra through the measurement of the detec- medical diagnostic imaging spectrum being a
tor response to the bowtie-attenuated beam: function of two independent variables: attenua-
Nk tion due to the photoelectric effect and that due to
S p  E    k Sk  E  , Compton scatter (Alvarez and Macovski 1976).
k Through a mathematical change of basis one can
where Sk(E) are the basis spectra of the fixed express the energy-dependent attenuation
kVps, Nk is the total number of the basis spectra, observed in two kVp measurements in terms of
and the αk are the weights of the basis spectra. two basis materials:
The self-normalized detector response to this
spectrum can be written as: Material density images represent the effective density
3 

for the material necessary to create the observed kVp

R d  
  G d  ,
k k k measurements. For example, pure water appears as

   G d 
d k k k
1000 mg/mL in a water image, 20 mg/mL of dilute iodine
is labeled as such an iodine image.
Dual-Energy: The GE Approach 49

p  I
low  ln 

    
  ln   Slow  E  exp   m1 1  E   m2 2  E   dE /  Slow  E  dE 

 Io low

p  I
high  ln 

    
  ln   Shigh  E  exp   m1 1  E   m2 2  E   dE /  Shigh  E  dE 

 Io high

where I represents the attenuated X-ray spec- measured with a monoenergetic X-ray source.
trum, Io represents the reference spectrum, μ1(E) For consistency with the Hounsfield unit, one can
and μ2(E) represent the mass attenuation coeffi- normalize the attenuation measurement with
cients of the basis materials, and m1 and m2 are respect to water (Chandra and Langan 2011).
their respective effective densities. Slow(E) and
Shigh(E) are defined by the source spectrum,
source filtration, and detector performance. The 5 Image Generation
solution for m1 and m2 accounts for spectral varia- and Post-processing
tion over the field of view due to the bowtie filter,
and multi-material beam hardening. As a conse- 5.1 Image Types
quence, projection-based material decomposition
provides the opportunity for more quantitative 5.1.1 Monochromatic
precision than may be achieved with single kVp Monochromatic images, ranging from 40 to
imaging. 140  keV, depict objects as if they were imaged
with a theoretical monochromatic beam whose
 I  X-ray energy is measured in kiloelectron volts
p   ln    m1 1  E   m2 2  E  (keV) instead of peak kilovoltage (kVp).
 Io  Monochromatic images with lower energy
Given the material basis density images, one levels generally improve the contrast-to-noise
can compute attenuation data (p) that would be ratio (CNR) and can even achieve higher CNR

Data Acquisition Projection Space Image Space

GSI Data Acquisition
Interleaved High and Monochromatic Images
low kVp projections 40-140 keV

Split
Image
Reconstruction
Material Density Images
(Water, Iodine, Calcium, HAP,
Uric Acid, Fat)

Material Density
Transformation (with
MAR Correction if GSI Image
MAR is applied) Reconstruction

VUE

Fig. 3  Schematic illustration of the GSI Material Decomposition process

50 S. Slavic and M. Danielsson

Fig. 4  A 30  cm  ×  40  cm QRM phantom, simulating a the liver ROI against the background was automatically
large patient, was scanned with 80  mm GSI helical at a computed using the GSI Viewer Optimal CNR tool
CTDIvol of 24 mGy. The contrast-to-noise ratio (CNR) of

than single-energy imaging at the same dose. images are proportional to material density (con-
This is because monochromatic images at lower centration) and are expressed in mg/ml.
energy levels (40–70  keV) are closer to the GSI can detect iodine contrast in concentra-
K-edge of iodine (33.2  keV), at which iodine tions as low as 0.5 mg/ml in density at a dose as
exhibits much higher attenuation as compared to low as 8  mGy.4 Iodine detectability is a perfor-
conventional single-energy CT at 120 kVp. mance metric that is key to the clinical value of
Figure  4 shows the increased CNR at lower dual-energy CT.
keV.  A big advantage over single-energy scan- As Patino/Sahani et  al. (Patino et  al. 2016)
ning is the availability of the range of keV images summarized, material-specific images generated
from a single exposure vs. single energy where by spectral CT expand the current role of CT and
the CNR is fixed based on the technique chosen. overcome several limitations of single-energy
Clinical studies confirm that low energy CT.
monochromatic images (40–70  keV) can depict The selection of the optimal material pairs for
more subtle contrast enhancement by improving reconstruction is based on the specific clinical
the CNR between a lesion and background paren- task.
chyma13. Monochromatic images provide more The most common material pair used clini-
reliable attenuation values than conventional cally is iodine and water, because the iodine
polychromatic CT images. (water) image can be used to assess iodine distri-
bution, to increase tissue contrast, and to amplify
subtle differences in attenuation between normal
5.1.2 Material Density and abnormal tissues. This capability facilitates
Material density (MD) images are generated improved lesion detection and characterization,
natively or via post-processing from spectral data tumor viability quantification, and treatment
and provide qualitative and quantitative informa- response monitoring (Patino et al. 2016; Agrawal
tion regarding tissue composition and contrast et al. 2014). Figure 5 illustrates how a color iodine
media distribution (Patino et al. 2016) overlay can aid in identifying cancer lesions.
MD images are generated as a material basis In chest imaging, iodine (water) images that
pair (e.g., water/iodine, uric acid/calcium). For represent the iodine distribution in the lung
example, when the water and iodine pair is gener- parenchyma (an example of which appears in
ated, the iodine images visualize the density of the Fig.  6) can also help to identify pulmonary
object with suppressed water information (written embolism-­ associated perfusion defects, espe-
as “iodine (water)”) and water images visualize
the density of the object with suppressed iodine
Detectability demonstrated on Revolution™ CT and
4 

information (written as “water (iodine)”). Note Revolution™ Apex Products. Detection of 0.5 mg/mL at
that the measured pixel intensity values in these 8 mGy was demonstrated in head phantom testing
Dual-Energy: The GE Approach 51

a b

Fig. 5  A patient with recurrent liver carcinoma. (b) Iodine color overlay on 55 keV can better visualize recurrent cancer
lesions (arrows) than (a) 55 keV image

cially in patients with underlying perfusion

abnormalities (Bauer et  al. 2011; Pontana et  al.
2008).
Fat (iodine) or fat (water) images can be used
to characterize fat content in abdominal lesions,
such as cholesterol stones in the gallbladder
(Fig. 7) and fat-containing renal lesions. Morgan
et al. demonstrated that fat (iodine) images per-
mit differentiation between low- and high-fat-­
containing adrenal lesions with a high specificity
of 94% (Morgan et al. 2013).

5.1.3 Material-Suppressed Iodine

Fig. 6  A patient with pulmonary embolus. An iodine
color overlay on this 70 keV image aids in the identifica- Spectral information in GSI Images may be uti-
tion of embolus (arrow) and the area of lung perfusion lized to generate virtual unenhanced images
abnormality (arrow heads) (VUE) by subtracting iodine from the images.
The VUE algorithm is based on multi-material
decomposition (MMD), which replaces the vol-
ume fraction of contrast by the same volume
fraction of blood, producing iodine-suppressed
images. The VUE images provide attenuation
information in Hounsfield units. Figure 8 shows
an example of a non-contrast scan acquired at
120 kVp, and GSI contrast enhanced 70 keV and
VUE images derived from a GSI contrast
enhanced scan. The iodine filled vessels are
removed in the VUE images.
VUE Imaging has shown promise in clinical
Fig. 7  A patient with gallbladder cholesterol stone. This
use cases where a True unenhanced (TUE) image
fat (water) image characterizes the fat contents of the
stone (arrows) is typically used (Mileto et al. 2021).
52 S. Slavic and M. Danielsson

a b

Fig. 8  Examples of (a) a non-contrast image acquired at 120 kVp, and (b) GSI contrast enhanced 70 keV image, (c)
VUE image derived from GSI contrast enhanced scan

5.1.4 Metal Artifact Reduction reduction used in conjunction with higher keV set-
Patients imaged with CT routinely have metal tings (>100 keV) that also provide beam hardening
implants which can cause artifacts due to photon reduction has allowed GSI and GSI MAR to realize
starvation, beam hardening, and scatter. GSI Metal enhanced metal artifact reduction that exceeds the
Artifact Reduction (GSI MAR) is a dual-energy performance of both of conventional MAR, and the
metal artifact reduction algorithm designed to reveal metal artifact suppression achieved in high keV vir-
anatomic details obscured by metal artifacts. GSI tual monoenergetic images.
MAR can reduce metal artifacts using a three-stage
correction to address all three factors. This approach
generates metal-­corrected images, while preserving 5.2 Noise Suppression
spatial resolution and data integrity near the metal
(Girijesh et al. 2014; Pal et al. 2013a, b; Hsieh et al. Noise suppression is very important due to flux
2013). The output sinograms are used as input to the challenges with the low kVp samples in any
improved spectral imaging chain to generate the multi-energy technology. To address this issue,
final GSI MAR images (Fig.  9). Metal artifact GE has used two methodologies to date in GSI
Dual-Energy: The GE Approach 53

a b c

Fig. 9  A patient with metal denture and mouth floor car- iodine color overlay with GSI MAR can reduce metal arti-
cinoma. (a) 60 keV image shows the artifacts caused by facts and reveal enhanced mouth floor carcinoma
the metal denture. (b) 60  keV with GSI MAR and (c)

imaging. Noise suppression in GSI by virtue of increase the noise reduction capability while
the projection-based material decomposition is maintaining and improving other imaging met-
applied in the material density space in order to rics such as texture and resolution. TrueFidelity™
enhance image quality without shifting mean val- was designed to produce image quality perfor-
ues. This allows for noise suppression techniques mance not easily achievable by conventional ana-
to propagate to all image types: Virtual lytical and iterative techniques. Rather than
Monochromatic, Material Density, Metal manually optimizing the reconstruction over
Artifact-Reduced Images, and Post-Processed many parameters to balance image noise, spatial
images such as Virtual Unenhanced images. resolution, overall texture, and other quality met-
rics across a multitude of clinical imaging sce-
5.2.1 Iterative Reconstruction narios, the deep learning approach uses a neural
The initial noise suppression method deployed network to learn the desired characteristics of the
with GSI leverages the statistical iterative recon- reconstructed images.
struction approach taken by ASIR5 (Fan et  al. While both noise reduction techniques repre-
2010) while simultaneously leveraging well-­ sent clinically relevant and robust performance,
known noise correlation properties of the projec- TrueFidelity™ GSI represents a step change in
tion space material decomposition process overall image quality. Examples of noise sup-
(Alvarez and Seppi 1979). pression and the improvements made from itera-
tive reconstruction and deep learning are
5.2.2 TrueFidelity™ GSI presented in Figs. 10, 11, 12 and 13.
Recent advancements in deep learning have led
to the application of deep learning techniques to
6 Workflow and Clinical
ASIR focuses on the statistical modeling of the noise
5  Processing
properties of the system in conjunction with the properties
of the scanned object. As a result, it provides significant GE’s GSI Workflow has evolved with the
benefit for those examinations that may experience limita- approach to assist the technologist through the
tions due to noise in the reconstructed images. In the case
workflow to ensure dose and image quality are
of spectral imaging, this is applied to reduce the noise in
the material density images to enhance image quality. optimized, organize the output images based on
54 S. Slavic and M. Danielsson

Fig. 10  Iterative reconstruction and deep learning noise suppression—monochromatic images. (Image courtesy of
Houston Methodist)

Fig. 11  Iterative reconstruction and deep learning noise suppression—material density images. (Image courtesy of
Houston Methodist)

clinical function, and transfer the relevant images • GSI Assist: personalizes GSI scan parameters
to PACS for the initial read. The images still con- to clinical indications and the anatomy of the
tain spectral information for later analysis and patient.
post-processing as needed on advanced worksta- • Xtream Recon: reconstructs GSI specific
tions (Slavic et al. 2017). images natively on the console with real-time
Key features that allow this workflow include: reconstruction speed, using parallel p­ rocessing
of the spectral information to greatly speed up
• GSI Profiles: available at a higher level in pro- the overall recon process.
tocol management to standardize, automate, • Direct transfer to PACS: offers the direct
and personalize GSI scan and recon parame- transfer of all natively reconstructed GSI
ters based on clinical needs. images to PACS.
Dual-Energy: The GE Approach 55

Fig. 12  Iterative reconstruction and deep learning noise suppression—virtual unenhanced images. (Image courtesy of
UZ Brussels)

Fig. 13  Iterative reconstruction and deep learning noise suppression—monochromatic images with metal artifact
reduction. (Image courtesy of Froedtert Hospital)

These workflow features allow the user to auto- 7 Spectral Applications

mate scanning, dose, and reconstructions to allow and Future
seamless usage in clinical workflow. The custom-
ization capability allows users to customize the Workflow, PACS clinical tools continue to
strategy per their clinical needs (Gauntt 2019) advance, facilitating spectral imaging solutions
Figure 14 shows how GSI Profiles and GSI and leading to widespread adoption. With GE,
Assist automate and assist in the GSI workflow. this spans the Advantage Workstation (AW) and
Figure 15 represents the Xtream Recon flow par- the Edison Platform to provide viewing, analysis,
allelizing the reconstruction process and allow- and analytics platforms to support clinical
ing the fast direct-to-PACS methodology. workflows.
56 S. Slavic and M. Danielsson

Fig. 14  GSI Assist/Profiles workflow

Material Decomposition Monochromatic images

Prep
Iterative Recon (40-140keV)
PACS
Material Density images
and/or
(Iodine, Water, Calcium,
post-
Uric Acid, HAP, Fat)
processing
workstations

VUE

Fig. 15  Illustration of Xtream Recon technology and tion iterative recon process. From this juncture multiple
direct transfer to PACS. An intermediate stage is created GSI DICOM images can be natively reconstructed
after projection data preparation and material decomposi-

With spectral imaging, in addition to the image spectral analysis, multi-material decomposition,
types, viewers and applications are able to utilize advanced segmentation, and colorization of
spectral data for quantitative analysis spanning, images to help automate clinical outputs and
material assessment, effective Z imaging and his- clinical findings.
tograms, spectral response curves, contrast-to-
noise (CNR) curves, overlays of spectral data, and
post-processed imaging. This tool suite continues 8 Photon Counting: The Next
to develop, provide more automated solutions, and Leap in Spectral Imaging
incorporate advanced imaging (AI) techniques.
Figures 16, 17 and 18 represent examples of The next revolutionary step in spectral imaging is
clinical outputs from GE clinical tools. These photon counting. GE is pursuing a novel approach
advanced tools incorporate a combination of in this area.
Dual-Energy: The GE Approach 57

Fig. 16  Abdomen colorization of iodine in liver lesions

8.1  hoton Counting with Deep

P image cases. This increases usability. Deep sili-
Silicon con and other photon-counting solutions also
offer significantly higher spatial resolution com-
Photon counting has emerged in recent years as a pared to state-of-the-art.
future alternative to dual-energy CT imaging. In Silicon is the number one industry standard
photon counting, each X-ray is counted sepa- semiconductor substrate and very large volumes
rately. The energy of each X-ray is measured of high purity crystalline material are produced
using a number of programmable thresholds. In in a multitude of fabricators around the world.
contrast to dual-energy CT, the number of energy Due to the enormous investments in silicon pro-
levels can be selected in the design of the front-­ duction over many years, it is basically impossi-
end electronics. For the deep silicon solution, we ble for any other material to compete in terms of
have eight threshold levels, meaning in principle cost or quality. A proposed alternative to silicon
that eight different materials can be separated in is to use a cadmium-based detector, either cad-
a so-called material basis decomposition. mium telluride (CdTe) or cadmium zinc telluride
Typically, only three materials will be recon- (CZT). Those materials exist in prototype vol-
structed, but having more thresholds increases umes, and an additional challenge is the environ-
dose efficiency and also means there is no need to mental and health hazards associated with
change energy threshold levels for different cadmium (World Health Organization 1992).
58 S. Slavic and M. Danielsson

Fig. 17  Automated and advanced segmentation of gout crystals

range. Another factor that will impact the dose

efficiency is the non-active “dead” layer at the
surface, since there is never a perfect crystal
structure all the way to the edge of the crystal and
charge collection will be limited in this region. A
low atomic number means that the sensitivity to
any dead layer at the entrance surface of the
detector is reduced. In Fig. 19, we show that the
overall detection efficiency (DE) is very high.

8.3 X-Ray Scatter in the Detector

Fig. 18  Automated lung segmentation and colorization
for pulmonary embolism studies An advantage of silicon is that its low atomic
number means there is a negligible probability
8.2  -Ray Detection Efficiency
X for K-fluorescence, which deteriorates both
for Deep Silicon energy resolution and spatial resolution, since
energy deposition for one interaction is scat-
Since silicon has a low atomic number, a signifi- tered over several pixels by the K-fluorescence
cant thickness is required to efficiently absorb photons. On the other hand, the cross section
incoming high energy X-rays in the diagnostic for Compton scatter is significant. All Compton
Dual-Energy: The GE Approach 59

Fig. 19  X-ray detection efficiency for different detector geometries for photon-counting CT

scatter energy depositions above the minimum tons are efficiently absorbed by the tungsten
threshold will be counted and will contribute to and have a negligible impact on spatial
the image. Energy deposited by Compton inter- resolution.
actions will be recorded in the low energy bins,
which allows these to be distinguished from the
photoelectric interactions, which tend to be 8.4 Count Rate Performance
recorded in the higher energy bins. For X-rays
with a primary Compton interaction, the energy One significant challenge with photon-counting
resolution will be reduced since all energy is detectors has been the very high count rates,
not deposited. For X-rays with a primary photo where the input flux can be up to 109 photons/
interaction, all the energy is deposited, and the mm2/s. In deep silicon, the depth of the silicon
energy resolution will be very high due to the can be segmented into sub-voxels to decrease the
high inherent energy resolution in silicon. count rate for each input channel. With a design
Intersecting tungsten foils stop the scattered using nine depth strata (with longer strata deeper
photons and work as a general anti-scatter col- in the detector), the count rate can be reduced by
limator without loss of geometrical efficiency. the same factor. The design shown in Fig. 22 out-
In Fig. 20, we outline the distributions of differ- lines the depth strata together with the measured
ent interactions in deep silicon for a 140  kVp count rate response as a function of the X-ray1.
spectrum, which is the worst-case diagnostic High count rate performance translates into
spectrum in that it produces the most Compton reduced detective quantum efficiency (DQE) loss
interactions. The point spread function in at high input flux and also enables a robust
Fig.  21 confirms that Compton scattered pho- design, far from the point of detector saturation.
60 S. Slavic and M. Danielsson

Fig. 20  Photoelectric and Compton interactions in deep silicon

Fig. 21  Measured point spread function and MTF for deep silicon

Fig. 22  Count Rate Performance and diagram of multi strata design
Dual-Energy: The GE Approach 61

8.5 Deep Silicon Summary Alvarez RE, Seppi E (1979) A comparison of noise
and dose in conventional and energy selec-
tive computed tomography. IEEE Trans Nucl Sci
In summary, the deep silicon approach to pho- NS-26(2):2853–2856
ton counting allows for the use of an industrial Bauer R, Frellesen C, Renker M et al (2011) Dual energy
standard material with extremely high purity CT pulmonary blood volume assessment in acute pul-
monary embolism: correlation with D-dimer level,
and crystal lattice quality. The depth of the sili-
right heart strain and clinical outcome. Eur Radiol
con will result in high detection efficiency and 21(9):1914–1921
enable depth strata for high and robust perfor- Chae EJ, Seo JB, Jang YM et al (2010) Dual-energy CT
mance for high X-ray flux imaging cases. X-rays for assessment of the severity of acute pulmonary
embolism: pulmonary perfusion defect score com-
that are Compton scattered in the detector are pared with CT angiographic obstruction score and
absorbed by intersecting tungsten foils that also right ventricular/left ventricular diameter ratio. AJR
work as a collimator for object scatter. Charge 194:604–610
collection in silicon is very fast, which trans- Chandra N, Langan DA (2011) Gemstone detector: dual
energy imaging via fast kVp switching. In: Johnson T,
lates into low diffusion of charges between pix- Fink C, Schönberg S, Reiser M (eds) Dual energy CT in
els, resulting in high energy resolution. This in clinical practice, Medical radiology. Springer, Berlin,
turn leads to high quantification accuracy and Heidelberg. https://doi.org/10.1007/174_2010_35
precision (for example, iodine concentration), Danielsson M, Persson M, Sjölin M (2021) Photon-­
counting x-ray detectors for CT.  Phys Med Biol
and to dose-efficient three-material decomposi- 66(3):03TR01
tion (Xu et al. 2013). Dubourg B, Caudron J, Lestrat J et  al (2014) Single-­
The deep silicon concept will deliver high spa- source dual-energy CT angiography with reduced
tial resolution combined with high energy resolu- iodine load in patients referred for aortoiliofemoral
evaluation before transcatheter aortic valve implanta-
tion and spectral performance for tissue tion: impact on image quality and radiation dose. Eur
differentiation and quantification (Danielsson Radiol 24:2659–2668
et al. 2021). One example is the evaluation of the Fan J, Hsieh, J, Sainath P, Crandall PS (2010) Head and
degree of stenosis in blood vessels, when so-­ Body CTDIw of dual energy x-ray CT with fast kVp
switching. Paper 7622–69, SPIE Medical Imaging,
called “blooming artifacts” can be avoided and San Diego
separation between calcium and iodine signifi- Gauntt D (2019) A suggested method for setting up
cantly improved. Another example is detection GSI profiles on GE Revolution CT Scanner. Med
and follow-up of cancer. Imaging:169–179. https://doi.org/10.1002/acm2.12754
Girijesh K, Pal D, Hsieh J (2014) Reduction of metal arti-
facts: beam hardening and photon starvation effects.
Compliance with Ethical Standards Proc. SPIE 9033, Medical imaging 2014: physics of
medical imaging, p 90332V
Disclosure of Interests  Scott Slavic is an employee of Hsieh J et  al (2013) Recent advances in CT image
GE Healthcare. Mats Danielsson is a consultant of GE reconstruction. Technical report, General Electric
Healthcare. Healthcare Company
Lin X, Wu Z, Tao R et al (2012) Dual energy spectral CT
Ethical Approval In this review article, the authors imaging of insulinoma value in preoperative diagnosis
obtained images from sites evaluating GE Healthcare compared with conventional multi-detector CT. Eur J
devices with permission from the institutions. This article Radiol 81:2487–2494
does not contain any studies with human participants per- Mileto A, Xiao J et  al (2021) Virtual unenhanced dual-­
formed by any of the authors. energy CT images obtained with a multimate-
rial decomposition algorithm: diagnostic value for
renal mass and urinary stone evaluation. Radiology
References 298(3):611–619
Morgan D et al (2013) Differentiation of high lipid con-
tent from low lipid content adrenal lesions using
Agrawal MD et  al (2014) Oncologic applications of
single-source rapid kilovolt (peak)-switching dual-­
dual-energy CT in the abdomen. RadioGraphics
energy multidetector CT.  J Comput Assist Tomogr
R34(3):589–612
37(6):937–943
Alvarez RE, Macovski A (1976) Energy-selective recon-
Pal D, Dong S, Genitsarios I, Hsieh J (2013a) Smart
structions in X-ray computerized tomography. Phys
metal artifact reduction white paper. Technical report,
Med Biol 21(5):733–744
General Electric Healthcare Company
62 S. Slavic and M. Danielsson

Pal D, Sen Sharma K, Hsieh J (2013b) Metal artifact cor- Schuman W, Green D, Busey J et al (2014) Dual-energy
rection algorithm for CT. In: Nuclear science sympo- liver CT: effect of monochromatic imaging on lesion
sium and medical imaging conference (NSS/MIC). detection, conspicuity, and contrast-to-noise ratio of
IEEE, pp 1–4 hypervascular lesions on late arterial phase. AJR Am J
Patino M et  al (2016) Material separation using dual-­ Roentgenol 203:601–606
energy CT: current and emerging applications. Slavic S et al (2017) GSI Xtream on revolution CT. GE
RadioGraphics 36(4):1087–1105 Healthcare, Waukesha
Pontana F, Faivre J, Remy-Jardin M et  al (2008) Lung World Health Organization (1992) Cadmium  – environ-
perfusion with dual-energy multidetector-row CT mental aspects. ECH 135
(MDCT): feasibility for the evaluation of acute pul- Xu C et al (2013) Transactions on nuclear science. IEEE
monary embolism in 117 consecutive patients. Acad 60(1):437–445
Radiol 15(12):1494–1504
 Dual-Energy: The Canon Approach

Kirsten Boedeker, Jay Vaishnav, Ruoqiao Zhang,

Zhou Yu, and Satoru Nakanishi

Contents
1 Introduction   63
2  he Aquilion ONE Prism and Its Technology 
T  64
2.1  Spectral Rapid kV Switching   64
2.2  Spectral Deep Learning Reconstruction   66
3 Spectral Performance   67
3.1  Spatial Resolution and CNR Performance   67
3.2  Wide Volume Detector   68
3.3  Virtual Monochromatic Images   68
3.4  Spectral Iodine Maps   69
3.5  Workflow   71
References   72

1 Introduction kV Switching technology with deep learning-­based

reconstruction to maximize the diagnostic value of
Canon Medical introduced the first multi-energy every photon at each kV level, making routine and
Computed Tomography (CT) system powered by wide volume multi-energy imaging possible.
deep learning artificial intelligence with the launch Spectral Imaging relies on the fundamental
of Spectral Imaging on the Aquilion ONE Prism collection of data at two energies, a high and a
CT system in 2020. Spectral Imaging’s multi- low kV, that can be used to represent two basis
energy approach combines the benefits of Rapid materials, such as iodine and water, for material
decomposition. The Aquilion ONE platform fea-
tures a key design consideration for the optimal
K. Boedeker · J. Vaishnav · S. Nakanishi implementation of multi-energy CT: wide vol-
Canon Medical Systems Corporation, ume coverage. The Aquilion ONE Prism offers
Otawara, Tochigi, Japan
e-mail: [email protected]; jvaishnav@ 16  cm of longitudinal coverage, allowing for
us.medical.canon; [email protected] acquisition of a whole organ, such as the head or
R. Zhang · Z. Yu (*) heart, in a single rotation as well as a 50 cm mid-
Canon Medical Research USA, plane Field of View (FOV) (Fig.  1). Therefore,
Vernon Hills, IL, USA Spectral Imaging is designed to encompass both
e-mail: [email protected]; standard helical and wide volume coverage.
[email protected]

© The Author(s), under exclusive license to Springer Nature Switzerland AG 2022 63

H. Alkadhi et al. (eds.), Spectral Imaging, Medical Radiology Diagnostic Imaging,
https://doi.org/10.1007/978-3-030-96285-2_5
64 K. Boedeker et al.

tral cardiac CT.  Spectral is also designed to be

compatible with the entire 50  cm FOV in both
standard helical modes and for wide volume
modes. Sufficiently large field of view has
become an important factor for CT systems, par-
ticularly when scanning large patients, as some
anatomical structures may be missed with a lim-
ited FOV, such as the periphery of lung, liver, and
bowel.
The Aquilion ONE Prism also utilizes Canon
Fig. 1  The Aquilion ONE Prism offers 16 cm of longitu-
dinal coverage, allowing for acquisition of a whole organ,
Medical’s SUREExposure Automatic Exposure
such as the head or heart, in a single rotation as well as a Control (AEC) to vary the tube current applied
50 cm midplane Field of View (FOV) during a scan, based on the size and composition
of the individual patient being scanned. The pur-
Rapid kV Switching works by collecting pose of AEC is to automatically optimize the CT
interleaved projection views at each energy, gen- scan exposure to reduce overall radiation dose
erated by rapidly toggling between kV levels as while tailoring the dose distribution to yield con-
the tube rotates around the patient. Rapid kV sistent image quality. This technology is particu-
Switching minimizes spatial and temporal mis- larly relevant for radiosensitive subjects, such as
alignment, permitting raw data-based material chest and pediatric patients. Canon Medical’s
decomposition. Combining Rapid kV Switching spectral CT is compatible with the routine use of
with the image quality improvement capabilities AEC in both standard helical and wide volume
of deep learning artificial intelligence permits the scanning modes.
data to be acquired with tube current modulation,
a key feature for optimizing radiation dose man-
agement. By infusing Rapid kV Switching with 2.1 Spectral Rapid kV Switching
the power of deep learning, a full suite of clinical
output—from crisp, clear iodine maps to low Every multi-energy CT system needs to generate
noise virtual monoenergetic images—can be two complete sinograms worth of raw data, one
generated for standard and wide volume CT. at each energy, to perform material decomposi-
tion, i.e., the process of separating the data into
two basis materials. In order to acquire views at
2  he Aquilion ONE Prism
T more than one energy, Rapid kV Switching oper-
and Its Technology ates by quickly and repeatedly switching the
energy of the beam from high to low as the tube
The Aquilion ONE wide volume CT system, first and detector rotate around the patient (Fig.  2).
launched in 2007, offers 16  cm of longitudinal Rapid kV Switching minimizes the temporal
coverage that permits isotropic volumes of an delay between high and low energy measure-
entire organ to be acquired in a single rotation of ments, resulting in images with minimal contrast
the gantry. The Aquilion ONE Prism has 320 delay and motion mismatch. In addition, Rapid
detector rows, each 0.5  mm thick. Canon kV Switching allows for the material decomposi-
Medical’s Spectral CT is designed to combine tion process to take place in the raw data domain
the temporal resolution benefits of Rapid kV itself (Zou and Silver 2008), rather than post-­
Switching with the ability of 16 cm coverage to reconstruction in the image domain. Raw data-­
reduce motion artifact throughout the volume. based decomposition has been demonstrated to
Performing Spectral CT on whole organs be less impacted by beam hardening and other
acquired in a single rotation provides increased biases that occur when material decomposition is
diagnostic potential in applications such as spec- performed in the image domain (Li et al. 2011).
Dual-Energy: The Canon Approach 65

learning-based spectral reconstruction was devel-

oped to overcome these challenges associated
with kV switching.

2.1.1 Deep Learning

Deep learning is a subfield of machine learning
that takes advantage of multi-layered artificial
neural network and represents the current state-­
of-­the-art in Artificial Intelligence. Unlike con-
ventional algorithms that are constrained by
Fig. 2  Rapid kV Switching operates by quickly and
pre-programmed rules for performing a complex
repeatedly switching the energy of the beam from high to
low as the tube and detector rotate around the patient task, deep learning occurs when a neural network
learns from its own intensive training process and
develops its own logic structure. With deep learn-
ing, a Deep Convolutional Neural Network
(DCNN) comprised of layers of neurons is
trained in the performance of a complex task
(Fig. 3). A neuron is a node where a mathemati-
cal operation takes places, the output of which is
connected with other neurons, forming a net-
work. The neural network derives its name from
the neuron-synapse paradigm found in biology
and mimics how humans draw conclusions,
based on learning from examples. As the network
“learns,” the mathematical parameters, known as
weight and bias, applied by each neuron are
adjusted until reaching a target level of perfor-
Fig. 3  A Deep Convolutional Neural Network (DCNN) mance. This ability to learn via a deep neural net-
comprised of layers of neurons is trained in the perfor-
work gives deep learning algorithms the freedom
mance of a complex task. A neuron is a node where a
mathematical operation takes places, the output of which to find the optimum way to perform the desired
is connected with other neurons, forming a network task.
The key to a successful DCNN lies in its train-
ing, the process by which the neural network
Earlier implementations of kV switching learns how to successfully perform its function.
using conventional reconstruction techniques are The network must compare its output to a gold
constrained by the need to acquire a large number standard reference in order to gauge its perfor-
of views at each energy to maintain spatial reso- mance and learn, i.e., adjust the weights and bias
lution. This requires highly specialized hardware of its neurons. To accomplish this the DCNN
that makes implementing dual-energy scanning uses a mathematical loss function to determine
across a variety of platforms difficult. In addition, the amount of error between its output and the
the need to maintain energy separation between reference datasets.
the high and low kV across such a large number Canon pioneered the successful application of
of views requires conventional kV switching to deep learning in CT reconstruction with the intro-
operate at speeds that prohibit the use of AEC for duction of the AiCE Deep Learning
radiation dose management. Furthermore, such Reconstruction in 2018. The implementation of
approaches face significant challenges when the AiCE dramatically lowered image noise over
cone angle of the system is large, as required for conventional reconstruction as the AiCE neural
wide volume coverage. Canon Medical’s deep network is able to distinguish and preserve signal
66 K. Boedeker et al.

while reducing unwanted noise. On the Aquilion which the X-ray beam is attenuated by the patient
ONE Prism, AiCE is able to reduce noise by 32% being scanned via Compton scattering and the
compared to the standard hybrid iterative recon- photoelectric effect.
struction approach. Spectral deep learning reconstruction takes
advantage of the fact a deep learning neural net-
work can be trained to convert the energy-­
2.2 Spectral Deep Learning dependent attenuation information associated
Reconstruction with one kV into that of another kV. Therefore,
spectral reconstruction works by generating Deep
Spectral deep learning reconstruction is a deep Learning Views (DLVs) created by transforming
learning-based reconstruction algorithm that uses attenuation information in views acquired at one
projection views acquired with Rapid kV energy into that of the opposite energy. DLVs are
Switching to perform raw data-based material generated by the trained neural network using
decomposition into basis pairs as well as recon- measured data from both the opposite-energy
struct the resultant images. Spectral reconstruc- views at a particular location and adjacent same-­
tion allows for the highly precise spatial and energy views. The DLVs then complement the
temporal alignment of the high and low energy measured views at each energy to generate a
views associated with Rapid kV Switching, with- complete sinogram for each kV.  Spectral deep
out sacrificing full anatomical coverage or auto- learning reconstruction then performs material
matic exposure control (Zhang et al. 2020). decomposition in the raw data domain and com-
A common challenge to all multi-energy CT pletes the reconstruction process, with a deep
technology is noise amplification of the material learning-based denoising step to create low noise
decomposition process. Computing the basis spectral CT image data.
materials from the two kV measurements requires The neural network used in spectral deep
solving what is referred to in mathematics as an learning reconstruction has a multi-scaled struc-
ill-posed inverse problem, meaning the solution ture, similar to that of U-Net (Ronneberger et al.
and resulting image output is very sensitive to 2015). With multiple layers at different scales,
noise. Recently, deep learning technologies have the network creates a large receptive field and is
been successfully applied in CT reconstruction, capable of capturing a multitude spatial and spec-
demonstrating great noise reduction capabilities tral features in the data. As with all deep learning
and thus well-suited to advancing the state-of-­ technologies, the key to successful reconstruction
the-art in multi-energy CT (Akagi et al. 2019). lies in the training of the neural network. Training
Previous implementations of kV switching of network parameters was conducted to mini-
have relied on brute force hardware approaches mize the mean absolution error loss function, as
to acquire enough views at each kV to both pre- shown in (1), between training input and target
serve image quality and effectively perform data,
material decomposition. Rather than compromise
ˆ = argmin ìí 1 ü
 yn* - f ( yn |Q )  1ý , (1)
wide volume coverage or AEC by greatly increas- Q
Q îN n
å þ
ing view rates, spectral reconstruction takes
advantage of the fact that much of the anatomical
information contained in a high kV view and a where Θ denotes the parameter set of the neu-
low kV view is common to both views. The high ral network, yn and yn* represent the nth training
spatial frequency information needed to maintain input and target sample, respectively, from a total
spatial resolution can be distilled from either of N training sample pairs, and f(y| Θ) represents
energy; for high spatial resolution content the neural network processing using input data y with
acquisition of a second independent sinogram is underlying parameter Θ.
superfluous. The unique information provided by Spectral deep learning reconstruction was
the high and low energy views is the degree to trained on complete measured sinograms acquired
Dual-Energy: The Canon Approach 67

at each energy for a wide variety of patient and sating for non-equal flux at 80 kV vs 135 kV and
phantom attenuation levels with a broad range of improving the utility of low keV Virtual
exposure levels. The sinogram data used for train- Monoenergetic Images (VMI). After the extensive
ing were processed with an array of sophisticated training process, Spectral deep learning recon-
models, such as a statistical model for noise struction was tested with independent validation
reduction. Other models utilized to ensure ultra- datasets and hundreds of thousands of image
high quality training include spectral, anatomical, results were reviewed extensively by engineers,
material, and advanced system models. Using medical physicists, and radiologists.
these ultra-high quality sinograms, the neural net-
work was trained to generate DLVs from mea-
sured opposite-energy views and adjacent 3 Spectral Performance
same-energy views. Spectral deep learning recon-
struction’s neural network has also been opti- 3.1  patial Resolution and CNR
S
mized for denoising in the image domain, using Performance
training data reconstructed with Canon’s AiCE
reconstruction algorithm (Fig. 4). AiCE has been By using high frequency information from the
demonstrated to preserve high contrast spatial views of both energies, spectral deep learning
resolution while greatly reducing noise magni- reconstruction is able to preserve the system’s
tude and producing a noise texture more similar to high contrast spatial resolution, as shown in
filtered backprojection than model-based iterative Fig. 5.
reconstruction approaches (Akagi et  al. 2019). The denoising capabilities of spectral deep
This denoising step plays a vital role in compen- learning reconstruction boost the contrast-to-­

Fig. 4  This figure illustrates the process of spectral deep sinogram material decomposition. The material decompo-
learning reconstruction algorithm (bottom) and the train- sition step produces basis material sinograms which can
ing of the neural networks (top). The kV switching data is be reconstructed into material images and monochromatic
input to the neural network. The neural network restores images
full sampled high kVp and low kVp views followed by
68 K. Boedeker et al.

noise (CNR) ratio of iodine. Below are CNR Prism, providing multi-energy imaging of entire
results for various densities in the 33 cm Gammex organs in a single rotation. This capability is par-
model 472, comparing Spectral CT across keV ticularly useful in cardiac CT where, in the last
levels to single energy CT (Fig. 6). 10  years, contrast-enhanced ECG gated CT has
become important in imaging the coronary arter-
ies. As an adjunct to coronary artery assessment,
3.2 Wide Volume Detector multi-energy CT is of increasing interest for eval-
uating myocardial blood supply (Vliegenthart
Canon Medical’s Spectral CT can be used in et al. 2012).
combination with the full 16  cm wide volume Below is an example of the image quality and
detector coverage available on the Aquilion ONE uniform temporal resolution that can be achieved
for cardiac imaging when spectral CT is com-
Body MTF bined with the use of a wide volume detector
1.2 (Fig.  7). Virtual monochromatic Images can be
Conventional 120kVp
1
Spectral 70keV
used to enhance contrast in the coronary arteries.
0.8
MTF

0.6
0.4 3.3 Virtual Monochromatic
0.2 Images
0
0 0.2 0.4 0.6 0.8 1 1.2 Conventional single energy acquisitions are spec-
lp/mm ified in terms of a peak kilovoltage, or kVp, in
which the indicated value refers to the maximum
Fig. 5  This graph compares the MTF of a monochro-
matic image at 70 keV from a Spectral Body scan with a energy contained in wide spectrum. The CT
120 kVp single energy Body scan Number, or Hounsfield Unit (HU), for a given

CNR
50.00
AIDR 120kVp
70 keV
40.00
60 keV
50 keV
30.00
CNR

40 keV

20.00

10.00

0.00
2mg/mL 2.5mg/mL 5mg/mL 7.5mg/mL 10mg/mL 15mg/mL 20mg/mL
Iodine Concentration

Fig. 6  This graph shows CNR results for various densi- show spectral CT consistently achieve higher CNR com-
ties in the 33 cm Gammex model 472, comparing spectral pared to single energy CT
CT across keV levels to single energy CT.  The results
Dual-Energy: The Canon Approach 69

Fig. 7  Virtual monoenergetic (70 keV) image of a spec- for the uniform temporal resolution. 70 keV DL Spectral.
tral cardiac scan. The left image shows the left ventricle in Rapid-kV switching (135–80 kVp); AEC; Spectral Body;
an approximate short axis view. The right image shows a CTDIvol: 9.6 mGy; DLP 153.4; 2.1 mSV *k-factor 0.014.
two-chamber view. These are examples of images that can (Courtesy Prof. Roy and Prof. Ohana, University Hospital
be obtained with a wide volume detector and are notable of Strasbourg)

pixel is largely determined by two energy-­ niques can diminish both high contrast spatial
dependent physical interactions between photons resolution and the noise texture. Spectral CT
and matter, viz. the photoelectric effect and takes advantage of the powerful noise reduction
Compton effect. The photoelectric effect is domi- capabilities of deep learning to improve the util-
nated at lower energies while the Compton effect ity of VMIs of all keVs for patient care. Phantom
is stronger at higher energies. In clinical practice, experiments show noise can be reduced by over
these two effects are discussed in terms of a rep- 50% at 70 keV relative to a conventional single
resentative material basis pair, such as iodine and energy scan while maintaining fine texture and
water. Acquiring images with two energy levels that the CNR of iodine is increased by 50%, even
allows for the individual impact each of basis at 40 keV (Fig. 8).
material on attenuation to be determined. This While contrast CT is helpful for diagnosing
material-specific information can then be com- GI pathology, contrast-induced nephropathy is a
bined to represent CT Numbers not just at the concern for patients with compromised renal
effective energy levels used for acquisition but function. Low energy virtual monochromatic
also at other energies in a spectrum, from images can enhance the contrast, and Canon’s
35–200 keV on the Aquilion ONE Prism. Because spectral solution is designed to offer low noise
images generated by recombining basis material and fine grain texture across the monochromatic
information represent data associated with single spectrum (Fig. 9).
point in the polyenergetic beam spectrum they
are called Virtual Monoenergetic Images (VMIs).
While the ability to create VMIs at keV levels 3.4 Spectral Iodine Maps
that yield improved contrast for various materials
has existed for some time, the promise of multi-­ In addition to structural information, spectral CT
energy CT VMIs for improving patient care has provides functional information. Images in
been stymied by the concomitant generation of Fig. 10 show iodine maps on which a perfusion
noise (Hanson et  al. 2018; Yu et  al. 2011). The defect is apparent as a large wedge-shaped region
CT Number of iodine increases over 15-fold in of hypoperfusion. An iodine map can both serve
reducing from 120  keV to 40  keV, but as the as a check on the original image, and also provide
share of photons comprising the recombined supporting information. An ROI placed on the
VMI image decreases the image noise increases iodine map can quantify the iodine concentration,
and can be prohibitive to visualization of low enabling a more detailed understanding of the
contrast objects. Conventional denoising tech- local perfusion (Fig. 10).
70 K. Boedeker et al.

Fig. 8  Virtual monochromatic images of a pulmonary or optimize the contrast difference between two adjacent
thromboembolism case. Canon’s Spectral CT allows the structures. Rapid kV switching (135–80  kVp); 550;
generation of virtual monochromatic images for the 165 Spectral Body; CTDIvol: 13  mGy; DLP 548; 7.7  mSV
energy levels between 35 and 200 keV. Spectral CT allows *k-factor 0.014. (Courtesy Prof. Yokoyama and Associate
easy tuning of the energy to suppress or enhance contrast, Prof. Machida, Kyorin University, Japan)

80 keV 70 keV 60 keV

Fig. 9  This post-nephrectomy patient with low renal vasculature with monochromatic imaging. (Courtesy
function was administered a reduced bolus of low density Prof. Roy and Prof. Ohana, University Hospital of
iodinated contrast. Spectral imaging was used to enhance Strasbourg)
Dual-Energy: The Canon Approach 71

Fig. 10  Iodine maps of a pulmonary thromboembolism 550  mA; Spectral Body; CTDIvol: 13  mGy; DLP 548;
case. These images show iodine maps on which a perfu- 7.7 mSV *k-factor 0.014. (Courtesy Prof. Yokoyama and
sion defect is apparent as a large wedge-shaped region of Associate Prof. Machida, Kyorin University, Japan)
hypoperfusion. Rapid-kV switching (135–80  kVp);

3.5 Workflow Postprocessing software applications offer the

ability to quantify images, as well as create multi-­
Rapid kV Switching with spectral deep learning layered images (Fig. 11).
reconstruction produces a wealth of spectral CT
data for clinicians to enhance diagnosis. Spectral Compliance with Ethical Standards
CT offers automatically generated monoener-
getic images, material-specific reconstructions, Disclosure of Interests subsidiaries.
and iodine maps, requiring no additional effort or
training for the technologist. Images are deliv- Ethical Approval  This article does not contain any stud-
ered directly to the reading station, making a rich ies with human participants performed by any of the
authors.
array of information readily available to assist the This article does not contain any studies with animals
radiologist with patient diagnosis. performed by any of the authors.
72 K. Boedeker et al.

Fig. 11  An example of a spectral abdominal image, pro- rial characterization; the blue curve shows increased
cessed on Vitrea™. Upper left: Virtual monochromatic enhancement at low energy as compared to the yellow
image. Upper center: Iodine blend image with one ROI curve. This information is helpful in understanding perfu-
selected. Upper right: Virtual non-contrast image. Bottom: sion differences in different parts of the liver. (Courtesy
spectral curve. An ROI is placed in the liver to measure Prof. Roy and Prof. Ohana, University Hospital of
the concentration of iodine in the iodine fusion image. Strasbourg)
The spectral curve offers an additional dimension of mate-

In International conference on medical image comput-

References ing and computer-assisted intervention 2015 Oct 5
(pp 234–241)
Akagi M, Nakamura Y, Higaki T, Narita K, Honda Vliegenthart R, Pelgrim GJ, Ebersberger U, Rowe
Y, Zhou J, Yu Z, Akino N, Awai K (2019) Deep GW, Oudkerk M, Schoepf UJ (2012) Dual-energy
learning reconstruction improves image quality of CT of the heart. AJR Am J Roentgenol 199(5
abdominal ultra-­ high-­resolution CT.  Eur Radiol Suppl):S54–S63
29(11):6163–6171 Yu L, Christner JA, Leng S, Wang J, Fletcher JG,
Hanson GJ, Michalak GJ, Childs R, McCollough McCollough CH (2011) Virtual monochromatic imag-
B, Kurup AN, Hough DM, Frye JM, Fidler JL, ing in dual-source dual-energy CT: radiation dose and
Venkatesh SK, Leng S, Yu L, Halaweish AF, image quality. Med Phys 38(12):6371–6379
Harmsen WS, McCollough CH, Fletcher JG (2018) Zhang R, Zhou J, Yu Z, Nemoto T (2020) A cascaded
Low kV versus dual-energy virtual monoener- deep-learning reconstruction method for sparse-view
getic CT imaging for proven liver lesions: what kV-switching dual-energy CT.  In Medical imaging
are the advantages and trade-offs in conspicuity 2020: physics of medical imaging 2020 Mar 16 (vol.
and image quality? A pilot study. Abdom Radiol 11312, p 1131223)
43(6):1404–1412 Zou Y, Silver MD (2008) Analysis of fast kV-switch-
Li B, Yadava G, Hsieh J (2011) Quantification of head and ing in dual energy CT using a pre-reconstruction
body CTDIvol of dual-energy x-ray CT with fast-kVp decomposition technique. In Medical imaging 2008:
switching. Med Phys 38(5):2595–2601 physics of medical imaging 2008 Mar 18 (vol 6913,
Ronneberger O, Fischer P, Brox T (2015) U-net: convo- p 691313)
lutional networks for biomedical image segmentation.
 Basic Principles and
Clinical Applications
of Photon-Counting CT

Thomas Flohr, Martin Petersilka, Stefan Ulzheimer,

Bernhard Schmidt, Klaus Erhard,
Bernhard Brendel, Marjorie Villien,
Philippe Coulon, Salim Si-Mohamed,
and Sara Boccalini

Contents
1  rinciples of Photon-Counting CT 
P  73
1.1  Properties of Current Solid-State Scintillation Detectors   73
1.2  Properties of Photon-Counting Detectors   75
1.3  Challenges for Photon-Counting Detectors   77
2 Material Decomposition for Photon-Counting CT   79
3 Pre-clinical Evaluation of Photon-Counting CT   81
References   90

1 Principles of Photon-
Counting CT

This review article gives an overview of the basic

principles of photon-counting detector CT, its
spectral capabilities, and of the clinical experi-
T. Flohr (*) · M. Petersilka · S. Ulzheimer ·
B. Schmidt ence gained so far in pre-clinical installations.
Siemens Healthcare GmbH, Computed Tomography, Other reviews of photon-counting detector CT
Forchheim, Germany may be found in Taguchi and Iwanczyk (2013),
e-mail: [email protected] Taguchi (2017), Willemink et  al. (2018), Leng
K. Erhard · B. Brendel et al. (2019), and Si-Mohamed et al. (2017a).
Philips GmbH Innovative Technologies, Research
Laboratories Hamburg, Hamburg, Germany
M. Villien · P. Coulon 1.1  roperties of Current Solid-
P
Philips Healthcare, Computed Tomography,
PA, Best, The Netherlands State Scintillation Detectors
S. Si-Mohamed · S. Boccalini
Hospices Civils de Lyon, Department of Radiology, To understand the properties of photon-counting
Lyon, France CT detectors it is helpful to recapitulate the prop-

© The Author(s), under exclusive license to Springer Nature Switzerland AG 2022 73

H. Alkadhi et al. (eds.), Spectral Imaging, Medical Radiology Diagnostic Imaging,
https://doi.org/10.1007/978-3-030-96285-2_6
74 T. Flohr et al.

a X-rays Collimator blades

TiO2 based
reflector

GOS

Photo-diodes

b 0.8 − 1mm z-direction

Detector element
top view

Separation Collimator
layers blades

Fig. 1  Schematic drawing of an energy-integrating scin- converted into an electrical current (two-step conversion).
tillator detector. (a) Side view, (b) top view. The The individual detector cells are separated by optically in-­
z-­direction is the patient’s longitudinal direction. Detector transparent layers (e.g., based on TiO2) to prevent optical
cells made of a scintillator such as GOS absorb the X-rays crosstalk. Collimator blades above the separation layers
(red arrows) and convert their energy into visible light suppress scattered radiation
(orange circles). This light is detected by photodiodes and

erties of solid-state scintillation detectors as they higher energy as consequence of the energy-­
are used in all medical CT scanners today. Solid-­ proportional weighting. In more mathematical
state scintillation detectors consist of detector terms, the detector responsivity D(E), a function
elements with a side length of 0.8–1 mm and a describing the generated photocurrent per inci-
depth of 1.2–2 mm, made of a scintillator (e.g., dent X-ray flux at energy E, is proportional to E,
gadolinium-oxysulfide GOS) with a photodiode see Fig. 2. The contrast-to-noise ratio (CNR) in
attached to its backside, see Fig.  1. The X-rays the CT images is negatively affected, in particu-
are absorbed in the scintillator. They produce vis- lar in CT scans with iodinated contrast agent—
ible light which is registered by the photodiode the X-ray absorption of iodine is highest at lower
and converted into an electrical current. Both the energies closely above its K-edge at 33 keV, see
intensity of the scintillation light and the ampli- Fig. 2.
tude of the induced current pulse are proportional Furthermore, scintillation detectors cannot
to the energy E of the absorbed X-ray photon. A provide spectrally resolved signals—all medical
CT scanner acquires 1000 and more projections CT scanners today capable of dual-energy imag-
(readings) during one rotation of the measure- ing rely on special acquisition techniques to miti-
ment system around the patient. For each of the gate this limitation, be it dual source CT, fast kV
detector elements, all current pulses registered switching, or use of dual layer detectors.
during the time of one projection are integrated. The low-level analog electric signal of the
X-rays with lower energy E, which carry most of photodiodes is distorted by electronic noise
the low-­contrast-­information, contribute less to which becomes larger than the quantum noise
the integrated detector signal than X-rays with (Poisson noise) of the X-ray photons at low X-ray
Basic Principles and Clinical Applications of Photon-Counting CT 75

Iodine K-edge Scintillator detector

Photon counting det

Detector responsivity D(E) [a.u.]

~E

~ const

Scintillator det: D(E) ~ E

Photon counting det: D(E) = const

20 60 100 140

X-ray energy [keV]

Fig. 2  Detector responsivity D(E) as a function of the by a vertical line. Low-energy X-rays closely above the
X-ray energy E for a GOS scintillator detector (dotted K-edge of iodine contribute less to the detector signal of a
black line, approximation solid black line) and a CdTe scintillator detector than of a photon-counting detector.
photon-counting detector (dotted blue line, approximation The iodine contrast in the image is therefore lower for a
solid blue line). The iodine K-edge at 33 keV is indicated scintillator detector than for a photon-counting detector

flux and causes a disproportional increase of the scintillators to increase spatial resolution
image noise and instability of low CT-numbers. while keeping the width of the separation layers
Electronic noise is, e.g., responsible for noise constant will reduce the geometric efficiency—
streaks in shoulder and pelvis images. The preva- therefore, it is problematic to increase the spatial
lence of electronic noise at low X-ray flux sets a resolution of solid-state scintillation detectors
limit to potential further radiation dose reduction beyond today’s performance levels (Flohr et  al.
in CT. 2007).
The detector elements are separated by opti-
cally in-transparent layers with a width of about
0.1 mm to prevent optical crosstalk between them 1.2 Properties of Photon-
which would significantly reduce spatial resolu- Counting Detectors
tion. X-ray photons absorbed in the separation
layers do not contribute to the measured signal Photon-counting detectors are made of semicon-
even though they have passed through the ductors such as cadmium-telluride (CdTe),
patient—from a radiation dose perspective they cadmium-­ zinc-telluride (CZT), or silicon (Si).
are wasted dose. The ratio of active detector cell We will focus on CdTe- and CZT-based photon-­
size and total size (including separation layers) is counting detectors.
called geometric dose efficiency. Current medical High voltage (800–1000 V) is applied between
CT detectors with an active cell size of about the cathode on top and pixelated anode electrodes
0.8  ×  0.8  mm2 to 1  ×  1  mm2 (Willemink et  al. at the bottom of a CdTe/CZT layer with a thick-
2018) have geometric dose efficiencies of 0.7– ness of 1.4–2  mm, see Fig.  3. The absorbed
0.8 (70–80%). Significantly reducing the size of X-rays produce electron-hole pairs which are
76 T. Flohr et al.

a X-rays Collimator blades

Cathode

CdTe/CZT High voltage ~1000V

Electrons
Pixelated anodes

z-direction
b
Detector element top view

Potential Collimator
sub-pixel blades
structure

Fig. 3  Schematic drawing of a direct converting photon-­ field without additional separation layers between them.
counting detector. (a) Side view, (b) top view. The X-rays Collimator blades are needed to suppress scattered radia-
(red arrows) absorbed in CdTe or CZT produce electron-­ tion. A potential division of the “macro pixels” between
hole pairs that are separated in a strong electric field two collimator blades into smaller sub-pixels is indicated
between cathode and pixelated anodes. The detector pix- for the three left detector cells. The pixelated anodes must
els are formed by the pixelated anodes and the electric then be correspondingly structured (not shown here)

separated in the strong electric field. The ­electrons threshold energy T0 ~ 20–25 keV. Low-amplitude
drift to the anodes and induce short current pulses baseline noise is well below this level and does
(10−9  s). A pulse-shaping circuit transforms the not trigger counts—even at low X-ray flux only
current pulses to voltage pulses with a full width the statistical Poisson noise of the X-ray quanta is
at half maximum (FWHM) of 10–15  nanosec- present in the signal. CT scans at very low radia-
onds. The pulse height of the voltage pulses is tion dose or CT scans of obese patients show
proportional to the energy E of the absorbed therefore less image noise, less streak artifacts,
X-rays. As soon as the pulses exceed a threshold and more stable CT-numbers than the corre-
they are counted, see Fig. 4. sponding scans with a scintillation detector, and
Photon-counting detectors have several advan- radiation dose reduction beyond today’s limits
tages compared to solid-state scintillation detec- seems possible.
tors. The detector elements are defined by the The detector responsivity D(E) in the X-ray
strong electric field between common cathode energy range from 30 to 100 keV is approximately
and pixelated anodes (Fig. 3) without additional constant (see Fig. 2)—all X-ray photons contribute
separation layers. The geometric dose efficiency equally to the measured signal regardless of their
is only reduced by anti-scatter collimator blades energy E, as soon as E exceeds T0. There is no
or grids. Different from scintillator detectors down-weighting of lower-energy X-ray photons as
each “macro pixel” confined by collimator blades in solid-state scintillation detectors. Photon-
can be divided into smaller sub-pixels which are counting detectors can provide CT images with
read-out separately if needed (see Fig.  3b) to potentially improved CNR, in particular in CT
improve spatial resolution. scans with iodinated contrast agent, see Fig. 2.
All current pulses produced by absorbed In a more advanced read-out mode, several
X-rays are counted as soon as they exceed a counters operating at different threshold energies
Basic Principles and Clinical Applications of Photon-Counting CT 77

100
T3

Pulse height (keV equivalent) 75 T2

50 T1

25 T0

Baseline
0 noise

0 100 200 300

Time (ns)

Fig. 4  The electrons produced by absorbed X-rays in a the low-amplitude baseline noise. Three additional thresh-
photon-counting detector induce signal pulses at the olds at higher energies (T1 at 50 keV, T2 at 75 keV, T3 at
anodes with a pulse height proportional to the X-ray 90 keV) are also indicated—simultaneous read-out of the
energy. The pulses are counted as soon as they exceed a counts exceeding different thresholds (in this example 4)
threshold T0 (dashed blue line, “counting” is indicated by provides spectrally resolved detector signals
a blue dot). T0 has a typical energy of 25 keV, well above

can be introduced for energy discrimination, see The simultaneous read-out of CT data in dif-
Fig.  4. Up to 6 different thresholds have so far ferent energy bins opens the potential of spec-
been realized in prototype settings (Schlomka trally resolved measurements and material
et al. 2008). In the example of Fig. 4, 4 different differentiation in any CT scan.
energy thresholds T0, T1, T2, and T3 are realized.
During the measurement time of one projection,
counter 1 counts all X-ray pulses with an energy 1.3 Challenges for Photon-
exceeding T0, while counter 2 simultaneously Counting Detectors
counts all X-ray pulses with an energy exceeding
T1, and so on. The photon-counting detector Despite their benefits, CdTe- or CZT-based
simultaneously provides 4 signals S0, S1, S2, and photon-­counting detectors need to cope with sev-
S3 with different lower-energy thresholds T0, T1, eral challenges. Their spectral separation is
T2, and T3. CT images reconstructed from these reduced by unavoidable physical effects. The
raw data are shown in Fig. 5. By subtracting the current pulses produced by X-rays absorbed
detector signals with adjacent lower-energy close to pixel borders are split between adjacent
thresholds, “energy bin” data can be produced. detectors cells (“charge sharing”). This leads to
Energy bin b0 = S1 – S0 as an example contains all erroneous counting of a high-energy X-ray pho-
X-ray photons detected in the energy range ton as several lower-energy hits. Incident X-rays
between T1 and T0. Physically, the thresholds are at an energy E may kick-out K-electrons of the
realized by different voltages which are fed into detector material (Cd and Te have K-edges at
pulse-height comparator circuits. 26.7 and 31.8  keV, respectively). The empty
78 T. Flohr et al.

Fig. 5  Contrast-enhanced kidney scan acquired with a contrast and the higher is the image noise in the recon-
pre-clinical hybrid dual source CT prototype with 4 structed images, because fewer low-energy X-ray photons
energy thresholds (25, 50, 75, and 90 keV, as indicated in contribute to the image. Courtesy of National Institute of
Fig. 4), operated at an X-ray tube voltage of 140 kV. The Health NIH, Bethesda, MD, USA
higher the low-energy threshold, the lower is the iodine

Fig. 6 Schematic X-ray X-ray

illustration of charge Energy E
sharing at pixel Cathode
boundaries and loss of
energy due to K-escape,
which lead to double + K-escape +
counting of X-ray pulses Direct − EFluoro + −
at wrong energies and converter −
reduction of spectral e. g. CdTe
separation. EFluoro is the
K-shell fluorescence − −
X-ray energy Pixelized − − −
E-EFluoro − EFluoro
anodes
Charge
sharing
Readout
electronics

K-shells are immediately refilled, and illustrated in Fig.  6. In summary, high-energy

­characteristic X-rays at the K-shell fluorescence X-ray photons are wrongly counted at lower
energy EFluoro are released which are re-absorbed energies, and spectral separation as well as spa-
and counted in the detector cell itself or in neigh- tial resolution are reduced. The low-energy bins
boring detector cells (“K-escape”). The incident of the detector will contain wrong high-energy
X-rays at the primary interaction site lose the information (“high-energy tails” as shown in
energy EFluoro and are counted at energy E- EFluoro. Fig. 7). For a realistic detector model including
Charge sharing, fluorescence, and K-escape are charge sharing, fluorescence, and K-escape, the
Basic Principles and Clinical Applications of Photon-Counting CT 79

0.045
(25,65)keV pulse bin (25,65)keV pulse bin
0.04 (65,140)keV pulse bin (65,140)keV pulse bin

0.035

0.03
225 µm

0.025 450 µm

0.02
“High-energy tail”
0.015

0.01

0.005

0
20 40 60 80 100 120 140 20 40 60 80 100 120 140
X-ray energy [keV] X-ray energy [keV]

Fig. 7  Computer simulation of the X-ray spectra recorded Right: pixel size 0.45 × 0.45 mm2. Increasing the pixel size
in the 2 energy bins of a realistic photon-counting detector reduces the characteristic “high-­energy tail” of the low-
for an incident 140  kV spectrum (bin1: 25–65  keV, blue energy bin caused by charge sharing and other effects such
line; bin2: 65–140 keV, green line). The dotted line is the as K-escape
incident X-ray spectrum. Left: pixel size 0.225 × 0.225 mm2.

spectral separation with two energy bins is equiv- 2 Material Decomposition

alent to existing dual-energy techniques (Kappler for Photon-Counting CT
et al. 2010).
Increasing the size of the detector pixels As the spectral X-ray attenuation of most materi-
improves spectral separation, because boundary als in the relevant energy regime from 30 to
effects such as charge sharing and K-escape con- 140  keV can be well approximated by a linear
tribute less to the total detector signal, see Fig. 7. combination of the attenuation due to the photo-­
The maximum size of the detector pixels is unfor- electric effect and Compton scattering, data
tunately limited by pulse pile-up. Medical CTs acquisition with two energy bins and a decompo-
are operated at high X-ray flux rates up to 109 sition into two base materials are often sufficient
counts per s and mm2—if the detector pixels are in the material decomposition task. Today’s
too large, too many X-ray photons hit them too established dual-energy applications are based on
closely in time to be registered separately (the decomposition into two base materials such as
width of the voltage pulses after pulse-shaping is calcium and water, iodine and water, or iodine
at least 10  ns). Several overlapping pulses are and calcium.
then counted as one hit only at a too high energy Data acquisition with more than two energy
(“pulse pile-up”). Pulse pile-up leads to non-­ bins enables multi-material decomposition under
linear detector count rates and finally to detector certain preconditions. Separation into two basis
saturation. It can be reduced by making the detec- materials plus an additional element with its
tor pixels smaller—however, smaller pixels lead K-edge position in the energy range of the source
to more charge sharing and K-escape. Finding spectrum of a clinical CT, such as gadolinium or
the optimum size of the detector cells to balance gold, requires the spectral data acquisition to
pulse pile-up, charge sharing, and K-escape is comprise at least three energy bins. The K-edge
one of the most challenging tasks in designing a attenuation needs to be represented by an addi-
photon-counting detector (Blevis 2020). tional base material for an accurate and consis-
80 T. Flohr et al.

tent material decomposition of the spectral data. reconstructed jointly with the use of a forward
Unfortunately, these K-edge materials do not model, which compares the current base material
naturally occur in the human body. Multi-material estimates with the measured spectral projection
decomposition will therefore be limited to clini- data and minimizes the residual error in an itera-
cal scenarios in which K-edge elements are intro- tive procedure. Both additional constraints on the
duced into the human body, e.g., to separate basis material images as well as statistical noise
different contrast agents (e.g., iodine and gado- models can therefore be easily incorporated at
linium, or iodine and bismuth), or to compute the cost of a computationally more expensive
material maps of other heavy elements (e.g., algorithm (Mory et al. 2018).
tungsten, or gold). One difference between dual-energy spectral
The computation of material-specific images CT systems and photon-counting spectral CT
from spectral projection data can essentially be systems regarding material decomposition is that
achieved with three different methods: image-­ in photon-counting CT systems the number of
domain decomposition, projection-domain spectral acquisitions (i.e., the number of energy
decomposition, and one-step inversion. In image-­ bins) is often higher than the number of basis
domain decomposition, each individual energy materials that should be decomposed. Thus, the
bin is first reconstructed separately and the actual decomposition task is over-determined, and the
material decomposition is performed in the image question especially for projection-domain
domain with linear approximations (Maaß et al. decomposition arises how to make optimal use of
2009). Prior knowledge on the object can be eas- the spectral information during decomposition. A
ily incorporated in these methods (Liu et  al. common option to use the spectral information
2016); however, beam-hardening artifacts are optimally in a projection-based decomposition is
difficult to be corrected in an image-domain the maximum likelihood material decomposition
decomposition approach and are better handled (Schlomka et al. 2008), an iterative minimization
with a projection-domain decomposition algorithm that aims to find the basis material line
algorithm. integrals, which best match the measured photon
In projection-domain decomposition (Roessl counts in the energy bins considering the noise
and Proksa 2007), the spectral information, given distribution of the measured data. A major ingre-
by the measurements of the photon counts in the dient of the maximum likelihood decomposition
various energy bins, is first decomposed into the is a spectral forward model that is used to calcu-
base materials directly in the projection domain, late for a given combination of material line inte-
for example with a maximum likelihood algo- grals the expectation values of the measured
rithm. Then, the material-specific sinograms are photon counts. The spectral forward model com-
separately reconstructed to provide the material-­ prises the photon emission of the tube, the photon
specific images in the image domain. This attenuation by the object, and the photon detec-
approach allows the incorporation of a statistical tion process by the photon-counting detector.
noise model and beam-hardening effects are Assuming that tube emission spectrum, material
properly modeled by the system model (Schirra attenuation spectra, and detector model are given,
et al. 2013). Compared to image-domain decom- the only unknown in the decomposition are the
position methods, prior (image domain) informa- basis material line integrals. If for a set of energy
tion and additional constraints are less easy to bins measured photon counts are available, the
implement in the projection domain. according basis material line integrals can be
Finally, one-step inversion methods aim at determined by minimizing the deviation between
combining the material decomposition and the measured counts and the expectation values
reconstruction task into one single step and solve of the counts, determined with the spectral for-
this problem iteratively (Mory et al. 2018). In this ward model. In maximum likelihood approaches,
way, the unknown base material images are the deviation is minimized in a statistical sense.
Basic Principles and Clinical Applications of Photon-Counting CT 81

For this, a model describing the noise in the mea- 3 Pre-clinical Evaluation
surements is needed. Commonly, it is assumed of Photon-Counting CT
that the noise of the measured photon counts fol-
lows a Poisson noise distribution. Currently, CT scanner prototypes based on CdTe/
Different strategies have been proposed to CZT detectors are used to evaluate the potential
solve the minimization problem (Schlomka et al. and limitations of photon-counting CT in clinical
2008). The final results of all strategies are virtu- practice.
ally the same. Thus, main selection criterion is A small-bore spectral micro-CT equipped
the computation time, since material decomposi- with a Medipix detector with 8 energy channels
tion can be time-consuming. Pre-computation has been translated to a large-bore photon-­
methods, look-up tables, and neural networks counting CT capable of obtaining diagnostic
may be applied for speed-up (Zimmerman and spectral CT images of a human within a clinical
Petschke 2017). radiation dose level (Panta et al. 2018); however,
The spectral image formation in a single-­ no further results have been published yet.
source spectral photon-counting CT prototype Recently, a clinical single-source spectral
(SPCCT - Philips Healthcare, Haifa, Israel) fol- photon-counting CT system (SPCCT—Philips
lows a two-step approach, with the material Healthcare, Haifa, Israel) with a full field-of-­
decomposition performed in the projection view (FOV) of 50  cm has been installed
domain. To this end, the photon counts from the (Si-Mohamed et  al. 2020). The system is
five energy bins are first decomposed into two or equipped with tiles of 2 mm thick CZT sensors
three basis materials using a maximum likelihood-­ with a 500 μm × 500 μm pixel pitch, bonded to
based algorithm as described above (Schlomka the proprietary ChromAIX2 application-specific
et al. 2008; Roessl and Proksa 2007). In a subse- integrated circuit (ASIC) (Steadman et al. 2017).
quent step, basis material images are recon- Each channel offers pulse-height discrimination
structed. For example, each basis material image with five programmable energy thresholds with
can be reconstructed separately using a filtered windows between 30 and 120  keV, which are
back-projection reconstruction algorithm (Grass advantageous with respect to achievable signal-­
2001; Heuscher 2004). to-­noise ratio (SNR) even if only two or three dif-
For standard spectral image reconstruction in ferent materials should be discriminated (Alvarez
other photon-counting CT prototypes (Siemens 2011). The single-source system acquires data
Healthcare GmbH, Forchheim), the photon from 64 rows to reconstruct a 50 cm FOV with a
counts from the four energy bins are combined z-coverage of 17.6 mm at the iso-center. The sys-
into two effective energy bins and statistically tem supports both axial and helical scan modes
decorrelated. Then, an iterative beam-hardening with a shortest gantry rotation time of 0.33 s and
correction in the projection domain is performed 2400 projections per rotation. Furthermore, the
based on a two-material decomposition into scanner is equipped with an ASG and can be
water and iodine—the result are pseudo-­ operated at clinically relevant flux levels for large
monoenergetic projection data at two distinct animal and human studies. A previous pre-­
energies. The pseudo-monoenergetic projection clinical single-source photon-counting CT sys-
data provide the basis for the established dual-­ tem (Philips Healthcare, Haifa, Israel) provided
energy applications, not only for direct recon- an in-plane field-of-view of 168  mm and a
struction of monoenergetic images at arbitrary z-­coverage of 2.5 mm, with a rotation time of 1 s
energies, but also for subsequent material decom- (Kopp et al. 2018).
position into water and iodine, water and iron, or A pre-clinical hybrid dual source CT scanner
calcium and iodine. For research, the projection prototype equipped with a conventional scintilla-
data of the four energy bins can be read-out sepa- tion detector and a CdTe photon-counting detec-
rately and decomposed into two, three, or four tor (Siemens Healthcare GmbH, Forchheim,
basis materials. Germany) was described and evaluated in
82 T. Flohr et al.

Kappler et  al. (2012, 2013, 2014). The photon-­ An improvement of iodine CNR by 11–38%
counting detector consists of sub-pixels with a with photon-counting CT was confirmed by mea-
size of 0.225 × 0.225 mm2. The detector provides surements in 4 anthropomorphic phantoms simu-
2 energy thresholds per sub-pixel. 2  ×  2 sub-­ lating 4 patient sizes (Gutjahr et  al. 2016).
pixels can be binned to a “UHR pixel” with a Improved iodine CNR can potentially be trans-
pixel size of 0.45 × 0.45 mm2, 4 × 4 sub-pixels lated into reduced radiation dose, or reduced
can be binned to a “macro pixel” with a size of amount of contrast agent. Better differentiation
0.9 × 0.9 mm2 comparable to today’s medical CT of gray and white brain matter was demonstrated
systems. By assigning alternating low-energy in a brain CT study with 21 human volunteers
and high-energy thresholds to adjacent detector (Pourmorteza et al. 2017) and attributed to both
sub-pixels in a “chess pattern mode,” the detector higher soft-tissue contrasts (10.3 ± 1.9 HU versus
provides 4 energy thresholds in “macro pixels.” 8.9 ± 1.8 HU), and lower image noise for photon-­
The in-plane FOV of the photon-counting detec- counting CT.
tor is 275  mm, the z-coverage is 8–16  mm, The impact of missing electronic noise on
depending on the read-out mode. A completion image quality was assessed for various clinical
scan with the energy-integrating sub-system can applications at low radiation dose. Less streaking
be used to extend the photon-counting FOV to artifacts in shoulder images acquired with the
500 mm. The shortest rotation time of the system photon-counting detector of the pre-clinical
is 0.5 s. hybrid dual source CT as compared to its scintil-
A clinical single-source CT scanner prototype lation detector were demonstrated (Yu et  al.
(Siemens Healthcare GmbH, Forchheim, 2016b), as well as better Hounsfield unit stability
Germany) is equipped with a CdTe photon-­ in a lung phantom in combination with better
counting detector consisting of sub-pixels with a reproducibility (Symons et al. 2017a). This is an
size of 0.275 × 0.322 mm2. Its FOV is 500 mm at important pre-requisite for further reduced radia-
the iso-center. 2 × 2 sub-pixels can be binned to a tion dose in lung imaging, e.g., in the context of
“macro pixel” in the “standard” mode, the lung cancer screening. In a study with 30 human
z-­coverage is then 57.6 mm (144 × 0.4 mm at the subjects undergoing dose-reduced chest CT
iso-center). The sub-pixels can as well be read imaging (Symons et al. 2017b), photon-counting
out separately in the “UHR” mode, the z-­coverage CT demonstrated higher diagnostic quality with
is then limited to 24  mm (120  ×  0.2  mm at the significantly better image quality scores, fewer
iso-center). The detector provides 4 energy beam-hardening artifacts, lower image noise, and
thresholds per sub-pixel. The shortest rotation higher CNR for lung nodule detection, see Fig. 8.
time of the system is 0.3 s (Flohr et al. 2020). The Improved quality of coronary artery calcium
system is installed in three pre-clinical settings (CAC) scoring with photon-counting CT at low
and operated using typical clinical scan protocols radiation dose was shown in a combined phan-
(Ferda et al. 2021). tom, ex  vivo and in  vivo study (Symons et  al.
The imaging performance of the pre-clinical 2019). The authors concluded that photon-­
hybrid dual source CT was evaluated by means of counting CT technology may play a role in fur-
phantom and cadaver scans (Yu et  al. 2016a; ther reducing the radiation dose of CAC scoring.
Gutjahr et  al. 2016), confirming clinical image Different photon-counting detector manufac-
quality at clinically realistic levels of X-ray pho- turers are using different compromises to balance
ton flux. In contrast-enhanced abdominal scans pulse pile-up and charge sharing effects, but all
of human volunteers, similar qualitative and use pixel sizes smaller than integrating detectors.
quantitative image quality scores as with conven- This results in increased spatial resolution, typi-
tional CT were achieved, with the addition of cally over 30  lp/cm or 0.15  mm (Kopp et  al.
spectral information for material decomposition 2018). It is important to notice that with small
(Pourmorteza et al. 2016). detector pixel size, the size of the X-ray tube
Basic Principles and Clinical Applications of Photon-Counting CT 83

a b

Fig. 8  Example of a low-dose lung scan acquired with heads) at low radiation dose because of the absence of
the pre-clinical hybrid dual source CT prototype. (a) electronic noise. (Courtesy of R Symons, NIH, Bethesda,
Energy-integrating detector image. (b) Photon-counting USA)
detector image, demonstrating less image noise (arrow-

focal spot practically determines the spatial evaluation of intra-stent restenosis with reliable
­resolution. Small focal spots can allow 30 lp/cm, results regarding the residual lumen for most
but at the expense of a reduced maximum X-ray tested stents and the clear identification or suspi-
flux or mA. Increased spatial resolution enabled cion of stenosis for all stents when, in contrast,
by the smaller detector pixels of photon-counting the residual lumen could not be detected for a
CT was evaluated in several phantom studies. single stent using dual layer CT.  The benefit of
0.15  mm in-plane spatial resolution and mini- spectral photon-counting CT for the assessment
mum slice widths down to 0.41 mm were demon- of stents was also demonstrated in vivo on rabbits
strated for the pre-clinical hybrid dual source CT as illustrated in Fig. 9. Figures 10 and 11 show
prototype, and better spatial resolution was con- first images of patient carotid and coronary CTA,
firmed in clinical images of the lung, shoulder, illustrating the added value of spectral photon-­
and temporal bone (Leng et  al. 2018). At equal counting CT for the assessment of atherosclerotic
spatial resolution, photon-counting images had plaques. The clinical benefits of this superior
less image noise than conventional CT images visualization of plaques in small vessels still need
because of the better modulation transfer func- to be validated but preliminary results are
tion (MTF) of the measurement system (Leng promising.
et  al. 2018). Significant improvements of coro- Kopp et al. (2018) demonstrated on a phantom
nary stent lumen were found (Symons et  al. and animal study that spectral photon-counting
2018a), as well as superior qualitative and quan- CT has the potential to improve the assessment of
titative image characteristics for coronary stent lung structures due to higher resolution com-
imaging when using a dedicated sharp convolu- pared to conventional CT.
tion kernel (von Spiczak et al. 2018). In a phan- Improved spatial resolution and less image
tom study (Bratke et  al. 2020) spectral noise with the “UHR mode” of the pre-clinical
photon-counting CT allowed for the noninvasive hybrid dual source CT were also demonstrated
84 T. Flohr et al.

a b

Fig. 9  Comparison of spectral photon-counting CT and lumen (about 5 mm diameter) compared to conventional
conventional CT for in vivo stent assessment in a rabbit. CT (a). Especially the stent’s struts and intra-stent lumen
Ultra-High Spatial Resolution images from SPCCT (b) can be better visualized. (Courtesy of S.  Si-Mohamed,
show a significantly improved visualization of the stent Hospices Civils de Lyon, France)

for CT scans of the brain, the thorax, and the kid- solutions of iodine (0.1–50  mg/ml) excellent
neys (Pourmorteza et  al. 2018), as well as for agreement between actual iodine concentrations
temporal bone anatomy (Zhou et al. 2018a). and iodine concentrations measured in the iodine
Superior visualization of higher-order bronchi maps was observed (Symons et al. 2018b). The
and third-/fourth-order bronchial walls at pre- authors assessed the use of iodine maps and
served lung nodule conspicuity compared with VMIs in head and neck CTA in 16 asymptomatic
clinical reference images was demonstrated in 22 volunteers and proposed VMIs as a method to
adult patients referred for clinically indicated enhance plaque detection and characterization as
high-resolution chest CT (Bartlett et  al. 2019). well as grading of stenosis by reconstructing
According to the authors, photon-counting CT is images at different keV.
beneficial for high-resolution imaging of airway The routine availability of VMIs with photon-­
diseases, and potentially for other pathologies, counting CT may pave the way to further stan-
such as fibrosis, honeycombing, and emphysema. dardization of CT protocols, provided that CNR
The achievable image quality with a photon-­ and image quality of the VMIs are enhanced by
counting detector in high-resolution chest CT is refined processing (see e.g. Grant et al. 2014). In
demonstrated in Fig. 12. Figure 13 illustrates the this approach, VMIs at standardized keV levels
resolution improvement for temporal bone tailored to the clinical question (e.g., 55–60 keV
anatomy. for contrast-enhanced examinations of parenchy-
The spectral performance of the pre-clinical mal organs, 45–55  keV for CT angiographic
hybrid dual source prototype with photon-­ studies) are the primary output of any CT scan
counting detector was evaluated in phantom stud- regardless of the acquisition protocol. Going one
ies (Leng et  al. 2017). CT number accuracy in step further, the acquisition protocol may be stan-
virtual monoenergetic images (VMIs) and iodine dardized as well. Some authors (Zhou et  al.
quantification accuracy were found to be compa- 2018b) already recommend a standardized acqui-
rable to dual source dual-energy CT. According sition protocol with 140  kV X-ray tube voltage
to the authors, photon-counting CT offers addi- for contrast-enhanced abdominal CT examina-
tional advantages, such as high spatial resolution, tions in all patient sizes, with standardized VMI
and improved CNR. In an anthropomorphic head reconstruction at 50  keV.  According to the
phantom containing tubes filled with aqueous authors, optimal or near optimal iodine CNR for
Basic Principles and Clinical Applications of Photon-Counting CT 85

a c

b d

Fig. 10  Axial (a, c) and modified sagittal (b, d) images of 64  ×  0.27  mm collimation, 0.33  s rotation time,
the right carotid artery of a 55-year-old asymptomatic CTDIvol = 15.9 mGy, DLP = 332.7 mGy.cm. Image recon-
man having undergone radiotherapy of the neck for struction: soft reconstruction kernel, 1024 × 1024 image
Hodgkin lymphoma, acquired with a single-source CT matrix, 0.25  mm slice width. Excellent visualization of
prototype with photon-counting detector (a, b) and angio the lumen of the carotid as well as plaques in both their
MRI (c, d). The injection protocol for the photon-­counting soft and calcific parts. The arrows show an ulceration con-
CT was of 40 mL of Iomeron 400 followed by 20 mL of firmed by angio MRI. (Courtesy of S. Boccalini, Hospices
saline at 4  mL/s. Data acquisition: helical mode with Civils de Lyon, France)

all patient sizes is obtained with this protocol. They found comparable overall performance to
Figure 14 shows spectral image types that can be state-of-the-art dual-energy CT in differentiating
routinely reconstructed for each thoraco-­ stone composition, while photon-counting CT
abdominal CT scan. was better able to help characterize small renal
Several authors assessed the performance of stones (Marcus et al. 2018).
spectral photon-counting CT for detection and Figure 15 demonstrates a promising dual-­
characterization of kidney stones, another estab- energy processing technique for vascular imag-
lished dual-energy CT application (Gutjahr et al. ing—the computation of virtual non-calcium
2017; Ferrero et  al. 2018; Marcus et  al. 2018). images based on a two-material decomposition
86 T. Flohr et al.

Fig. 11  Cardiac images of a 48-year-old woman with DLP = 475.7 mGy.cm. Image reconstruction: sharp recon-
chest pain, acquired with a single-source CT prototype struction kernel, 1024  ×  1024 image matrix, 0.25  mm
with photon-counting detector, after injection of 45 ml of slice width. Excellent visualization of the coronary tree
iodine (400 mg/ml) at a rate of 4 ml/s. Data acquisition: including small branches and small fat peri-coronary
ECG gated helical mode with 64 × 0.27 mm collimation, inflammation. (Courtesy of Prof. Douek, Hospices Civils
0.33  s rotation time, CTDIvol  =  25.7  mGy, de Lyon, France)

Fig. 12  Lung images of a 74-year-old woman with breast CTDIvol = 3.89 mGy, DLP = 126 mGycm. Image recon-
cancer and signs of fibrosis after radiation therapy, struction: sharp convolution kernel, 1024  ×  1024 image
acquired with a single-source CT prototype with photon-­ matrix, 0.4  mm slice width. Excellent visualization of
counting detector. Data acquisition: “UHR” mode, fibrosis and fine details such as fissures. (Courtesy of
120  ×  0.2  mm collimation, 0.3  s rotation time, J. Ferda, Pilsen, Czech Republic)
Basic Principles and Clinical Applications of Photon-Counting CT 87

a b c

Fig. 13 (a) Bones of the middle ear—the stapes (yellow photon-counting detector. Data acquisition: “UHR” mode,
circle) has a size of about 2 mm × 3 mm. (b) Specimen 120  ×  0.2  mm collimation. Spatial resolution is signifi-
image acquired with a state-of-the-art medical CT cantly improved. (Courtesy of A Persson, CMIV,
(SOMATOM Force, Siemens Healthcare). (c) Specimen Linköping, Sweden)
image acquired with a single-source CT prototype with

Fig. 14  Abdominal images of a 67-year-old woman with DLP = 450 mGycm. Image reconstruction: 0.4 mm slice
adrenal adenoma and parapelvic renal cyst, acquired with width. VMIs at 190, 65, 55, and 45  keV, virtual non-­
a single-source CT prototype with photon-counting detec- contrast image VNC, and iodine image. (Courtesy of
tor. Data acquisition: “standard” mode, 144 × 0.4 mm col- J. Ferda, Pilsen, Czech Republic)
limation, 0.3  s rotation time, CTDIvol  =  10.2  mGy,

into iodine and calcium images. In the iodine techniques removing calcified plaques, the lumen
images—which correspond to virtual non-­ of the vessels is restored.
calcium images—calcified plaques are removed If the photon-counting detector is operated with
which hamper the assessment of the true vessel more than two energy bins, multi-material decom-
lumen in standard CT images, in particular for position is possible if K-edge elements are present.
small vessels. In contrast to other processing In a canine model of myocardial infarction,
88 T. Flohr et al.

a b c

Fig. 15  Contrast-enhanced abdominal scan of a 73-year-­ the small aortic branch. (b) Conventional HU threshold-­
old patient acquired with a single-source CT prototype based Ca-removal—reliable evaluation of the small vessel
with photon-counting detector. Data acquisition: “stan- is still not possible (arrow). (c) Virtual non-Ca image—the
dard” mode, 144  ×  0.4  mm collimation, 0.5  s rotation calcified plaque is removed, and the true vessel lumen is
time, 120  kV, 130 eff. mAs, DLP  =  316 mGycm. (a) restored (arrow). (Courtesy of J.  Ferda, Pilsen, Czech
Standard diagnostic image (VMI at 65 keV). A calcified Republic)
plaque (arrow) prevents the assessment of the patency of

Symons et  al. (2017c) performed dual-­ contrast imaging by injecting iodinated and gadolinated
agent imaging of the heart to simultaneously contrast agents in healthy rabbits at different
assess both first-pass and late enhancement of the times so that the first contrast agent visualized
myocardium. The authors concluded that com- the portal phase and the second the arterial
bined first-pass iodine and late gadolinium maps phase.
allowed quantitative separation of blood pool, The ability to perform absolute quantification of
infarct scar, and remote myocardium. The same multiple contrast agents can be used for the simul-
authors also investigated the feasibility of simulta- taneous assessment of different pathophysiological
neous material decomposition of three contrast processes. Spectral photon-counting CT has been
agents (bismuth, iodine, and gadolinium) in vivo used to image macrophages inside atherosclerotic
in a canine model (Symons et  al. 2017d). They plaque with gold nanoparticles, and simultane-
observed tissue enhancement at multiple phases in ously image the arterial lumen with an iodine con-
a single CT acquisition, opening the potential to trast agent (Cormode et al. 2010). In a study from
replace multiphase CT scans by a single CT acqui- the same team, spectral photon-­counting CT has
sition with multiple contrast agents, see Fig. 16. been used in phantoms and in rabbits to simultane-
Several phantom and animal studies have also ously discriminate and quantify a gold blood pool
demonstrated the feasibility of multiple contrast agent (Au-NP) from an iodinated contrast agent
imaging using a combination of material decom- imaging tissue and calcium-­ rich matter, from a
position and K-edge imaging. The concept of single scan (Cormode et  al. 2017). Si-Mohamed
assessing the liver in different time phases from et al. (2018) have demonstrated that spectral pho-
a single-scan, double injection of iodine and ton-counting CT can be used to perform a complete
gadolinium contrast agents, was demonstrated in peritoneal dual-­contrast protocol, enabling a good
2017 (Muenzel et al. 2017) in a simulation study. assessment of the peritoneal cavity and abdominal
Si-Mohamed et al. (2019) demonstrated the fea- organs in rats thanks to dual-contrast agents within
sibility of this dual-contrast multiphase liver peritoneal and blood compartments, see Fig. 17.
Basic Principles and Clinical Applications of Photon-Counting CT 89

a b c

Fig. 16 Simultaneous imaging of 3 different contrast to simultaneously visualize different phases of renal
agents (iodine, gadolinium, and bismuth) by multi-­material enhancement. (a) Image acquired at 30 s after start of gad-
decomposition in a dog model. Scan data were acquired olinium injection, at the peak of gadolinium enhancement
with the pre-clinical hybrid dual source CT prototype and in the renal cortex. (b) Image acquired at 220 s, at the peak
read-out in four energy bins (25–50, 50–75, 75–90, and of iodine enhancement in the renal cortex. (c) Enhancement
90–140  keV). Bismuth was administered more than one curves of gadolinium and iodine in the aorta, renal cortex,
day prior to scanning. Intravenous administration of gado- medulla, and pelvis. (Courtesy of R Symons, NIH,
linium-based contrast agent was followed by intravenous Bethesda, MD, USA, see also Symons et al. 2017d)
administration of iodine-based contrast agent after 3 mins

Conventional Gadolinium Iodine Overlay

CT images K-edge images images images

1
2
3 3
4
5
10 mm
6

Injection of iodine agent IV and

gadolinium agent IP 10 mm

−100 HU 300 0 mg/mL 11 0 mg/mL 11

Fig. 17 Abdominopelvic spectral photon-counting CT purple) and the peritoneal cavity on the gadolinium
images and contrast material images acquired post-IV K-edge map (in green) for increasing the contrast in each
injections of intraperitoneal gadolinium and intravenous compartment potentially enabling better lesion detection
iodine-based contrast agents. The contrast material images in comparison to conventional imaging. (Courtesy of
allowed a specific enhancement of the abdominal organs S.  Si-Mohamed, Hospices Civils de Lyon, France, see
and vessels on the iodine map (top row, head arrow: also Si-Mohamed et al. 2018; Thivolet et al. 2020)
hepatic vein; bottom row, head arrow: urinary cavity) (in
90 T. Flohr et al.

While several novel imaging contrast agents integrating detectors and photon-counting detec-
based on high atomic number elements are being tor based spectral micro-CT.  The material
explored, iodine and gadolinium are particularly concentration maps confirmed expected biodis-
attractive because of their existing approval for tributions of contrast agents in the blood, liver,
clinical use. However, these iodinated agents have spleen, and kidneys. Photon-counting CT dem-
a number of limitations (nephropathy, gadolinium onstrated to be useful for functional characteriza-
retention, non-specificity, K-edge at too low tion of solid tumors and could aid in the
energy for iodine, etc.). Thus there is a compelling characterization of nanoparticles that show prom-
need to develop photon-counting specific contrast ise in the developing field of cancer theranostics.
agents to expand the field of CT-based molecular In clinical practice, the use of multi-material
imaging. Nanoparticles are a promising platform maps may be hampered by the unavoidable
for contrast agent development. Cormode et  al. increase of image noise in a multi-material
(2017) described the biodistribution and pharma- decomposition. Similar to ultra-high resolution
cokinetics of gold and iodine contrast agents. They scanning non-linear data and image denoising
found persistently high concentrations of the gold techniques will play a key role to fully exploit the
nanoparticles in the blood vessels of rabbits over potential of multi-material decomposition in clin-
the duration of the experiment (41 min), allowing ical routine, see, e.g., Tao et al. (2018).
both arterial and venous mapping. Photon-
counting CT imaging may allow the use of blood Compliance with Ethical Standards
pool agents for delayed steady-state imaging, and
can simultaneously perform first-pass arterial Ethical Approval This is a review article. No patient
imaging using a different contrast agent, such as scans were performed for this chapter.
iodine. Further potential applications of blood
pool contrast agents include detection of bleeding,
visualization of tumor vasculature, and quantifica- References
tion of tissue blood volume. In another study,
Riederer et  al. (2019) have investigated a tanta- Alvarez RE (2011) Estimator for photon counting energy
lum-based contrast agent (K-edge energy of tanta- selective x-ray imaging with multibin pulse height
analysis. Med Phys 38(5):2324–2334
lum is 67.4 keV). They were able to demonstrate Badea CT, Clark DP, Holbrook M, Srivastava M, Mowery
in a phantom study that spectral photon-counting Y, Ghaghada KB (2019) Functional imaging of tumor
CT can provide tantalum density maps and allows vasculature using iodine and gadolinium-based
for material decomposition and differentiation nanoparticle contrast agents: a comparison of spectral
micro-CT using energy integrating and photon count-
between tantalum and iodine in vitro which may ing detectors. Phys Med Biol 64(6):065007. https://
enable for an improved follow-up diagnosis in doi.org/10.1088/1361-­6560/ab03e2
patients after vascular occlusion therapy. Bartlett DJ, Koo WC, Bartholmai BJ et al (2019) High-­
Photon-counting CT can also be used to study resolution chest computed tomography imaging of
the lungs: impact of 1024 matrix reconstruction and
the biodistribution of nanoparticles in vivo over photon-­counting detector computed tomography.
time. In particular, longitudinal imaging of the Investig Radiol 54(3):129–137
biodistribution of nanoparticles would be highly Blevis I (2020) Chapter 10: X ray detectors for spectral
attractive for their clinical translation. Spectral photon counting CT. In: Taguchi K, Blevis I, Iniewski
K (eds) Spectral, photon counting computed tomogra-
photon-counting CT allowed repetitive and quick phy technology and applications, 1st edn. CRC Press,
acquisitions in vivo, and follow-up of changes in Boca Raton. https://doi.org/10.1201/9780429486111.
the Au-NP biodistribution over time as demon- First Published 2020; eBook Published 15 July 2020.
strated in 2017 (Si-Mohamed et  al. 2017b). In See Chapter 10
Bratke G, Hickethier T, Bar-Ness D et al (2020) Spectral
another study from Badea et al. (2019) the team photon-counting computed tomography for coro-
investigated the in vivo discrimination of iodine nary stent imaging: evaluation of the potential clini-
and gadolinium nanoparticle contrast agents cal impact for the delineation of in-stent restenosis.
using both dual-energy micro-CT with energy-­ Investig Radiol 55(2):61–67
Basic Principles and Clinical Applications of Photon-Counting CT 91

Cormode DP, Roessl E, Thran A et al (2010) Atherosclerotic In: Medical imaging 2014: physics of medical imag-
plaque composition: analysis with multicolor CT and ing 90331C
targeted gold nanoparticles. Radiology 256:774–782 Kopp FK, Daerr H, Si-Mohamed S et al (2018) Evaluation
Cormode DP, Si-Mohamed S, Bar-Ness D et  al (2017) of a preclinical photon-counting CT prototype for pul-
Multicolor spectral photon-counting computed tomog- monary imaging. Sci Rep 8(1):1738
raphy: in vivo dual contrast imaging with a high count Leng S, Zhou W, Yu Z et al (2017) Spectral performance
rate scanner. Sci Rep 7:4784. https://doi.org/10.1038/ of a whole-body research photon-counting detector
s41598-­017-­04659-­9 CT: quantitative accuracy in derived image sets. Phys
Ferda J, Vendiš T, Flohr T et  al (2021) Computed Med Biol 62(17):7216–7232
tomography with a full FOV photon-counting detec- Leng S, Rajendran K, Gong H et al (2018) 150-μm spatial
tor in a clinical setting, the first experience. Eur resolution using photon-counting detector computed
J Radiol 137:109614. https://doi.org/10.1016/j. tomography technology: technical performance and
ejrad.2021.109614. Online ahead of print first patient images. Investig Radiol 53(11):655–662
Ferrero A, Gutjahr R, Halaweish AF, Leng S, McCollough Leng S, Bruesewitz M, Tao S et al (2019) Photon-counting
CH (2018) Characterization of urinary stone composi- detector CT: system design and clinical applica-
tion by use of whole-body, photon-counting detector tions of an emerging technology. Radiographics
CT. Acad Radiol 25(10):1270–1276 39(3):729–743
Flohr TG, Stierstorfer K, Süß C et al (2007) Novel ultra- Liu J, Ding H, Molloi S, Zhang X, Gao H (2016) TICMR:
high resolution data acquisition and image recon- Total image constrained material reconstruction via
struction for multi- detector row CT.  Med Phys nonlocal total variation regularization for spectral
34(5):1712–1723 CT. IEEE Trans Med Imaging 35(12):2578–2586
Flohr T, Petersilka M, Henning A, Ulzheimer S, Ferda J, Maaß C, Baer M, Kachelrieß M (2009) Image-based dual
Schmidt B (2020) Photon counting CT review. Phys energy CT using optimized precorrection functions: a
Med 79:126–136 practical new approach of material decomposition in
Grant KL, Flohr TG, Krauss B et  al (2014) Assessment image domain. Med Phys 36(8):3818–3829
of an advanced image-based technique to calculate Marcus RP, Fletcher JG, Ferrero A et al (2018) Detection
virtual monoenergetic computed tomographic images and characterization of renal stones by using photon-­
from a dual-energy examination to improve contrast-­ counting-­based CT. Radiology 289(2):436–442
to-­noise ratio in examinations using iodinated contrast Mory C, Sixou B, Si-Mohamed S, Boussel L, Rit S (2018)
media. Investig Radiol 49(9):586–592 Comparison of five one-step reconstruction algorithms
Grass M (2001) Angular weighted hybrid cone-beam CT for spectral CT. Phys Med Biol 63(23):235001
reconstruction for circular trajectories. Phys Med Biol Muenzel D, Daerr H, Proksa R et al (2017) Simultaneous
46(6):1595–1610 dual-contrast multi-phase liver imaging using spectral
Gutjahr R, Halaweish AF, Yu Z et al (2016) Human imag- photoncounting computed tomography: a proof-of-­
ing with photon-counting-based computed tomogra- concept study. Eur Radiol Exp 1:25
phy at clinical dose levels: contrast-to-noise ratio and Panta RK, Butler APH, de Ruiter NJA et al (2018) First
cadaver studies. Investig Radiol 51(7):421–429 human imaging with MARS photon-counting CT. In:
Gutjahr R, Polster C, Henning A et al (2017) Dual-energy 2018 IEEE nuclear science symposium and medical
CT kidney stone differentiation in photon-counting imaging conference proceedings (NSS/MIC)
computed tomography. Proc SPIE Int Soc Opt Eng Pourmorteza A, Symons R, Sandfort V et  al (2016)
10132 Abdominal imaging with contrast-enhanced photon-­
Heuscher D (2004) Redundant data and exact heli- counting CT: first human experience. Radiology
cal cone-beam reconstruction. Phys Med Biol 279(1):239–245
49(11):2219–2238 Pourmorteza A, Symons R, Reich DS, Bagheri M, Cork
Kappler S, Niederlöhner D, Stierstorfer K, Flohr T (2010) TE, Kappler S, Ulzheimer S, Bluemke DA (2017)
Contrast-enhancement, image noise and dual-energy Photon-counting CT of the brain: in  vivo human
simulations for quantum-counting clinical CT.  In: results and image-quality assessment. AJNR Am J
Proceedings of the SPIE medical imaging conference Neuroradiol 38(12):2257–2263
2010, Vol 7622, pp 76223H Pourmorteza A, Symons R, Henning A, Ulzheimer S,
Kappler S, Hannemann T, Kraft E et al (2012) First results Bluemke DA (2018) Dose efficiency of quarter-­
from a hybrid prototype CT scanner for exploring ben- millimeter photon-counting computed tomography:
efits of quantum-counting in clinical CT. In: Medical first-in-human results. Investig Radiol 53(6):365–372
imaging 2012: physics of medical imaging 83130X Riederer I, Bar-Ness D, Kimm M et  al (2019) Liquid
Kappler S, Henning A, Krauss B, et  al (2013) Multi-­ embolic agents in spectral X-ray photon-­ counting
energy performance of a research prototype CT scan- computed tomography using tantalum K-edge
ner with small-pixel counting detector. In: Medical imaging. Sci Rep 9:5268. https://doi.org/10.1038/
imaging 2013: physics of medical imaging 86680O s41598-­019-­41737-­6
Kappler S, Henning A, Kreisler B, et  al (2014) Photon-­ Roessl E, Proksa R (2007) K-edge imaging in x-ray com-
counting CT at elevated x-ray tube currents: contrast puted tomography using multi-bin photon counting
stability, image noise and multi-energy performance. detectors. Phys Med Biol 52(15):4679–4696
92 T. Flohr et al.

Schirra CO, Roessl E, Koehler T et al (2013) Statistical Symons R, Reich DS, Bagheri M et  al (2018b) Photon-­
reconstruction of material decomposed data in spectral counting computed tomography for vascular imag-
CT. IEEE Trans Med Imaging 32(7):1249–1257 ing of the head and neck: first in vivo human results.
Schlomka JP, Roessl E, Dorscheid R et  al (2008) Investig Radiol 53(3):135–142
Experimental feasibility of multi-energy photon-­ Symons R, Sandfort V, Mallek M, Ulzheimer S,
counting K-edge imaging in pre-clinical computed Pourmorteza A (2019) Coronary artery calcium scor-
tomography. Phys Med Biol 53(15):4031–4047 ing with photon-counting CT: first in  vivo human
Si-Mohamed S, Bar-Ness D, Sigovan M et  al (2017a) experience. Int J Cardiovasc Imaging 35(4):733–739
Review of an initial experience with an experimental Taguchi K (2017) Energy-sensitive photon-counting
spectral photon counting computed tomography sys- detector-based X-ray computed tomography. Radiol
tem. NIMA 873:27–35 Phys Technol 10(1):8–22
Si-Mohamed S, Cormode DP, Bar-Ness D et  al (2017b) Taguchi K, Iwanczyk JS (2013) Vision 20/20: single
Evaluation of spectral photon counting computed photon-­counting x-ray detectors in medical imaging.
tomography K-edge imaging for determination of Med Phys 40(10):100901
gold nanoparticle biodistribution in  vivo. Nanoscale Tao S, Rajendran K, McCollough CH, Leng S (2018)
9(46):18246–18257 Material decomposition with prior knowledge aware
Si-Mohamed S, Thivolet A, Bonnot P et  al (2018) iterative denoising (MD-PKAID). Phys Med Biol
Improved peritoneal cavity and abdominal organ 63(19):195003
imaging using a biphasic contrast agent protocol Thivolet A, Si-Mohamed S, Bonnot P et al (2020) Spectral
and spectral photon counting computed tomography photon-counting CT imaging of colorectal peritoneal
K-edge imaging. Investig Radiol 53(10):629–639 metastases: initial experience in rats. Sci Rep 10:13394.
Si-Mohamed S, Tatard-Leitman V, Laugerette A https://doi.org/10.1038/s41598-­020-­70282-­w
et  al (2019) Spectral Photon-Counting Computed von Spiczak J, Mannil M, Peters B et al (2018) Photon-
Tomography (SPCCT): in-vivo single-acquisition counting computed tomography with dedicated
multi-phase liver imaging with a dual contrast agent sharp convolution kernels  – tapping the potential of
protocol. Sci Rep 9:8458. https://doi.org/10.1038/ a new technology for stent imaging. Investig Radiol
s41598-­019-­44821-­z 53(8):486–494
Si-Mohamed S, Boussel L, Douek P (2020) Clinical per- Willemink MJ, Persson M, Pourmorteza A, Pelc NJ,
spectives of spectral photon-counting CT. In: Spectral, Fleischmann D (2018) Photon-counting CT: tech-
photon counting computed tomography: technology nical principles and clinical prospects. Radiology
and applications. CRC Press, Taylor & Francis Group, 289(2):293–312
Boca Raton, p 97 Yu Z, Leng S, Jorgensen SM et al (2016a) Evaluation of
Steadman R, Herrmann C, Livne A (2017) Chro-mAIX2: conventional imaging performance in a research CT
a large area, high count-rate energy-resolving pho- system with a photon-counting detector array. Phys
ton counting ASIC for a spectral CT prototype. Nucl Med Biol 61:1572–1595
Instrum Methods Phys Res, Sect A 862:18–24 Yu Z, Leng S, Kappler S et al (2016b) Noise performance
Symons R, Cork T, Sahbaee P et  al (2017a) Low-dose of low-dose CT_ comparison between an energy inte-
lung cancer screening with photon-counting CT: a grating detector and a photon-counting detector using
feasibility study. Phys Med Biol 62(1):202–213 a whole-body research photon-counting CT scanner. J
Symons R, Pourmorteza A, Sandfort V et  al (2017b) Med Imaging 3(4):043503
Feasibility of dose-reduced chest CT with photon-­ Zhou W, Lane JI, Carlson ML et al (2018a) Comparison
counting detectors: initial results in humans. of a photon-counting-detector CT with an energy-­
Radiology 285(3):980–989 integrating-­detector CT for temporal bone imag-
Symons R, Cork TE, Lakshmanan MN et al (2017c) Dual-­ ing: a cadaveric study. AJNR Am J Neuroradiol
contrast agent photon-counting computed tomogra- 39(9):1733–1738
phy of the heart: initial experience. Int J Cardiovasc Zhou W, Abdurakhimova D, Bruesewitz M et al (2018b)
Imaging 33:1253–1261 Impact of photon-counting detector technology on kV
Symons R, Krauss B, Sahbaee P et  al (2017d) Photon-­ selection and diagnostic workflow in CT.  Proc SPIE
counting CT for simultaneous imaging of multiple Int Soc. Opt Eng:10573. pii: 105731C. https://doi.
contrast agents in the abdomen: an in vivo study. Med org/10.1117/12.2294952
Phys 44(10):5120–5127 Zimmerman KC, Petschke A (2017) Empirical neural net-
Symons R, de Bruecker Y, Roosen J et al (2018a) Quarter-­ work forward model for maximum likelihood material
millimeter spectral coronary stent imaging with decomposition in spectral CT.  In: Medical imaging
photon-counting CT: initial experience. J Cardiovasc 2017: physics of medical imaging 101323S
Comput Tomogr 12:509–515
 Contrast Media for Modern
Computed Tomography

Hubertus Pietsch and Gregor Jost

Contents
1 Introduction   93
2 Iodinated Contrast Media: Current Standard in Safety and Tolerability   94
2.1   tructure and Physicochemistry 
S  94
2.2  Tolerability and Safety   95
2.3  LOCM, the Reference Standard   95
3  he Technology Is a Determining Factor for the Efficiency of Contrast
T
Media   95
4  harmacokinetics of Iodinated Contrast Media and Their Relevance for
P
Modern CT   97
4.1  Volume of Distribution and Excretion   97
4.2  Basic Mechanisms of Contrasting Inside the Body   99
5 Contrast Media in Dual-Energy CT   99
6 Clinical Applications of Dual-Energy Material Decomposition   101
6.1  Oncology Applications   101
6.2  Cardiovascular Applications   102
6.3  Pulmonary Applications   102
7 Summary and Outlook   103
References   104

1 Introduction
H. Pietsch (*)
Bayer AG, MR and CT Contrast Media Research, Contrast media are an important element in mod-
Berlin, Germany ern computed tomography and are the basis for a
Universitätsklinikum Essen, Institute of Diagnostic profound diagnosis. The number of computed
and Interventional Radiology and Neuroradiology, tomography (CT) examinations performed is
Essen, Germany increasing by approximately 5% each year. In
e-mail: [email protected] 2020, roughly 280 million CT scans were per-
G. Jost formed worldwide, and about 50% of these CT
Bayer AG, MR and CT Contrast Media Research, examinations were contrast enhanced.
Berlin, Germany

© The Author(s), under exclusive license to Springer Nature Switzerland AG 2022 93

H. Alkadhi et al. (eds.), Spectral Imaging, Medical Radiology Diagnostic Imaging,
https://doi.org/10.1007/978-3-030-96285-2_7
94 H. Pietsch and G. Jost

Intravenous iodine-containing low- and iso-­ sis, acute cerebral or myocardial infarction,
osmolar contrast media, such as iopromide or advanced cancer, severe heart failure, impaired
iodixanol, are standard and are commonly used renal or hepatic function, and also pregnant
for contrast enhancement in CT imaging (Pietsch women as well as newborns have been given
et al. 2012). these contrast media via all conceivable routes of
Thus, the physical properties of these contrast administration.
media for X-ray attenuation are limited exclu-
sively to one element—iodine. One option for the
future would be the development of contrast 2.1 Structure
media with other contrasting elements, prefera- and Physicochemistry
bly with a much higher atomic number (Nowak
et al. 2011; Pietsch et al. 2009). All currently available iodinated contrast media
Such contrast media would be more suitable are based on the tri-iodinated benzene ring. They
for higher voltage ranges of the X-ray tube used are classified, based on the presence of 1 or 2 tri-­
in CT. On the one hand, this would allow more iodinated benzene rings, as monomeric or
flexibility in imaging protocols, allowing exami- dimeric, and based on the electric charge of the
nations with lower radiation doses to be made molecule, as ionic and nonionic contrast media,
with equivalent diagnostic value (Nowak et  al. respectively (Pietsch et al. 2012; Schöckel et al.
2011; Roessler et  al. 2016). In addition, these 2020).
contrast media could allow new applications or The structure and charge of the molecule
indications due to optimized spectral differentia- define the two main physicochemical properties
tion in X-ray attenuation compared to iodine. of contrast media:
Furthermore, iodine-free contrast media would
have advantages for patients with known reac- 1. Osmotic pressure or osmolality
tions to contrast media or with thyroid High osmolality ionic contrast media
disorders. (HOCM) such as sodium diatrizoate
The continuous technological progress in (Urografin) has an osmolality up to five times
X-ray imaging and CT should be an incentive for higher than that of human blood. In contrast,
the development of a new dedicated contrast nonionic low osmolality CM (LOCM) have a
media, as new technologies in detectors and markedly lower osmolality than HOCM,
tubes, as well as the application of artificial intel- which is nonetheless higher than that of blood.
ligence algorithms will further develop and Nonionic iso-osmolar contrast media (IOCM)
improve the next generation of CT scanners in have an osmolality the equivalent to that of
the long term. blood.
2. Viscosity
The viscosity is a crucial parameter for the
2 I odinated Contrast Media: flow properties of solutions, which increases
Current Standard in Safety with increasing concentration and decreasing
and Tolerability temperature in a non-linear relationship. In
practice, this factor is relevant for the possible
Iodinated nonionic contrast media are the stan- speed of injection of contrast media through a
dard in contrast-enhanced CT.  They were clini- needle or catheter.
cally approved and established over 40 years ago. At comparable concentrations, dimeric
Millions of patients with a wide variety of dis- contrast media are much more viscous than
eases, multimorbid elderly patients, patients with monomerics at similar concentrations. All
severe diseases such as generalized atherosclero- new contrast media developments must take
Contrast Media for Modern Computed Tomography 95

into account, in addition to excellent tolerabil- By far the most common reactions occur
ity, the physicochemical parameters which are within 1 h and are usually mild and self-limiting
crucial for the application of such agents in (Dawson 2006).
CT (Behrendt et al. 2013).

2.3 LOCM, the Reference

2.2 Tolerability and Safety Standard

In general, iodinated CM have a very high toler- The properties of LOCM, in particular, its safety
ability and safety profile (Rosovsky et al. 1996; profile, set a high standard. All new contrast
Palkowitsch et  al. 2013; Dawson 2006). The media to be developed for CT must have compa-
known types of adverse reactions after applica- rable characteristics. The molecules must be
tion can in principle be divided according to their characterized by a high hydrophilicity and solu-
pathophysiology into idiosyncratic dose-­bility in order to achieve a low viscosity and low
independent hypersensitivity reactions or organ- osmolality in the final formulation, combined
otoxic dose-dependent reactions. Hypersensitivity with high stability (important for manufacturing,
reactions are allergy-like and are not immuno- basic requirement for autoclavability).
modulated. Direct biological reactions to the The important basics for good tolerability
administration of the molecules of iodinated CM in vivo are lowest possible protein binding, and
solution and their physicochemical properties no metabolism in the body as a basis for rapid
are, for example, contrast media-induced and complete renal elimination (Jost et al. 2009).
nephropathy (CIN) following intra-arterial A new contrast agent must compete with all
administration (e.g., cerebral angiography). these important properties.
Furthermore, reactions involving thyroid func-
tion due to free iodide can also be regarded as
such. Mild and transient adverse reactions are 3 The Technology Is
observed in up to 3% of patients after intravenous a Determining Factor
injection of nonionic compounds. Nevertheless, for the Efficiency of Contrast
some rare but occasionally serious adverse reac- Media
tions to iodine-containing contrast media may
occur. In patients with a serious thyroid dysfunc- The diagnostic capabilities and performance of
tion, these include the effects of free iodide, today’s contrast media is still largely dependent
which may be present in contrast media in the on or influenced by the radiological technology
smallest traces or may be generated in the body used and has changed drastically and effectively
itself following administration. Adverse reac- over the evolutionary stages, particularly of CT
tions are usually observed more frequently with machine development. Compared with conven-
ionic than with nonionic contrast media (Schöckel tional X-ray, computed tomography has signifi-
et al. 2020; Kopp et al. 2008). cantly improved the effectiveness of iodinated
Adverse effects of contrast media are also cat- CM (Schockel et al. 2020).
egorized according to the time of their In CT, in addition to the concentration of the
occurrence: contrasting element on site, the high sensitivity of
the CT techniques is also the basis for the detec-
• acute (up to 1 h following injection), tion and spatially resolved visualization of iodin-
• delayed/late (between 1 h and 7 days follow- ated CM. In contrast, in standard projection X-ray
ing injection), procedures, such as angiography, only the amount
• very late reactions (more than 1 week follow- of iodine present along the path of the X-ray radi-
ing injection of an iodine-containing CM). ation penetrating the object is relevant (Fig. 1).
96 H. Pietsch and G. Jost

Lesion 1

Projection imaging
14 · 10x Iodine atoms on the way

CT imaging
5 mg I/ml + native tissue = 150 HU

X- Ra
y

Lesion 2

Projection imaging
7 · 10x Iodine atoms on the way

CT imaging
2 mg I/ml + native tissue = 60 HU
X- R
ay

Fig. 1  Comparison of the contrast in 2D radiography versus CT

The modern CT is much more sensitive to Today, modern CT uses tube voltages between
iodine than conventional X-ray imaging. 70 and 150 kV (Lusic and Grinstaff 2013). The use
Projection radiography requires at least 20  mg of tube voltages in the range of 70–100 kV yields
iodine/ml to visualize tissue with a thickness of greater iodine attenuations up to a factor of two
1 cm (Langer et al. 1985). This is demonstrated compared to 120  kV (Fig.  2). The resulting
by an example of angiography: despite very increase in sensitivity to contrast media can be
high dosages of contrast media, which are rap- used to reduced radiation dose and/or contrast
idly and locally injected, intravenous DSA has a media dose specifically to the patient or to increase
relatively low sensitivity to iodine. In contrast, image quality in certain indications (Fleischmann
CT is able to clearly visualize iodine concentra- et al. 2018). The introduction of high-power X-ray
tions of 1 mg/ml in a volume of less than 0.1 ml. tubes allows CT imaging at lower kV for multiple
The CT attenuation given in Hounsfield units indications and for a broad patient population.
(HU) is based on the linear attenuation coeffi- The development and widespread use of inter-
cient μ, which describes how monochromatic active reconstruction techniques further increased
X-rays are attenuated when they pass through the sensitivity for visualization of iodine enhance-
an object along the path they cross. However, it ment by reducing the image noise level.
is not identical to this as μ strongly depend on A further step is the wide availability of differ-
the X-ray energy and in turn to the X-ray tube ent dual-energy technologies and the introduc-
spectrum. Therefore, on the HU scale the CT tion of counting detectors in CT, with the potential
values are normalized to the attenuation of water for spectral imaging. This offers great opportuni-
to correct for different X-ray tube voltages and ties and will change the use and spectrum of con-
filtrations. trast media (Flohr et al. 2006).
Contrast Media for Modern Computed Tomography 97

Fig. 2 Increasing
attenuation of iodine at
lower tube voltages.
Samples of Ultravist 200
containing solutions

Relative Attenuation (%)

were investigated in a
human equivalent body 150
phantom. Measurement
was performed in a
100
dual-source CT Siemens
Somatom Force
50

0
70 80 90 100 110 120 130 140 150
Tube Voltage (kV)

4 Pharmacokinetics tion of the body. Finally, the contrast media mol-

of Iodinated Contrast Media ecules are filtered by the glomeruli, concentrated
and Their Relevance in the renal tubules, and excreted in the urine
for Modern CT (Pietsch et al. 2012).
Because iodinated contrast media can neither
With the introduction of CT, the clinical signifi- penetrate cells of the body nor pass through bio-
cance and the huge diagnostic benefit of contrast logical barriers such as the blood–brain barrier,
media became evident, as the majority of neces- they are also termed extracellular CM (Claussen
sary diagnoses in CT can only be achieved with et al. 1984). This is important for the physician
the use of contrast media. The first clinical appli- and the patient because these agents are neither
cation of iodine-containing contrast media after absorbed enterically nor do they accumulate in
intravenous injection was visualization of the uri- healthy organs or tissues other than the kidney to
nary tract. In addition to urography, contrast an extent that can be used for diagnostic pur-
media were also used from the beginning to visu- poses. Thus, the iodine-containing contrast media
alize open and enclosed body cavities and blood are non-specific in the broadest sense, as they can
vessels because they could easily mix with the be circulated throughout the body, with the
contents of the organ or body cavities. exception of the central nervous system due to
the blood–brain barrier. On the other hand, pre-
cisely this feature is the basis for visualizing
4.1 Volume of Distribution pathologies within the body, such as inflamma-
and Excretion tion, necrosis, or tumors. All changes or pro-
cesses in the body that are associated with an
Contrast media cannot pass through cell mem- increase in the number of blood vessels (angio-
branes due to their hydrophilicity. In general, genesis), an increase in the permeability of the
these molecules are very small and do not inter- vessel walls, size changes of the interstitial space,
act with any parts, membranes or cells of the or simply an increased blood perfusion can be
body, nor should they affect its functionality. Due visualized with extracellular contrast media (Bae
to their size, they are able to diffuse through tiny et al. 2000, 2004).
pores. The general distribution takes place with Figure 3 illustrates the pathway of the con-
the blood flow analogously to the blood circula- trast media into the body following intravenous
98 H. Pietsch and G. Jost

Fig. 3  Dilution and

early pharmacokinetics
of extracellular contrast
agents after intravenous
injection depend on the Broadening of
perfusion, vascular bolus in the lung due to:
permeability and size of - Length of vessels/ cappilaries
the interstitial space of - Blood volume
the respective tissue
* Tissue: The interstitial result in a delay + initial dilution
space includes the until equilibrium
plasma volume (e.g. 5% is searched
of tissue volume,
contributing to about 30
HU) and the interstitial
space between the cells right heart left heart
of the solid tissue (e.g. dilution by 80 ml/s
10% of tissue volume). ~20 mg l/ml

perfusion: fast

tissue
permeability: high
interstital space: large
large arteries < 20 mg l/ml ~450 HU

perfusion: fast

tissue
permeability: low
interstital space: large

perfusion: fast
permeability: high tissue
interstital space: small

perfusion: slow
tissue

permeability: high
interstital space: small

perfusion: slow
tissue

permeability: high
interstital space: large

administration and also visualizes the basic 80 ml/s) and passage through the lungs reduces
mechanisms of the contrasting pathological or dilutes the concentration in the arterial sys-
­tissues. Based on a standard human dosage of tem to about 10 to <20 mg iodine/ml. However,
approximately 1 ml/kg body weight, extracellu- this concentration is still sufficient to result in a
lar CM is injected intravenously at a rate of contrast enhancement of several 100 Hounsfield
3–8  ml/s at concentrations of 300–370  mg units in the arterial blood on CT (Claussen et al.
iodine/ml. Dilution by the cardiac circulation (− 1984).
Contrast Media for Modern Computed Tomography 99

4.2 Basic Mechanisms iodine with elements with a higher atomic num-
of Contrasting Inside the Body ber, ideally tuned to the emission spectrum of a
modern CT tube. This approach is highly inter-
In addition to the intrinsic perfusion of the tissue, esting and offers attractive opportunities in the
the ability of the contrast medium to diffuse field of dual-energy imaging towards photon-­
through the tiny pores in the capillaries into the counting detector technology (Flohr et al. 2006;
interstitial space is an important prerequisite for Frenzel et al. 2015; Gutjahr et al. 2016).
contrasting pathological processes in the body.
This provides information about the nature of the
tissue, as well as the pathophysiological changes 5  ontrast Media in 
C
characteristic of many diseases. For example, Dual-Energy CT
blood flow differs between different tissues in the
body and is often increased in inflammation or Important clinical applications of dual-energy
fast-growing tumors. In contrast, it is heavily (DE) CT are contrast-enhanced procedures, for
reduced in ischemia, in certain tumors and example, the quantification of the iodine uptake
severely restricted or non-existent in necroses or in oncology, the evaluation of lung perfusion in
cysts. case of suspected pulmonary embolism or the
The permeability of the vessels can also be characterization of myocardial perfusion defects.
described by the contrast medium—which could The physical basis of these applications is the
also differ in the respective tissues of the body. spectral X-ray attenuation characteristic of
For example, inflammation, angiogenesis, spe- iodine, the attenuating element in clinically avail-
cial tumors as well as disruptions of the blood– able X-ray contrast media. In the energy range of
brain barrier are characterized by an increase in CT, the iodine mass attenuation coefficient shows
the permeability of the vessels, which in turn a large energy dependence with a relatively low
leads to altered contrast media concentrations in K-edge energy at 33.2  keV (Fig.  4, upper left).
the interstitial space. Consequently, the CT attenuation strongly
However, this specific distribution is transient, depends on the X-ray tube spectrum, mainly on
sometimes only for seconds after intravascular tube potential and filtering but, to a lesser extent,
injection or during the first passage of the con- also on the object composition and size (Fig. 2).
trast medium through the tissues. It can already Measurements using Ultravist in a thoracic phan-
disappear in the recirculation phase. The great tom revealed an attenuation of 50.6 HU per mgI/
advantage of CT is the ability to scan the respec- mL at 70  kV that continuously decreased with
tive organ or the entire body quickly and repeat- tube voltage to 21.6 HU per mgI/mL at
edly in order to use the specific dynamics and the 150 kV. This more than 2.3-fold difference can be
early distribution pattern of the contrast medium further increased using dedicated filters; the addi-
efficiently for diagnosis. tional 0.6 mm tin filtration available for the third-­
Two fundamental attributes of contrast media generation dual-source CT results in a further
are the basic prerequisites for their use in CT: the decrease of attenuation to 11.8 HU per mg I/ml at
ability to efficiently absorb X-rays due to the ele- 150 kV. The improved spectral separation allows
ment iodine and, due to the physicochemical a better material decomposition of iodine vs. soft
properties of the molecules, to depict numerous tissue that showed no significant change in atten-
organs or systems and pathologically altered uation in dependent on the kV setting and filter-
structures within the human body via the phar- ing. Notably, many innovations in dual-energy
macokinetics. The next step would be to change CT technology are tailored to the spectral attenu-
fundamental properties such as the distribution ation profile of iodine.
spaces by adjusting the size of the molecules. It is An alternate approach for dual-energy CT is
possible to change the spectral attenuation prop- the change of the contrast media attenuating
erties. This can be achieved by replacing the element. Elements with higher K-edge energy
100 H. Pietsch and G. Jost

Mass Attenuation Coefficient (cm2/g)

Mass Attenuation Coefficient (cm2/g)

60 1.0 60 1.0

X-Ray Tube Spectrum

X-Ray Tube Spectrum

40 40

0.5 0.5

20 20

0 0.0 0 0.0
20 40 60 80 100 120 20 40 60 80 100 120
Energy (keV) Energy (keV)
Iodine Xe Dy Hf W

Attenuation (HU per mgX/mL; [10% Xe])

Attenuation (HU per mg/mL)

60 60

40 40

20 20

0 0
80 100 120 140 20 100 120 140
Tube Voltage (kV) Tube Voltage (kV)
Ultravist Xe Dy Hf W

Fig. 4  Spectral X-ray attenuation of Ultravist (Iodine) coefficient superimposed with a typical 120  kV X-Ray
and alternative promising elements with different k-edge spectrum (upper row). Respective CT attenuations at dif-
energies as Dysprosium (Dy), Hafnium (Hf), Tungsten ferent tube voltages (lower row)
(W) and Xenon (Xe). Energy dependent mass attenuation

such as dysprosium (53.8  keV), hafnium Dual-energy adds additional information to the
(65.4  keV), or tungsten (69.5  keV) possess a visualization of morphology in CT images. In the
different spectral mass attenuation coefficient context of contrast media, this is realized by com-
that better match the photon energy distribution putation of a material map usually containing the
of the X-ray tube (Fig.  4; upper right). This amount and distribution of contrast media and vir-
results in more similar CT attenuations for the tual non-contrast (VNC) images containing ana-
different tube voltages, particularly for tung- tomical information. Different material
sten and hafnium. Tungsten, for example, has decomposition algorithms exist but all require dif-
almost identical attenuations for 70/150 kV or ferent spectral attenuation characteristics of the
80/140 kV (Fig.4, lower right). The total atten- materials to be separated. Two-material decompo-
uation per mass concentration (mg Dy/mL) is sition algorithms can separate contrast media from
highest for dysprosium. However, the attenua- endogenous tissues (e.g., iodine and VNC) or a
tion also shows a significance dependence on specific endogenous material from contrast media
tube voltage albeit not at the level of iodine. and tissue (e.g., calcium and tungsten/tissue) or
Another interesting element is xenon even two contrast media (e.g., xenon and tungsten/
(K-edge = 34.65 keV), which possesses a simi- tissue). Higher order material decomposition algo-
lar attenuation profile with comparable signal rithms are feasible but also require a higher order of
ratios between high and low kV than iodine. input data and can lead to high image noise levels.
Contrast Media for Modern Computed Tomography 101

6  linical Applications of Dual-

C the full information about morphology and
Energy Material
iodine uptake can be provided by a single dual-
Decomposition
energy scan (Flohr et  al. 2006). This approach
was further evaluated for treatment response
6.1 Oncology Applications monitoring in oncology. Targeted anti-tumor
drug therapies as antiangiogenetic drugs or
Iodine, with its large spectral attenuation differ- tyrosine kinase inhibitors primarily induce
ence between low and high X-ray energies, can changes in the tumor microvasculature rather
be effectively separated from tissue that has a than a decrease in tumor size. In an experimen-
low or almost no energy-dependent difference tal glioma model in rats, treatment effects were
in attenuation (Fig.  5, upper right). One of the successfully monitored with Ultravist enhanced
early dual-source CT studies was the calculation dual-energy CT followed by the quantification
of iodine maps and virtual non-contrast (VNC) of the iodine uptake in the tumor. Notably, the
images in an experimental liver tumor rabbit dual-energy method was more effective than
model using an investigational iodine-based conventional single energy CT and showed
liver contrast medium to visualize hepatocytes comparable results at much lower radiation dose
(Fig. 5, lower left). This investigation on a first-­ than dynamic-contrast enhanced CT (Knobloch
generation dual-source CT demonstrated that et al. 2014).

low kV high kV low kV high kV

Mass Attenuation Coefficient (cm2/g)

Mass Attenuation Coefficient (cm2/g)

100 100

10 10

1 1

0.1 0.1
20 40 60 80 100 120 140 20 40 60 80 100 120 140
Energy (keV) Energy (keV)
Soft Tissue Clacified Tissue Iodine Soft Tissue Clacified Tissue Tungsten

Parenchymal spectral CT: Iodine and VNC images Vascular spectral CT: calcified plaque removal

Iodine VNC

Spin: 0 Spin: 0
Tilt: –90 Tilt: –90

Carotid Specimen

CT VNCa Ca

Fig. 5  Spectral X-ray attenuation of iodine, tungsten in the tumor, while the VNC images shows the morphology
comparison to soft and calcified tissues (upper row). The (Flohr et al. 2006). In vascular spectral CT tungsten offers
different spectral behavior of iodine and tungsten between accurate calcified plaque removal. In the human carotid
low and high tube potentials (marked with triple lines) artery specimen (lower right) investigated in experimental
enables different applications. Material decomposition is tungsten-­based contrast media solution (15mgW/ml) the
used for computation of iodine and virtual non-contrast lumen narrowing is visible on the virtual non calcium recon-
(VNC) images in parenchymal dual-energy CT. In the rabbit struction (VNCa) and calcium (Ca) burden is visible on the
tumor model (lower row left) the iodine map clearly shows calcium reconstruction map (Sartoretti et al. 2021)
102 H. Pietsch and G. Jost

6.2 Cardiovascular Applications sclerotic plaques (Fig.  5, lower right). In this

study the noise levels in the VNCa images were
Cardiovascular diseases represent a second major significantly improved by using an experimental
indication for contrast-enhanced CT.  Important tungsten (W3O2)-based contrast medium com-
clinical questions for CT angiography (CTA) are, pared to Ultravist as the iodine reference standard
for example, the accurate assessment of vascular (Sartoretti et  al. 2021). This results in superior
stenosis and the characterization of arterioscle- vessel visualization and better vessel wall delin-
rotic plaques. Dual-energy has a great potential eation. Thus, tungsten and hafnium-based con-
for improving CTA image quality and diagnostic trast media would offer new promising
confidence by reducing the negative effects of opportunities for cardiovascular dual-energy CT
blooming and beam hardening often induced by (Berger et  al. 2017; Sülzle et  al. 2015; Yu and
calcified plaques. The efficacy of material decom- Watson 1999).
position between iodinated contrast media and
calcified structures is limited, however. The spec-
tral attenuation characteristic of the two materials 6.3 Pulmonary Applications
is similar, resulting in comparable ratios between
high and low X-ray energies (Fig. 5, upper left). Several lung diseases such as chronic obstructive
Contrast media with higher K-edge energies such pulmonary disease (COPD) are associated with
as tungsten and hafnium can close this gap as regional ventilation disorders that are difficult to
they offer a significantly different dual-energy assess with global pulmonary function tests such
pattern than calcified tissues (Fig. 5., upper right). as spirometry or gas transfer tests. On the other
This in turn improves the material decomposition hand, new therapeutic options such as implanta-
into virtual non-calcium (VNCa) images contain- tion of endobronchial valves or improved lung
ing a contrast medium-soft tissue mixture and a volume reduction surgery require a precise defi-
calcium map containing the calcified tissue. The nition of regional lung function. Xenon-enhanced
resulting better plaque removal in the VNCa CT can be used for spatially resolved imaging of
reconstructions and the lower image noise in lung ventilation. Xenon is a radiopaque gas that
these images can significantly increase the diag- can be safely inhaled in concentrations of about
nostic accuracy of vascular imaging. This was 30%. This results in an image enhancement of
demonstrated with a coronary artery stenosis 20–30 HU (120  kV) in normally ventilated
phantom mimicking different degrees of stenosis regions. However, subtraction between unen-
at different plaque density levels (Sartoretti et al. hanced and Xe-enhanced images is challenging
2020). The vessel lumen was filled with different because imaging cannot be performed at exactly
investigational contrast media based on tungsten, identical respiration levels. Dual-energy CT cir-
hafnium, bismuth, holmium, or iodine. Dual- cumvents this procedure since the xenon enhance-
energy imaging was performed with a photon- ment can be determined at a single time-point
counting detector CT prototype at energy thresh- (Thieme et al. 2008, 2009). The spectral attenua-
olds optimized for the specific element of the tion of xenon is comparable to that of iodine,
contrast medium. Qualitative and quantitative leading to a large change in attenuation between
image evaluation performed on the VNCa images low and high X-ray energies (Fig.  4). Thus, a
revealed improved image quality and diagnostic xenon and a lung parenchyma image can be
evaluability of stenosis for tungsten, hafnium, reconstructed using material decomposition
and bismuth compared to iodinated contrast techniques.
media. Furthermore, the used tungsten- and Another important parameter that can be
hafnium-­ containing contrast media result in affected by disease is lung perfusion. Iodine-
lower VNCa image noise levels. Further experi- enhanced dual-energy CT significantly improved
mental evidence was obtained in a similar study the diagnostic workup, as this enables the recon-
on a human carotid artery specimen with athero- struction of morphological images and functional
Contrast Media for Modern Computed Tomography 103

images that visualize perfusion (Dawson 2006). perfusion. These elements only show as little or
In terms of physiological aspects, pulmonary almost no dependence of attenuation on the tube
function depends on lung ventilation and lung potential and can be effectively separated from
perfusion to ensure an effective gas exchange. xenon by dual-energy material decomposition
Both can be affected by disease, i.e., a ventilation (Fig.  6a). The feasibility of this approach was
perfusion mismatch is a typical characteristic in demonstrated in a study on healthy pigs. The
COPD patient. The degree of this mismatch and anesthetized animals were ventilated with a 30%
its regional distribution have a potential to xenon, 70% oxygen mixture for 120  s and an
improve the diagnosis and therapeutic decisions. investigational tungsten-based contrast medium
To date, such ventilation and perfusion mea- was injected in temporal coordination. Repeated
surements can be made by means of nuclear med- dual-energy imaging (80/140 kV) was performed
icine method such as SPECT, using radiolabeled to capture the lung during different contrast
gases and tracers in sequential mode. Dual- medium phases on a third-generation dual-source
energy CT might be a promising alternative that CT.  A xenon material map and a virtual non-­
may even enable simultaneous lung ventilation xenon (VNXe) image, containing soft tissue and
and perfusion imaging. Furthermore, it would tungsten were reconstructed for each time-point
offer a much higher spatial resolution, an easier (Fig.  6b). Xenon and tungsten could be clearly
examination taking at much less time, and impor- separated enabling simultaneous assessment of
tantly also a high availability of scanning units. lung ventilation and perfusion.
However, xenon for ventilation and iodine for
perfusion imaging have a very similar attenuation
profile and cannot be separated using dual-energy 7 Summary and Outlook
methods. This requires the sequential application
of Xe and iodine resulting in repeated scanning There had already been systematic investigations
and images acquired at different breathing phases of alternatives to iodine as a contrasting element
or breath-hold levels (Thieme et al. 2008). This in the early days of X-ray imaging. Since then,
might be overcome, however, by using contrast numerous contrast media approaches with ele-
media elements with high K-edge energies such ments other than iodine have been explored. For
as hafnium and tungsten to visualize pulmonary a variety of reasons, however, none has made it to

a b
low kV high kV
Mass Attenuation Coefficient (cm2/g)

Tissue/Tungsten

100

10

1
Xenon

0.1
20 40 60 80 100 120 140
Energy (keV)
Xenon Tissue Tungsten time

Fig. 6  Separation of two contrast materials. (a) Spectral generation Dual Source CT) of the lung in a porcine
X-ray attenuation of xenon, tungsten and tissue. The dif- model. Inhalation of 30% xenon for 2  min followed by
ferent spectral behavior of xenon and tungsten between intravenous administration of an experimental tungsten-­
low and high tube potentials (marked with triple lines) based contrast medium. Xenon and tungsten could be
enables dual-energy-based material decomposition of clearly separated enabling simultaneous assessment of
both contrasting materials. (b) Functional imaging (third lung ventilation and lung perfusion
104 H. Pietsch and G. Jost

the market. As described earlier, the advantage of such contrast media could make an important
such elements is the higher attenuation of X-rays contribution by improving the contrast enhance-
compared to iodine, which depends on the mass ment per molecule, which would allow, on the
attenuation coefficient of the contrast element one hand, a reduction of the radiation dose with
and the energy distribution of the X-ray spec- an acceptably higher noise level but the same
trum. Elements with higher atomic numbers and contrast-to-noise ratio. Alternatively, a better
thus higher k-edge energies better fit to the CT image quality in respect of contrast-to-noise can
tube energy spectrum than iodine. Therefore, ele- be obtained with the same radiation dose. Both
ments such as the lanthanides, hafnium, tanta- would allow the use of contrast-enhanced CT in
lum, tungsten, or even gold are better suited as further clinical areas.
contrasting elements. In view of the rapid developments in CT tech-
In the recent past, there have been very prom- nology, the use of CM will continue to be of cen-
ising investigations of hafnium with the aim of tral importance as an integral part of clinical
reducing the radiation exposure in contrast-­ routine. In particular, the trend spectral CT
enhanced CT imaging (Frenzel et al. 2015, 2016; through photon-counting CT technology will
Berger et al. 2017). In this context, the element expand the application possibilities of new con-
hafnium was combined with a strongly chelating trast media with higher atomic numbers. Their
polydentate ligand. Both the physicochemical spectral attenuation characteristics will certainly
properties and the tolerance in animal studies allow completely new ways of improved material
were encouraging. A new generation of multi- separation and tissue characterization. This holds
dentate tungsten metal clusters with high chemi- great potential to increase diagnostic perfor-
cal stability, generally, an essential requirement mance and to enable additional clinical fields and
for CM, was also found (Sülzle et  al. 2015). indications.
These compounds also proved to be quite well The development of X-ray technology has
tolerated and showed high solubility in order to always gone hand in hand with the development
achieve high tungsten concentrations comparable and continuous improvement of suitable contrast
to the iodine concentrations of commercial CM. media (Barrs 2006). This will no doubt also con-
Another interesting approach are nanoparti- tinue in the future.
cles (NPs) which contain the absorbing element
in a very high density. Depending on their size Compliance with Ethical Standards 
and coating, nanoparticles have a different phar-
macokinetic profile and biodistribution compared The manuscript has not been submitted to more than one
to currently available iodinated CM, thus limiting publication simultaneously. The submitted work is original
and has not been published elsewhere in any form or lan-
their distribution to the intravascular space (Dong guage (partial or complete). The authors have followed the
et al. 2019). One concern is the slow and incom- rules for obtaining, selecting, and processing data and have
plete elimination of NPs from the body. Gold as a not presented any data, texts, or theories of other authors as
source material for NPs has been experimented if they were their own.
with again and again. The problem of rapid and
sufficient excretion via the kidneys and the high
cost have limited the progress of gold-based NPs References
for use in humans (Allijn et  al. 2013; Cormode
et al. 2010). Allijn IE et al (2013) ACS Nano 7:9761
CT imaging has great clinical utility and can Bae KT, Tran HQ, Heiken JP (2000) Radiology 216:872
Bae KT, Tran HQ, Heiken JP (2004) Radiology 231:732
only be partially replaced by other radiation-free Barrs TJ (2006) Am J Health Syst Pharm 63:2248
techniques. Therefore, reducing the radiation Behrendt FF et al (2013) AJR Am J Roentgenol 200:1151
exposure required for CT is an important goal Berger M et al (2017) Inorg Chem 56:5757
(Schockel et al. 2020; Lusic and Grinstaff 2013; Claussen CD, Banzer D, Pfretzschner C, Kalender WA,
Schorner W (1984) Radiology 153:365
Dong et al. 2019; Shahid et al. 2020). Here, too,
Contrast Media for Modern Computed Tomography 105

Cormode DP et al (2010) Radiology 256:774 Palkowitsch PK, Bostelmann S, Lengsfeld P (2013) Acta
Dawson P (2006) BMJ 333:663 Radiol 55:707
Dong YC et al (2019) Sci Rep 9:14912 Pietsch H et al (2009) Eur J Radiol 80:349
Fleischmann U et al (2018) Investig Radiol 53:264 Pietsch H, Hübner-Steiner U, Seidensticker P (2012)
Flohr T et al. (2006) Paper presented at the RSNA; 2006, Ullmann’s encyclopedia of industrial chemistry, 7th
Chicago 2006 edn. Wiley-VCH, Weinheim
Frenzel T et al (2015) Investig Radiol 50:671 Roessler AC et al (2016) Investig Radiol 51:249
Frenzel T et al (2016) Investig Radiol 51:776 Rosovsky MA et al (1996) Radiology 200:119
Gutjahr R et al (2016) Investig Radiol 51:421 Sartoretti T et al (2020) Invest Radiol
Jost G et al (2009) Investig Radiol 44:114 Sartoretti T et al (2021) Atherosclerosis 310:11
Knobloch G, Jost G, Huppertz A, Hamm B, Pietsch H Schöckel L et al (2020) Investig Radiol 55:592
(2014) Eur Radiol 24:1896 Shahid I, Lancelot E, Desche P (2020) Investig Radiol
Kopp AF et al (2008) Acta Radiol 49:902 55:598
Langer M, Felix R, Keysser R, Speck U, Banzer D (1985) Sülzle D, Bauser M, Frenzel T (2015) J Clust Sci 26
Digitale Bilddiagn 5:154 Thieme SF et al (2008) Eur J Radiol 68:369
Lusic H, Grinstaff MW (2013) Chem Rev 113:1641 Thieme SF et al (2009) Eur Radiol 20:2882
Nowak T, Hupfer M, Brauweiler R, Eisa F, Kalender WA Yu SB, Watson AD (1999) Chem Rev 99:2353
(2011) Med Phys 38:6469
 Part II
Clinical Applications
 Neuroradiological Imaging

Sebastian Winklhofer, Dominik Nakhostin,
and Mohammed Fahim Mohammed

Contents
1 Introduction   110
2 Technical Background and Workflow Considerations   110
2.1  Background   110
2.2  Training Considerations for Technologists and Radiologists   110
2.3  Patient Selection and Scan Acquisition   111
2.4  Protocol and Dose Optimization   112
2.5  Image Reconstruction and Storage   112
2.6  Considerations for Image Interpretation   115
3 Clinical Applications   115
3.1  I ntroduction to Clinical Applications   115
3.2  Differentiation Between Contrast Staining and Intracranial Hemorrhage   115
3.3  Further Material and Tissue Differentiation   115
3.4  Image Quality, Radiation Dose, and Artifact Reduction   118
3.5  Other Applications of DECT in Emergency Neuroradiology   121
4 Photon Counting   122
5 Outlook and Conclusion   122
References   123

S. Winklhofer (*)
Department of Neuroradiology, Clinical
Neuroscience Center, University Hospital Zurich,
University of Zurich, Zurich, Switzerland
e-mail: [email protected]
D. Nakhostin
Department of Neuroradiology, Clinical
Neuroscience Center, University Hospital Zurich,
M. F. Mohammed
University of Zurich, Zurich, Switzerland
Medical Imaging Department – CR and Corporate
Department of Diagnostic and Interventional Clinical Performance and Innovation Department,
Radiology, University Hospital Zurich, University of Ministry of the National Guard Health Affairs,
Zurich, Zurich, Switzerland Riyadh, Saudi Arabia
e-mail: [email protected] e-mail: [email protected]

© The Author(s), under exclusive license to Springer Nature Switzerland AG 2022 109
H. Alkadhi et al. (eds.), Spectral Imaging, Medical Radiology Diagnostic Imaging,
https://doi.org/10.1007/978-3-030-96285-2_8
110 S. Winklhofer et al.

Abstract example is trauma and emergency imaging,

where CT plays a major role in the rapid and
Dual-energy computed tomography (DECT) accurate triage and diagnosis of patients. DECT
has proven itself as an important innovation inmight not only add benefit for image interpreta-
neuroimaging. Technical aspects and practical tion in such clinical situations, but may also
considerations to incorporate dual-energy reduce the required radiation dose or the number
scanning into routine practice including of required follow-up imaging, and therefore
patient selection, recommendations for recon- might result in a reduction of costs for the health-
structions, and image interpretation for rou- care system (Wong et al. 2020).
tine head CT examinations are discussed.
Selected established or newly introduced
applications have found their way into daily 2 Technical Background
clinical routine, which include various topics and Workflow
such as emergency neuroradiology, stroke Considerations
imaging, or applications affecting the image
quality or radiation dose. In addition, this 2.1 Background
chapter provides an outlook to the future of
spectral neuroimaging including photon-­ Here we will discuss technical aspects and prac-
counting detector-based CT and applications tical considerations to incorporate dual-energy
for artificial intelligence. scanning into routine practice including patient
selection, approaches to image reconstruction,
issues related to image storage and recom-
mended reconstructions for routine head CT
1 Introduction examinations.
The precise technical principles of the image
A new tool in computed tomography based neu- acquisition are described in detail in the first part
roimaging? Yes! Spectral imaging, in particular of this book. In summary, several dual-energy CT
dual-energy computed tomography (DECT), has systems are available today which are suitable for
proven itself as an important innovation in neuro- neuroimaging. They can be broadly classified
imaging. DECT has come a long way from its into scanner systems with two x-ray sources
first experimental studies and today is a well-­ (dual-source DECT) or with one x-ray source
understood and clinically applicable technology (single-energy DECT). The latter includes sev-
in neuroradiology. Current literature does not eral different technical backgrounds including
only demonstrate the feasibility of clinical DECT the fast kV switching, the dual-layer detector, the
applications, but also the evidence-based value twin-beam or the dual-spiral approach. All of
for the patient. This in itself is very exciting as these commercially available techniques of the
just a few decades ago the future of CT in neuro- main scanner vendors have found their way into
imaging in particular was believed to be limited clinical routine DECT in neuroimaging today.
and that the wide availability of MRI would over-
shadow it in clinical practice.
CT has made great advances and is now 2.2 Training Considerations
increasingly in the focus of modern diagnostic for Technologists
and therapeutical clinical guidelines. One exam- and Radiologists
ple is the rapidly evolving topic of stroke imag-
ing, where highly relevant clinical decisions (i.e., To ensure the smooth introduction of a new ser-
the decision to treat the patient with a certain vice, the involvement of key stakeholders and the
therapy such as mechanical thrombectomy or engagement of technologically and clinically
not) are mainly based on CT imaging. Another well-trained professionals is essential. Identifying
Neuroradiological Imaging 111

these dedicated colleagues and training them to case of a highly agitated patient that cannot lay
be super-users will facilitate an improved transi- still). Having these algorithms clearly outlined
tion to the launch of any new service. Assigning a can streamline workflows and avoid confusion at
handful of technologists to become super-users the time of scanning.
by training them on scanning techniques and Although selective scanning may have the
concepts through in-service lectures, simulated benefit of limiting the workflow impact of intro-
training with the vendors, and hands-on scanning ducing DECT into a department’s practice, it is
and reconstruction techniques by application spe- sometimes impossible to predict which patients
cialists or through workshops are all essential to may benefit from a dual-energy scan. As the list
build capacity and improve confidence so that of clinically validated indications of DECT of the
they may scan and troubleshoot as needed. It head continues to grow beyond what is high-
would also provide them with the judgment lighted in this chapter, it might become extremely
needed to adjust protocols as clinically appropri- challenging to continue updating patient selec-
ate. They can then transfer knowledge to col- tion criteria based on an indication or prospec-
leagues and ensure that the service is running tively predicted benefit. Alternatively, routine
efficiently from a technical standpoint. implementation of DECT scans for all patients
Similarly, training and capacity building of may be more feasible, especially after gaining
radiologists begins with an introduction of the more experience with DECT scanning and under-
applications of DECT through lectures and scien- standing any potential impact on both technolo-
tific meetings. Familiarity with the applications gist and radiologist workflows.
can be improved through hands-on or online The number and physical distribution of
workshops. DE-capable scanners is an important consider-
ation in designing the DECT workflow. For
example, in practices that have a mix of
2.3  atient Selection and Scan
P DE-capable scanners and SECT-only scanners, it
Acquisition may be more practical to scan all or the majority
of patients on the DE-capable scanners in DE
The decision to acquire dual-energy scans must mode in order to streamline the workflow and
be made prospectively on nearly all the current improve technologist familiarity with DE tech-
DECT platforms available on the market—the niques and protocols.
exception being the dual-layer (or “sandwich”) However, if a practice or division only has
detector DECT platform which can display access to a single scanner that offers DECT capa-
DECT data retrospectively as the low and high bilities (such as a scanner in the emergency
energy x-ray spectra are split at the detector level. department or outpatient imaging centers), the
Patients could be selected based on referral scanner’s technical limitations may guide the
departments (e.g., all patients from the oncology decision to selectively scan in the dual-energy
or the emergency department), specific clinical mode. For example, while dual-energy scanning
indications (e.g., post intra-arterial thrombec- time is generally comparable to single-energy
tomy for stroke, intracranial hemorrhage assess- acquisitions, some scanners such as earlier rapid-
ment, etc.) or based on broader criteria, such as kV switching scanners may have prolonged
­
the anatomic area assessed (such as all adult reconstruction times which may impact patient
brain CTs). To ensure a smooth adoption within flow through the department and must be taken
the department, the selection criteria must be into consideration, particularly in emergency
agreed upon by all stakeholders and a clear algo- settings.
rithm must be created, including what should be Due to the comparatively small field of view
the default protocols for certain indications and (FOV) associated with scanning of the head, the
details on when it is acceptable to switch to scanner’s FOV is of little concern when consider-
single-­energy scan mode (for example, in the ing single- or dual-energy scan modes.
112 S. Winklhofer et al.

2.4 Protocol and Dose energy acquisitions to simulate a single-energy

Optimization acquisition. Alternatively, virtual monoenergetic
images (VMI) reconstructed at 65–70  keV are
In the last few years, there has been a prolifera- considered 120 kVp equivalents as well. These
tion of various DE/Spectral CT systems on the basic images should automatically be generated
market (at least seven by the authors’ count), all for every scan and sent to PACS for primary
with unique approaches to image acquisition and interpretation.
data reconstruction. Hence, it would be beyond The more advanced dual-energy reconstruc-
the scope of this chapter to provide a comprehen- tions may be generated through several
sive list of protocols suitable for each scanner approaches, each offering various advantages and
type. Instead, engaging the vendors is essential. drawbacks. Broadly, we can divide reconstruc-
When creating new DECT protocols, we recom- tion approaches to automated reconstructions,
mend creating a team consisting of a radiologist, technologist-driven reconstructions, and
technologist, medical physicist, and application radiologist-­driven reconstructions. Automated
specialist from the vendor. A clear expectation of reconstructions may be performed on the CT
image quality and radiation dose must be out- console or on a dedicated workstation depending
lined at the start. With this approach, protocols on the vendor and the reconstruction options
can be created with a specific benchmark in mind. available. For example, the CT console may
The protocols then can be tested on various phan- allow for VMI reconstructions or gray-scale vir-
toms to ensure appropriate image quality, noise tual non-contrast images but not colored overlay
levels, and radiation dose as outlined by govern- maps or multiplanar reformats. In addition, scan-
ment or institutional guidelines. We recommend ner console-based reconstruction may limit
testing the protocols on dedicated DECT phan- access to the CT scanner and delay imaging of
toms which are commercially available. Optimal the next patient, in which case this approach may
image parameters have to be adapted according be unfavorable especially in an already busy
to the available imaging system and to the indi- department operating near capacity. Hence,
vidual situation and needs of each radiology workstation-based automation is preferred by the
department. authors whenever available. Another drawback of
Similar aspects are valid regarding the injec- automated reconstructions is a significant
tion of contrast materials. However, the amount increase in number of images generated per
and the injection rate of iodine are quite similar study. In our practice, the number of images in a
compared to standard SECT protocols. head CT has increased from an average of 300–
It is important to note that DE/Spectral CT 400 to 1200–1500 images. This will increase the
scans are either dose-neutral or dose-negative for amount of storage needed and may have a nega-
the majority of indications as compared to SECT tive impact on study interpretation if the radiolo-
scans (van Ommen et al. 2019; Pomerantz et al. gists are not familiar with the additional
2013a; Neuhaus et al. 2017). reconstructions or their clinical use. However, the
presence of images at the time of study interpre-
tation increases the likelihood of utilization of
2.5 Image Reconstruction DECT reconstructions, especially when radiolo-
and Storage gists are aware of the clinical applications and
method of interpretation.
Following the acquisition of the raw data, the Technologist-driven image reconstructions
images are processed and a “single-energy equiv- could be performed manually at the CT console
alent” image is generated to mimic a 120 kVp or on a dedicated workstation, especially if the
acquisition. This may be generated by linear-­ department has a 3D/advanced imaging lab. The
blended (weighted average or mixed images) main advantage of this approach is that the recon-
techniques which mix data from low and high structions can be generated on-demand and as
Neuroradiological Imaging 113

needed based on the findings on the images pro- For all cases, the single-energy equivalent
vided by the available raw data. This may also images should always be sent to the
reduce the load on PACS storage as not all studies PACS. Additionally, it is strongly recommended
will have advanced reconstructions sent for that the DE raw data are sent to PACS as well.
archiving. However, this approach may increase This will allow for any additional advanced DE
the workload of technologists which may delay reconstructions should the need arise, enable
scanning of patients or increase turn-around time research, and allow for retrospective application
in the CT room. It also requires that technologists of any new techniques added to the reconstruc-
are well trained on DECT systems and their dedi- tion software. Additionally, in the case of a non-­
cated reconstruction techniques. Another disad- contrast head CT, we recommend the inclusion of
vantage of this approach is the possible low a 65  keV VMI series to improve overall image
utilization of advanced DECT reconstructions by quality, a 190 keV VMI series to improve detec-
radiologists if they are not readily available at the tion of intracranial hemorrhage and decrease
time of interpretation as they may not be aware metal artifacts or streaking artifacts in the poste-
that the images were acquired using a DE tech- rior fossa and finally, a brain hemorrhage algo-
nique. This could be the case if the image data are rithm to differentiate between hemorrhage or
not labeled appropriately or due to the loss in calcium. In the case of post intra-arterial throm-
productivity by requesting the reconstructions bectomy, the addition of a colored iodine map
and waiting for them to be sent over to the PACS. and virtual non-contrast (VNC) images is recom-
The radiologist-driven approach is usually ini- mended. For post-contrast studies (CTA, CTV, or
tiated at the time of image interpretation. The routine post-contrast), the addition of bone-­
images may be reconstructed at a stand-alone subtraction images and low keV (40–50  keV)
workstation; however, it is in the authors’ experi- VMI images add value to interpretation (Table 1).
ence that this approach is the least likely to drive Each practice must identify the scope and
utilization or engagement as it takes away from goals of implementing a DECT service in their
the task of interpreting the study and negatively practice. A careful assessment of the pros and
affects productivity, particularly in a high-­volume cons of each approach and an understanding of
radiology setting, such as an emergency radiol- how each might impact or compliment the cur-
ogy department. A more practical approach is the rent workflow is essential to the successful inte-
utilization of a thin client—a server-based ver- gration of DECT into practice. At one of our
sion of the advanced imaging workstation that institutions, a high-volume academic practice,
can be accessed through the internal hospital net- the service was launched by intensive training of
work or over the internet—integrated into the the residents and fellows on DECT concepts,
PACS.  This would allow the radiologist to pro- techniques, reconstruction methods, and the use
cess the DECT data while interpreting the study. of the thin client. The required images were gen-
This approach would likely have the lowest erated by the resident at the time of primary inter-
impact on the number of images sent for perma- pretation and were reassessed by the attending
nent archiving as the radiologist could select and radiologist during the final readout. Any addi-
send only the significant and pertinent images, tionally required images were generated at read-
ignoring the rest. It would also improve the radi- out by the attending. Currently with the increase
ologist’s understanding of DECT reconstruc- in familiarity with DECT, the practice has shifted
tions, improving confidence and encouraging to automated generation of the highest impact
experimentation. However, this approach also reconstructions based on experience and litera-
has the largest negative impact on the productiv- ture, so that the reconstructions are available to
ity of the radiologist, as each study that requires all at the time of interpretation without delays in
DECT data reconstruction has to be loaded into workflow. The projected storage requirements
the thin client and each algorithm must be pro- have also been scaled to consider the increase in
cessed and assessed individually. number of images per study.
114 S. Winklhofer et al.

Table 1 Commonly DECT Recommended

Reconstructions for Head CT
Reconstruction Type Description Function Image Example

utilized DECT reconstruc- Standard Mixed Images Single-energy equivalent First line image assessment
tions in clinical practice
and their practical
applications

Iodine Overlay Images A fused set of images that • Assessment of contrast

displays the distribution of enhancement
Iodine content overlayed on a • Differentiating blood vs.
grey-scale image iodine

Virtual Monoenergetic Extrapolated images that Intermediate Energy VMI (65-

Images (VMI) simulate tissue behavior at a 70 keV):
single selected energy level • Improved image quality
(keV) • Reduced posterior fossa
artifacts

High Energy VMI (140-190

keV):
• Reduced artifacts
• Improved conspicuity of
hemorrhage

Low Energy VMI (40-50 keV): 190 keV Image to improve

• Increase tissue contrast detection of subdural
• Increase iodine hematomas
attenuation

Bone-subtracted Images Subtraction of calvarial bone • Allows for improved image

based on dual-energy reconstruction in post
material identification and contrast studies
decomposition • May also contribute to
improved detection of
hemorrhage along the
calvarium

Virtual Non-Contrast Images Subtraction of Iodine; based • Complimentary to iodine

(VNC) on dual-energy material overlay images
identification and • Differentiates hemorrhage
decomposition or calcium from iodine

VNC image of the above post-

thrombectomy case

Raw Data Raw data file or source • Allows for prospective

images used for dual-energy DECT reconstructions
reconstruction

Low Energy (80 kVp)Images

High Energy (Sn140 kVp)

Images
Neuroradiological Imaging 115

2.6 Considerations for Image entiation of hyperdensities in a non-enhanced CT

Interpretation (NECT), which could signify both intracranial
hemorrhage (ICH) and contrast extravasation
When implementing a DECT service for the first (after prior administration of intravenous contrast
time, there may be some apprehension from col- agent, such as in diagnostic angiographies
leagues due to the sudden introduction of several (DSA)). Zaouak et  al. (Zaouak et  al. 2020)
unfamiliar images into routine imaging studies. It is reported a specificity and accuracy of up to 100%
essential to communicate clearly what changes will for the differentiation between cerebral hemor-
be taking place, which images will be included, the rhage and iodine leakage due to a disturbed
clinical value of any additions based on current lit- blood–brain barrier after neurointerventional
erature, and how to best utilize any additional procedures. A large meta-analysis by Choi et al.
images. It is also normal and expected that there will (Choi et  al. 2020) published in 2020 confirmed
be a learning curve at the launch of the service which the excellent diagnostic performance of DECT in
may increase the time spent on each study. Creating said regard, however showing a relatively large
awareness early on will help with adoption. risk of publication bias. In addition, said inter-
A systematic approach to image interpretation ventional application also has a direct clinical
will reduce errors. It is our recommendation that impact: Chen et  al. (Chen et  al. 2020) showed
for any study, the single-energy equivalent that contrast extravasation is a predictor of poor
images be utilized for primary interpretation in a clinical outcomes in patients undergoing endo-
similar way to standard practice. The dual-energy vascular therapy for acute ischemic stroke
reconstructions can then be used as a secondary (Fig. 1).
read and for problem-solving depending on the Furthermore, the generation of virtual non-­
specific clinical scenario. contrast (VNC) images from contrast-enhanced
CT images (particularly CT angiographies)
promises to be a valuable tool in order to generate
3 Clinical Applications precontrast images without the additional radia-
tion dose of acquiring a true non-contrast image.
3.1 Introduction to Clinical However, a study from Bonatti et al. from 2017
Applications (Bonatti et  al. 2017) showed that intracranial
hemorrhages were somewhat less conspicuous
A number of promising clinical DECT applica- on VNC images compared to true non-contrast
tions have been introduced and developed in the images and particularly the extent of an ICH
past years. Selected established and proven appli- might be underestimated in VNC images.
cations have found their way into daily clinical An application which yet has to be scientifi-
routine, which include various topics such as cally evaluated but shows great potential benefit
emergency neuroradiology, stroke imaging, or in contrast-enhanced DECT is the differentiation
applications affecting the image quality or radia- between hemorrhage and neoplastic contrast
tion dose. These may add value to already exist- enhancing cerebral lesions (hyperdense) within a
ing imaging concepts or show new CT tumor associated (hyperdense) surrounding hem-
applications in neuroimaging. orrhage, as seen in Fig. 2.

3.2 Differentiation Between 3.3  urther Material and Tissue

F
Contrast Staining Differentiation
and Intracranial Hemorrhage
In addition to differentiating between hemor-
One of the most widely used and accepted appli- rhage and iodine from contrast agents, DECT
cations of DECT in neuroradiology is the differ- can also differentiate between iodine and cal-
116 S. Winklhofer et al.

a Standard mixed b VNC c Iodine Map

100 [HU]

10mm

0 [HU]

Fig. 1  Dual-energy CT after mechanical thrombectomy ous thrombectomy. In VNC images (b), the hyperdensities
in a patient with acute left-sided ischemic stroke. are not visible anymore, indicating the absence of hemor-
Parenchymal hyperdensities are visible in the left lenti- rhage. Iodine images (c) demonstrate color coded hyper-
form and caudate nucleus in standard mixed images (a). densities in the area of the hyperdensities seen in the
In this case, it is unclear whether these hyperdensities are mixed images. These findings indicate an iodine extrava-
due to hemorrhage or iodine extravasation from the previ- sation without any intracranial hemorrhage

a Non Contrast Mixed b Contrast Enhanced c Iodine Map

100 [HU]

15mm 15mm

0 [HU]

Fig. 2 (a) Non-contrast CT of the head demonstrates a from DECT demonstrates a color coded part within the
right-sided parenchymal hyperdensity with a perifocal hyperdensity in A and B indicating a contrast enhancing
hypodense edema. This scan was interpreted as an intra- tumor. Parts of the hyperdense lesion in A and B do not
cranial hyperdensity with a potential underlaying tumor or show iodine uptake in C, indicating hemorrhagic compo-
vascular malformation. (b) Contrast-enhanced scan dem- nents (arrows). The patient received surgery and histopa-
onstrated no obvious enhancement. (c) The iodine map thology revealed an atypical teratoid rhabdoid tumor

cium containing structures in NECT (Tran et al. virtual rendering technique (VRT) or maximum
2009). This possibility leads to several clinical intensity projection (MIP) images for an
applications, among the most frequently used improved (clinical) vessel visualization (Figs. 3
ones being the detection of (intraosseous) cal- and 4). Bone subtraction with DECT can be
cium and subsequent bone removal (bone sub- used to avoid artifacts adjacent to the vessel
traction) in CT angiography, providing clear from bony or calcified structures so that vascu-
Neuroradiological Imaging 117

a b c

d e f

Fig. 3  Arterial phase dual-energy CT angiography with image, (d) bone subtraction, (e) volume rendering tech-
bone subtraction of the intracranial arteries. (a) 80 kVp nique (VRT) for advanced 3-dimensional visualization,
acquisition, (b) 150 kVp acquisition, (c) mixed (weighted) (f) cinematic rendering reconstruction

lar pathologies such as aneurysms or vessel ste- Another often encountered problem in neuro-
noses can be nicely visualized (Zhang et  al. radiology, particularly in the emergency setting,
2010; Korn et  al. 2015) in particular by using is the uncertainty of differentiation of hyperdense
VRT or other 3-dimensional rendering tech- foci into (benign) calcifications and intracranial
niques. The problem of a potential blooming- hemorrhage (Fig.  5). Wiggins et  al. showed in
associated overestimation of the grade of a their 2019 paper that with the application of
vessel stenosis could be overcome by applying DECT and subsequent reconstruction of virtual
an improved and modified 3-material decompo- non-calcium and calcium overlay images, the
sition algorithm for calcium removal as shown diagnostic accuracies can be improved substan-
by Mannil et  al. (Mannil et  al. 2017). This tially (Wiggins et al. 2020).
approach increases the accuracy of calcified By using a 3-material decomposition in DECT
plaques removal and shows similar stenoses postprocessing with parameters adjusted to sup-
degrees as compared to the gold standard of press the gray/white matter contrast, these recon-
digital subtraction angiography (DSA) images structions are able to more accurately detect
as reference. edema and the end-infarct volume as compared
118 S. Winklhofer et al.

a b c

d e f

Fig. 4  A patient with a large acute parenchymal hemor- visible in the bone removal images (d) where one can also
rhage seen in a single-energy CT scan of the head (a). see enlarged veins indicating an arterio-venous malforma-
Virtual non-contrast (VNC) images (b) reconstructed tion (AVM). 3-D maximum intensity projection (MIP) (e)
from the dual-energy CT angiography confirm the pres- and volume rendering technique (VRT) (f) reconstruc-
ence of hemorrhagic hyperdensities. Iodine images (c) tions from DECT allow for a better understanding of the
demonstrate a clear iodine containing anterior part of the complex malformation with the arterial inflow (arrow)
lesion with a tubular structure. These findings are better and the venous outflow (arrowhead)

with the initial true non-contrast images. This the image quality for the edema assessment can
approach might allow for a better assessment of be improved, but DECT also allows to reduce the
the degree and extent of infarction and may fur- mentioned artifacts which can be helpful to delin-
ther serve to better guide stroke therapy eate posterior fossa ischemia. However, sensitiv-
(Mohammed et  al. 2018; Grams et  al. 2018; ity from DECT is still lower compared to MRI,
Taguchi et  al. 2018; Hopf-Jensen et  al. 2020) so that MRI remains to be the preferred modality
(Fig.  6). Other studies showed similar potential for lesions in this region (Hixson et al. 2016).
benefits for the detection of ischemic brain paren-
chyma also based on other DECT systems
(Lennartz et al. 2018) or DECT reconstructions 3.4 I mage Quality, Radiation
such as virtual monoenergetic reconstructions Dose, and Artifact Reduction
(van Ommen et al. 2021).
Furthermore, DECT has been shown to be A prerequisite for the implementation of DECT
beneficial for the diagnosis of ischemia in the in everyday clinical practice is an image quality
posterior fossa. The latter is a region which is fre- that meets the clinical requirements while at the
quently affected by artifacts due to beam harden- same time keeping the radiation dose as low as
ing and photon starvation. With DECT, not only possible. Even if additional information is gained
Neuroradiological Imaging 119

a b c

Fig. 5 Differentiation intracranial hemorrhage versus former hyperdensity is no longer visible, indicating that
calcifications by using DECT. A faint parenchymal hyper- this is in fact a calcified lesion. This was confirmed by a
density in the right frontal brain which is not easy to be previous MRI scan in the susceptibility weighted image
classified as hemorrhage or calcification in standard brain (SWI) (c) from two years prior to the CT, which was not
CT (a). Dual-energy can help to answer this challenging available at the time of the initial CT (a) evaluation by the
question by using virtual non-calcium images (b). The radiologist

from spectral imaging, it must be ensured that the et al. 2013). Furthermore, DECT also has potential
two main aspects, image quality and radiation image quality applications in pediatric patients.
dose, correspond mostly to a standard SECT Weinmann et  al. showed that DECT can at least
image acquisition. The SECT equivalent stan- maintain or even improve the image quality in pedi-
dard weighted (mixed) image reconstruction of atric head CT while at the same time significantly
the head is used for first-line image evaluation; reduce the radiation dose (Weinman et al. 2019).
hence, its quality should be prioritized so that it is Another often encountered challenge is the
indistinguishable from a true SECT acquisition differentiation between naturally occurring beam
and must be able to show subtle changes of the hardening artifacts close to the neurocranium and
brain parenchyma for example, as in the case of actual intracranial hemorrhage. Beam hardening
an acute ischemic stroke. occurs when a polychromatic x-ray beam passes
Weighted average DECT images of the brain through a very dense object (such as the bone in
have been shown to yield less artifacts at lower radi- the calvaria), leading to preferably attenuation of
ation doses compared to standard SECT acquisi- low-energy photons and consequent artifacts,
tions. At a comparable dose level, the signal to noise such as streaking artifacts. The reconstruction of
ratio (SNR) has also been shown to be higher and virtual monochromatic images (at different keV
the image noise has been shown to be lower in values) allows to virtually reconstruct images
DECT. However, when comparing the gray matter/ which minimize artifacts (particularly in the pos-
white matter (GM/WM) contrast-­ to-­
noise ratio terior fossa), while at the same time maintaining
(CNR), SECT was superior to DECT (Dodig et al. a high CNR and signal to noise ratio. Pomerantz
2020). Nevertheless, in several studies it has been et  al. showed in their work that the maximum
shown that the latter disadvantage could be com- CNR and SNR values were observed at 65 keV
pensated by monoenergetic DECT reconstructions while posterior fossa artifacts were best reduced
and CNR could even be improved compared to at 75 keV (Pomerantz et al. 2013a).
polychromatic SECT CT with the same or even In addition, another clinical challenge in neu-
lower radiation doses (Neuhaus et  al. 2017; roradiology is beam hardening artifacts associ-
Pomerantz et al. 2013b; Zhao et al. 2018; Kamiya ated with surgical clips or endovascular coils
120 S. Winklhofer et al.

a b

c d

Fig. 6  Improved visibility of ischemic brain parenchyma improved visibility of the ischemic stroke with a higher
in a patient with right-sided acute ischemic stroke after difference between the HU in the ROI in the ischemic
mechanical thrombectomy. Standard DECT mixed image parenchyma versus in the healthy parenchyma in the con-
(a) demonstrates faint hypodensities (arrows) which are tralateral side in VNC (c) images (difference  =  10 HU)
better visible in the VNC images (b) (arrows). Quantitative compared to standard mixed images (d) (difference  =  5
region of interest (ROI) measurements confirm the HU)

used for the treatment of intracranial aneurysms. getic extrapolation from DECT leads to a signifi-
Using DECT and monoenergetic reconstructions cant reduction of clip artifacts, especially when
(e.g., 95 keV), artifacts from surgical clips can be used in combination with iterative reconstruction
reduced relatively well, whereas artifacts from methods (Winklhofer et al. 2018).
endovascular coils can hardly be reduced The generation of virtual monoenergetic
(Mocanu et al. 2018) (Fig. 7). In another study, it images has numerous other implications for clin-
was shown that the usage of virtual monoener- ical practice, for instance in improving the detec-
Neuroradiological Imaging 121

tion of early ischemic changes in stroke patients. to the reduced beam hardening artifacts, increased
Ståhl et  al. showed in their 2020 work (Ståhl contrast index (CI, a measure of relative promi-
et  al. 2020) that with the generation of VMI, a nence compared to background HU) of the hema-
very high diagnostic accuracy regarding early toma at blood–brain interface, improved spatial
ischemic changes in the cerebral parenchyma resolution, and decreased partial volume averag-
could be achieved, with the best result at ing rendered by thin-section images.
70 keV. Karino et al. demonstrated the value of Bodanapally et  al. showed in another paper
VMI images for the investigation of brain metas- from 2019 (Bodanapally et al. 2019b) that a pseu-
tasis by DECT.  They described an improved dohematoma after traumatic hemorrhagic contu-
tumor visibility with optimal VMI (63  keV) sion (by means of quantifying iodine leakage in
which can supplement an accurate delineation of DECT head scans) is a reliable marker of post-
brain metastases (Karino et  al. 2020). Lastly, traumatic blood–brain barrier permeability, which
VMI (particularly at low keV levels) can also in turn is predictive of subsequent neurosurgical
lead to superior quantitative image quality in management. Pseudohematoma was defined as an
carotid and intracerebral angiography (Leithner enhancing penumbra caused by iodine leakage on
et al. 2018). follow-up 120 kV images. In another paper by the
same authors, it could be shown that by analyzing
conventional 120 kV images, the size of hemor-
3.5  ther Applications of DECT
O rhagic cerebral contusions is routinely overesti-
in Emergency Neuroradiology mated, which in turn can be corrected by using a
190  keV image set (Bodanapally et  al. 2018).
As of recently, DECT has also found its way into Furthermore, Can et  al. showed that in DECT
the acute setting of emergency radiology. It has scans by means of properly identifying quantifi-
been shown that there are several applications cation of iodine extravasation, hematoma expan-
regarding the evaluation of intracranial hemor- sion in acute cerebral hemorrhage could be
rhages. By using monoenergetic reconstructions reliably predicted (Tan et al. 2019).
from DECT, the reliability to detect traumatic If DECT is appropriately included in the clini-
intracranial hemorrhages can be increased com- cal workflow, it might not only add benefit by
pared to standard 120 kV Images. Subdural and increasing the radiologist image interpretation
epidural hematomas in particular as well as con- confidence but also might reduce the need for
tusions can be better visualized in high keV follow-up imaging as shown in a larger study
images (190 keV) as shown by Bodanapally et al. with 3159 emergency CT scans from all body
(Bodanapally et al. 2019a). This can be attributed regions (Wong et al. 2020).

a b c d

Fig. 7 Dual-energy CT for metal artifact reduction. monoenergetic images (VMI) with higher or lower keV
Severe hypodense streak artifacts are seen in a patient levels. b–d demonstrate higher keV levels (90–130 keV)
after surgical aneurysm clipping in the standard 70  keV with markedly decreasing artifacts (arrows)
image (a). DECT allows to reconstruct various virtual
122 S. Winklhofer et al.

4 Photon Counting 5 Outlook and Conclusion

Photon-Counting Detector-Based CT (PCD-CT) With the increasing availability and implementa-

is currently the subject of intense investigations tion of DECT systems, its future impact in clini-
and is a promising and emerging new develop- cal neuroimaging will further increase. New
ment in CT which is based on a high resolution, DECT techniques might be available and already
multi-energy imaging technique. In contrast to established systems will be improved.
DECT where the conventional detector converts One example is the recent revival of the dual-­
x-rays to light and consequently to an electrical spiral technique by one of the major vendors
signal, photon-counting detectors do not convert which is actually the earliest and simplest way to
the x-ray photons to light, but directly to electri- obtain DECT data, in which the patient is exam-
cal pulses. The energy of the x-ray of each pho- ined using two CT spirals in quick succession.
ton is determined by the magnitude of the What initially started as the beginning of DECT
electrical pulse. This results in the ability to also imaging could now be optimized so that the spec-
register spectral information with multiple tral examination can be performed in a dose-­
(more than two) energies (Rajendran and neutral fashion with high quality on single source
McCollough 2020; Willemink et  al. 2018). DECT scanners.
Details about the technical background of New DECT or multi-energy contrast agents
PCD-CT can be found in the chapters “Basic are currently under investigation as specific dis-
Principles and Clinical Applications of Photon- ease biomarkers or for local disease determina-
Counting CT” and “Photon-Counting CT: Initial tion (Gutjahr et  al. 2021). Biocompatible
Clinical Experience” of this book. Several pro- high-atomic number contrast materials with a
totypes of PCD-CT are currently under investi- different biodistribution and a different x-ray
gation and the first commercial systems for attenuation than iodine might expand the diag-
clinical applications available. The major nostic power of DECT or multi-energy CT with-
advantages are the increased spatial resolution out an increase of the scan time or the radiation
and the potentially lower radiation dose com- dose (Yeh et al. 2017).
pared to conventional single-energy or current Also, AI might play a role for DECT by means
DECT systems (McCollough et al. 2015; Rajiah of data acquisition, data postprocessing, and data
et  al. 2020; Zhou et  al. 2019). This ultra-high reconstruction. One example is the application of
resolution capability enables a substantially bet- Deep Learning to improve the image quality
ter delineation of fine anatomy, e.g., for the tem- while reducing the required radiation dose at the
poral bones or the sinuses compared to standard same time in pediatric patients (Lee et al. 2020).
clinical CT imaging with even a potentially Another truly interesting aspect has been recently
lower radiation dose (Rajendran et  al. 2020; investigated: DECT information could be gener-
Zhou et al. 2018). A higher soft-tissue contrast ated from SECT images by using Deep Learning
and lower image noise for photon-counting CT approaches. This could allow to gain the benefits
allows for an improved gray-white matter con- of DECT such as monoenergetic reconstructions
trast compared with conventional CT as shown or material differentiation from a non-DECT
by Pourmorteza et al. (Pourmorteza et al. 2017). single-­
energy acquisition (Liu et  al. 2021;
Other potential applications of PCD-CT include Kawahara et  al. 2021; Cong et  al. 2020). The
advanced metal artifact reduction (Do et  al. results of the preliminary research indicate great
2020; Sigovan et al. 2019), the differentiation of potential; however, the approach has to be clini-
blood and iodine (Riederer et  al. 2019), and cally validated in the future, but might heavily
improved vascular imaging (Harvey et al. 2019) influence future CT imaging with several poten-
(Mannil et al. 2018). tial applications.
Neuroradiological Imaging 123

With the upcoming implementation of clinical Bodanapally UK, Shanmuganathan K, Gunjan YP,
Schwartzbauer G, Kondaveti R, Feiter TR (2019b)
photon-counting scanners, multi-energy CT Quantification of iodine leakage on dual-energy CT as
imaging will reach a new era and as a result, these a marker of blood-brain barrier permeability in trau-
technical advantages may be translated into matic hemorrhagic contusions: prediction of surgical
advantages for the patients. intervention for intracranial pressure management.
AJNR Am J Neuroradiol 40(12):2059–2065
Spectral imaging has developed into a power- Bonatti M, Lombardo F, Zamboni GA, Pernter P, Pozzi
ful tool in neuroimaging and potential applica- Mucelli R, Bonatti G (2017) Dual-energy CT of the
tions have been increasing over the past decade brain: comparison between DECT angiography-­
with many other highly interesting and promising derived virtual unenhanced images and true unen-
hanced images in the detection of intracranial
projects currently under development. A growing haemorrhage. Eur Radiol 27(7):2690–2697
number of studies are available which demon- Chen Z, Zhang Y, Su Y, Sun Y, He Y, Chen H (2020)
strate that physicians and subsequently patients Contrast extravasation is predictive of poor clinical
substantially benefit from selected technical outcomes in patients undergoing endovascular therapy
for acute ischemic stroke in the anterior circulation. J
innovations implemented in clinical routine. A Stroke Cerebrovascular Dis Off J Natl Stroke Assoc
key driver of this adoption will be an optimiza- 29(1):104494
tion and standardization of workflows in the Choi Y, Shin N-Y, Jang J, Ahn K-J, Kim B-S (2020)
future. To increase the utilization of spectral neu- Dual-energy CT for differentiating acute intracranial
hemorrhage from contrast staining or calcification: a
roimaging, vendor-neutral approaches must be meta-analysis. Neuroradiology 62(12):1617–1626
adopted, and the ability to seamlessly view and Cong W, Xi Y, Fitzgerald P, De Man B, Wang G (2020)
process DECT images within PACS is of utmost Virtual monoenergetic CT imaging via deep learning.
importance. Patterns (N Y) 1(8):100128
Do TD, Sawall S, Heinze S, Reiner T, Ziener CH, Stiller
Nevertheless, the implementation of these W et  al (2020) A semi-automated quantitative com-
advantages into the daily operation of a radiology parison of metal artifact reduction in photon-counting
institute requires some effort which emphasizes computed tomography by energy-selective threshold-
the need for further technical improvements and ing. Sci Rep 10(1):21099
Dodig D, Kovačić S, Matana Kaštelan Z, Žuža I, Benić
an optimized workflow in each individual radiol- F, Slaven J et al (2020) Comparing image quality of
ogy department. single- and dual-energy computed tomography of the
brain. Neuroradiol J 33(3):259–266
Compliance with Ethical Standards Grams AE, Djurdjevic T, Rehwald R, Schiestl T, Dazinger
F, Steiger R et  al (2018) Improved visualisation of
early cerebral infarctions after endovascular stroke
Disclosure of Interests  No financial disclosure. No ethi- therapy using dual-energy computed tomography
cal conflict. oedema maps. Eur Radiol 28(11):4534–4541
Gutjahr R, Bakker RC, Tiessens F, van Nimwegen SA,
Schmidt B, Nijsen JFW (2021) Quantitative dual-­
energy CT material decomposition of holmium micro-
References spheres: local concentration determination evaluated
in phantoms and a rabbit tumor model. Eur Radiol
Bodanapally UK, Shanmuganathan K, Issa G, Dreizin 31(1):139–148
D, Li G, Sudini K et  al (2018) Dual-energy CT in Harvey EC, Feng M, Ji X, Zhang R, Li Y, Chen GH et al
hemorrhagic progression of cerebral contusion: over- (2019) Impacts of photon counting CT to maximum
estimation of hematoma volumes on standard 120- intensity projection (MIP) images of cerebral CT angi-
kV images and rectification with virtual high-energy ography: theoretical and experimental studies. Phys
monochromatic images after contrast-enhanced Med Biol 64(18):185015
whole-body imaging. AJNR Am J Neuroradiol Hixson HR, Leiva-Salinas C, Sumer S, Patrie J, Xin W,
39(4):658–662 Wintermark M (2016) Utilizing dual energy CT to
Bodanapally UK, Archer-Arroyo KL, Dreizin D, improve CT diagnosis of posterior fossa ischemia. J
Shanmuganathan K, Schwartzbauer G, Li G et  al Neuroradiol 43(5):346–352
(2019a) Dual-energy computed tomography imaging Hopf-Jensen S, Anraths M, Lehrke S, Szymczak S, Hasler
of head: virtual high-energy monochromatic (190 keV) M, Müller-Hülsbeck S (2020) Early prediction of final
images are more reliable than standard 120 kV images infarct volume with material decomposition images
for detecting traumatic intracranial hemorrhages. J of dual-energy CT after mechanical thrombectomy.
Neurotrauma 36(8):1375–1381 Neuroradiology
124 S. Winklhofer et al.

Kamiya K, Kunimatsu A, Mori H, Sato J, Akahane M, Mohammed MF, Marais O, Min A, Ferguson D, Jalal S,
Yamakawa T et  al (2013) Preliminary report on vir- Khosa F et  al (2018) Unenhanced dual-energy com-
tual monochromatic spectral imaging with fast kVp puted tomography: visualization of brain edema.
switching dual energy head CT: comparable image Investig Radiol 53(2):63–69
quality to that of 120-kVp CT without increasing the Neuhaus V, Abdullayev N, Große Hokamp N, Pahn G,
radiation dose. Jpn J Radiol 31(4):293–298 Kabbasch C, Mpotsaris A et  al (2017) Improvement
Karino T, Ohira S, Kanayama N, Wada K, Ikawa T, Nitta of image quality in unenhanced dual-layer CT of the
Y et al (2020) Determination of optimal virtual mono- head using virtual Monoenergetic images compared
chromatic energy level for target delineation of brain with Polyenergetic single-energy CT. Investig Radiol
metastases in radiosurgery using dual-energy CT. Br J 52(8):470–476
Radiol 93(1106):20180850 Pomerantz SR, Kamalian S, Zhang D, Gupta R, Rapalino
Kawahara D, Saito A, Ozawa S, Nagata Y (2021) Image O, Sahani DV et  al (2013a) Virtual monochromatic
synthesis with deep convolutional generative adver- reconstruction of dual-energy unenhanced head CT
sarial networks for material decomposition in dual-­ at 65-75 keV maximizes image quality compared
energy CT from a kilovoltage CT. Comput Biol Med with conventional polychromatic CT.  Radiology
128:104111 266(1):318–325
Korn A, Bender B, Brodoefel H, Hauser TK, Danz S, Pomerantz SR, Kamalian S, Zhang D, Gupta R, Rapalino
Ernemann U et  al (2015) Grading of carotid artery O, Sahani DV et  al (2013b) Virtual monochromatic
stenosis in the presence of extensive calcifications: reconstruction of dual-energy unenhanced head CT
dual-energy CT angiography in comparison with at 65-75 keV maximizes image quality compared
contrast-enhanced MR angiography. Clin Neuroradiol with conventional polychromatic CT.  Radiology
25(1):33–40 266(1):318–325
Lee S, Choi YH, Cho YJ, Lee SB, Cheon JE, Kim WS et al Pourmorteza A, Symons R, Reich DS, Bagheri M, Cork
(2020) Noise reduction approach in pediatric abdomi- TE, Kappler S et al (2017) Photon-counting CT of the
nal CT combining deep learning and dual-energy tech- Brain: in vivo human results and image-quality assess-
nique. Eur Radiol ment. AJNR Am J Neuroradiol 38(12):2257–2263
Leithner D, Mahmoudi S, Wichmann JL, Martin SS, Rajendran K, McCollough CH (2020) Photon-counting
Lenga L, Albrecht MH et  al (2018) Evaluation of detector-based computed tomography. In: Mannil M,
virtual monoenergetic imaging algorithms for dual-­ Winklhofer SFX (eds) Neuroimaging techniques in
energy carotid and intracerebral CT angiography: clinical practice: physical concepts and clinical appli-
effects on image quality, artefacts and diagnostic per- cations. Springer, Cham, pp 87–94
formance for the detection of stenosis. Eur J Radiol Rajendran K, Voss BA, Zhou W, Tao S, DeLone DR, Lane
99:111–117 JI et al (2020) Dose reduction for sinus and temporal
Lennartz S, Laukamp KR, Neuhaus V, Große Hokamp N, bone imaging using photon-counting detector CT with
Le Blanc M, Maus V et al (2018) Dual-layer detector an additional tin filter. Investig Radiol 55(2):91–100
CT of the head: initial experience in visualization of Rajiah P, Parakh A, Kay F, Baruah D, Kambadakone
intracranial hemorrhage and hypodense brain lesions AR, Leng S (2020) Update on multienergy CT:
using virtual monoenergetic images. Eur J Radiol physics, principles, and applications. Radiographics
108:177–183 40(5):1284–1308
Liu CK, Liu CC, Yang CH, Huang HM (2021) Generation Riederer I, Si-Mohamed S, Ehn S, Bar-Ness D, Noël
of brain dual-energy CT from single-energy CT using PB, Fingerle AA et al (2019) Differentiation between
deep learning. J Digit Imaging blood and iodine in a bovine brain-initial experience
Mannil M, Ramachandran J, Vittoria de Martini I, with spectral photon-counting computed tomography
Wegener S, Schmidt B, Flohr T et al (2017) Modified (SPCCT). PLoS One 14(2):e0212679
dual-energy algorithm for calcified plaque removal: Sigovan M, Si-Mohamed S, Bar-Ness D, Mitchell J,
evaluation in carotid computed tomography angiogra- Langlois JB, Coulon P et  al (2019) Feasibility of
phy and comparison with digital subtraction angiogra- improving vascular imaging in the presence of metal-
phy. Investig Radiol 52(11):680–685 lic stents using spectral photon counting CT and
Mannil M, Hickethier T, von Spiczak J, Baer M, Henning K-edge imaging. Sci Rep 9(1):19850
A, Hertel M et  al (2018) Photon-counting CT: high-­ Ståhl F, Gontu V, Almqvist H, Mazya MV, Falk DA
resolution imaging of coronary stents. Investig Radiol (2020) Performance of dual layer dual energy CT
53(3):143–149 virtual monoenergetic images to identify early isch-
McCollough CH, Leng S, Yu L, Fletcher JG (2015) emic changes in patients with anterior circulation
Dual- and multi-energy CT: principles, technical large vessel occlusion. J Neuroradiol = Journal De
approaches, and clinical applications. Radiology Neuroradiologie
276(3):637–653 Taguchi K, Itoh T, Fuld MK, Fournie E, Lee O, Noguchi
Mocanu I, Van Wettere M, Absil J, Bruneau M, Lubicz K (2018) “X-Map 2.0” for edema signal enhancement
B, Sadeghi N (2018) Value of dual-energy CT angiog- for acute ischemic stroke using non-contrast-enhanced
raphy in patients with treated intracranial aneurysms. dual-energy computed tomography. Investig Radiol
Neuroradiology 60(12):1287–1295 53(7):432–439
Neuroradiological Imaging 125

Tan CO, Lam S, Kuppens D, Bergmans RHJ, Wong WD, Mohammed MF, Nicolaou S, Schmiedeskamp
Parameswaran BK, Forghani R et al (2019) Spot and H, Khosa F, Murray N et  al (2020) Impact of dual-­
diffuse signs: quantitative markers of intracranial energy CT in the emergency department: increased
hematoma expansion at dual-energy CT.  Radiology radiologist confidence, reduced need for follow-
290(1):179–186 ­up imaging, and projected cost benefit. AJR Am J
Tran DN, Straka M, Roos JE, Napel S, Fleischmann D Roentgenol 215(6):1528–1538
(2009) Dual-energy CT discrimination of iodine Yeh BM, FitzGerald PF, Edic PM, Lambert JW, Colborn
and calcium: experimental results and implications RE, Marino ME et al (2017) Opportunities for new CT
for lower extremity CT angiography. Acad Radiol contrast agents to maximize the diagnostic potential of
16(2):160–171 emerging spectral CT technologies. Adv Drug Deliv
van Ommen F, de Jong HWAM, Dankbaar JW, Bennink Rev 113:201–222
E, Leiner T, Schilham AMR (2019) Dose of CT pro- Zaouak Y, Sadeghi N, Sarbu N, Ligot N, Lubicz B (2020)
tocols acquired in clinical routine using a dual-layer Differentiation between cerebral hemorrhage and
detector CT scanner: a preliminary report. Eur J contrast extravasation using dual energy computed
Radiol 112:65–71 tomography after intra-arterial neuro interventional
van Ommen F, Dankbaar JW, Zhu G, Wolman DN, Heit procedures. J Belgian Soc Radiol 104(1):70
JJ, Kauw F et al (2021) Virtual monochromatic dual-­ Zhang LJ, Wu SY, Niu JB, Zhang ZL, Wang HZ, Zhao
energy CT reconstructions improve detection of YE et  al (2010) Dual-energy CT angiography in the
cerebral infarct in patients with suspicion of stroke. evaluation of intracranial aneurysms: image quality,
Neuroradiology 63(1):41–49 radiation dose, and comparison with 3D rotational
Weinman JP, Mirsky DM, Jensen AM, Stence NV (2019) digital subtraction angiography. AJR Am J Roentgenol
Dual energy head CT to maintain image quality while 194(1):23–30
reducing dose in pediatric patients. Clin Imaging Zhao XM, Wang M, Wu RZ, Dharaiya E, Feng F, Li ML
55:83–88 et al (2018) Dual-layer spectral detector CT monoen-
Wiggins WF, Potter CA, Sodickson AD (2020) Dual-­ ergetic reconstruction improves image quality of non-­
energy CT to differentiate small foci of intracranial contrast cerebral CT as compared with conventional
hemorrhage from calcium. Radiology 294(1):129–138 single energy CT. Eur J Radiol 103:131–138
Willemink MJ, Persson M, Pourmorteza A, Pelc NJ, Zhou W, Lane JI, Carlson ML, Bruesewitz MR, Witte
Fleischmann D (2018) Photon-counting CT: tech- RJ, Koeller KK et al (2018) Comparison of a photon-­
nical principles and clinical prospects. Radiology counting-­ detector CT with an energy-integrating-­
289(2):293–312 detector CT for temporal bone imaging: a cadaveric
Winklhofer S, Hinzpeter R, Stocker D, Baltsavias G, study. AJNR Am J Neuroradiol 39(9):1733–1738
Michels L, Burkhardt J-K et  al (2018) Combining Zhou W, Bartlett DJ, Diehn FE, Glazebrook KN, Kotsenas
monoenergetic extrapolations from dual-energy CT AL, Carter RE et  al (2019) Reduction of metal arti-
with iterative reconstructions: reduction of coil and facts and improvement in dose efficiency using
clip artifacts from intracranial aneurysm therapy. photon-­counting detector computed tomography and
Neuroradiology 60(3):281–291 tin filtration. Investig Radiol 54(4):204–211
 Head and Neck Imaging

David Zopfs

Contents
1  ead and Neck Oncology
H  128
1.1  Delineation and Visibility of Tumors  128
1.2  Lymph Node Imaging  131
2 Salivary Glands  131
3 Inflammation  132
4 Metal Artifact Reduction  132
5 Thyroid and Parathyroid Lesions  135
6 Angiography of the Head and Neck  135
7  uture Directions: Assessment of Therapy Response and 
F
Prediction of Recurrence  137
References  139

Abstract
nodes. High keV VMI are an established tool
in the metal artifact reduction arsenal and are
CT is an important imaging method in the head beneficial in the assessment of tumorous carti-
and neck region and accumulating evidence lage invasion. Material-specific DECT recon-
indicates an added value of dual-energy CT structions, such as quantitative iodine maps
(DECT) reconstructions in this area, especially might allow for a better tissue characterization
for head and neck cancer. In general, DECT or distinction of benign and metastatic lymph
derived low keV virtual monoenergetic images nodes. While VMI and iodine maps are both
(VMI) facilitate an increased contrast and sub- relatively well researched and already in fre-
sequently an improved delineation of tumor quent clinical use, data on radiomics is com-
tissue and its boundaries, vessels, and lymph paratively sparse and principal application
across different platforms is not yet given.
Overall, a multiparametric approach of
D. Zopfs (*) various DECT reconstructions seems most
Institute for Diagnostic and Interventional Radiology, promising to develop a clinically relevant
University Hospital Cologne, Cologne, Germany advantage over conventional CT, allowing for
e-mail: [email protected]

© The Author(s), under exclusive license to Springer Nature Switzerland AG 2022 127
H. Alkadhi et al. (eds.), Spectral Imaging, Medical Radiology Diagnostic Imaging,
https://doi.org/10.1007/978-3-030-96285-2_9
128 D. Zopfs

a more accurate imaging of head and neck CT examinations. Numerous studies demon-
pathologies to achieve the best possible patient strated the ability of DECT derived low keV vir-
treatment. The following chapter will provide tual monoenergetic images (VMI) to boost
an overview of the most significant applica- contrast in vascularized tumors, resulting in an
tions of multiparametric DECT in head and increased signal-to-noise (SNR) and contrast-to-­
neck imaging and outline challenges and noise ratio (CNR) of tumorous tissue compared
obstacles to further deployment, as well as to conventional image reconstructions (Albrecht
future developments. et al. 2015; Forghani 2019; Forghani et al. 2017;
Lohöfer et al. 2018; Roele et al. 2017; Wichmann
et al. 2014). While in general 40 keV VMI tend to
yield highest SNR and CNR in head and neck
1 Head and Neck Oncology cancer, image noise might simultaneously
increase at 40  keV, depending on the applied
Head and neck cancer causes up to 300.000 can- denoising algorithms and technical approach to
cer deaths each year, representing the 7th most DECT (Forghani 2019; Große Hokamp et  al.
common cancer in men worldwide. Squamous 2020). Beyond improved quantitative improve-
cell carcinomas account for the largest propor- ment, low keV VMI similarly enhance subjective
tion of head and neck cancer (Bray et al. 2018; image analysis. For example, May et al. reported
Forghani 2019). Precise imaging of the head and superior image quality in 40  keV VMI in com-
neck region is essential for initial, pretreatment parison to conventional images and other VMI
workup as well as follow-up examinations to levels for a dual-source DECT system (May et al.
allow for adequate treatment decisions. In the ini- 2019). Albrecht et al. recommended 55 keV for
tial assessment of newly diagnosed head and clinical practice in patients with head and neck
neck cancer, one of the crucial points of radio- cancer (Albrecht et al. 2015). The improvement
logical imaging is to determine the extent of in detection and delineation of head and neck
tumor growth and to assess metastatic spread, cancer by low-energy VMI in staging CT exami-
which both may not be detected in clinical and nations has been widely described in many stud-
endoscopic evaluation, thus resulting in a poten- ies. Toepker et  al. highlighted the capability of
tial upstaging (Forghani 2019). Compared to low keV VMI to improve the depiction of tumor
MRI, CT is more widely available; however, due margins, which might be especially beneficial in
to its limited soft tissue contrast, precise tumor initial assessment of T-stage or for surgical resec-
delineation in CT of the head and neck reason tion planning (Toepker et  al. 2014). In general,
may be challenging. Moreover, the head and neck VMI at 40–60 keV are to be favored when assess-
region has a rather difficult anatomy and the ing head and neck cancer with individual adjust-
tumors are often close to many important ana- ment depending on the case. While many
tomical structures - there is no other region where retrospective studies have demonstrated the ben-
the human body is as narrow in its cross-section. efits of various DECT reconstructions, larger-­
Different DECT derived reconstructions seem a scale prospective, preferably cross vendor studies
promising tool to mitigate this issue and facilitate exploring the impact on patient treatment and
a more accurate staging. disease outcome are still lacking. Figure 1 illus-
trates the increase in contrast of hyper vascular-
ized tumor areas in low keV VMI and an enhanced
1.1 Delineation and Visibility delineation in iodine overlay maps.
of Tumors
1.1.1 Invasion of Cartilage
Contrast enhancement and tissue asymmetry are Adequate assessment of potential cartilage inva-
among the underlying mechanisms integral to sion is a cornerstone in the workup of patients
head and neck cancer identification in cervical with hypopharyngeal or laryngeal cancers, in
Head and Neck Imaging 129

a b

c d

Fig. 1  Initial imaging study of a 73-year-old male patient 60 keV (c), 50 keV (d), down to 40 keV (e). Improved
with histopathological proven squamous cell carcinoma in delineation of tumor boundaries (white arrows) can be
conventional image reconstruction (a) the contrast appreciated in low keV VMI and iodine overlay maps
between the lesions and liver parenchyma is gradually compared to CI.  Iodine overlay maps depict areas of
increased with decreasing keV levels from 70 keV (b) to necrosis and hypervascularization (white asterisk)
130 D. Zopfs

particular detection of thyroid cartilage invasion DECT derived iodine overlay maps and weighted-­
(Kuno et  al. 2014; Pérez-Lara and Forghani average images (Kuno et al. 2012). Iodine over-
2018). The reason for that being that the pres- lay maps facilitate the depiction of iodine
ence/absence of cartilage invasion has a direct distribution in contrast enhances scans and
impact on treatment strategies: If cartilage inva- thereby are likely to diminish the rate of false
sion is present, more aggressive surgical treat- positive diagnoses for tumor infiltration in hya-
ment is needed while larynx-preserving strategies line cartilage (Kuno et  al. 2012; Zopfs et  al.
may be advantageous in patients without carti- 2021). Kuno et  al. demonstrated in a follow-up
lage invasion (Kuno et  al. 2012, 2014; Roele study a higher specificity of iodine overlay maps
et al. 2017; Sheahan 2014). Detection of cartilage and weighted-average images compared to MRI
invasion with CT is challenging, as non-­ossified with similar sensitivity (Kuno et  al. 2018).
cartilage and tumor show a considerable overlap Another promising approach to differentiate hya-
in HU values (Kuno et  al. 2012; Roele et  al. line cartilage and tumor tissue was demonstrated
2017), rendering accurate delineation rather dif- by Forghani et al.: High keV of ≥95 keV enabled
ficult. Furthermore, hyaline cartilage ossifies a reliable, attenuation-based method for differen-
with age, which causes healthy laryngeal carti- tiation (see Fig. 2). This is due to the relatively
lage to differ in appearance from patient to high intrinsic attenuation of hyaline cartilage,
patient, further complicating accurate differentia- which is not decreased in high keV contrary to
tion between healthy cartilage and adjacent the attenuation of vascularized tumor tissue, in
tumor. In this respect, Kuno et  al. reported that which enhancement largely depends on iodinated
the combination of DECT derived iodine overlay contrast media (Forghani et al. 2015; Roele et al.
maps and weighted-average images (which are 2017).
similar to conventional CT images) significantly Given the results of these studies, implemen-
improved specificity of detection of cartilage tation of low keV and high keV VMI as well as
invasion compared to weighted-average images iodine maps in clinical routine represents a clear
only, while sensitivity did not decrease (Kuno benefit over conventional single energy CT
et  al. 2012). Additionally, inter-observer agree- images in the staging of laryngeal and hypopha-
ment increased when using the combination of ryngeal cancer.

a b

Fig. 2  Virtual monoenergetic image (VMI) reconstruc- bordering cartilage (white arrow). At 150 keV VMI (b),
tion at 50 keV (a) display a hyperattenuating tumor (black the attenuation of the tumor is reduced, while the attenua-
arrows) adjacent to the posterior part of the left thyroid tion of the cartilage remains high and the delineation of
cartilage with considerable overlap of attenuation to the the cartilage boundary is improved
Head and Neck Imaging 131

1.2 Lymph Node Imaging healthy and metastatic lymph nodes in patients
with papillary thyroid cancer (Liu et  al. 2015).
Presence/Absence of lymph node metastases is However, while Tawfik et al. reported significant
an important prognostic factor in head and neck lower iodine concentrations in metastatic lymph
cancer (Axelsson et  al. 2017; Magnano et  al. nodes, Liu et al. contrary found an increased nor-
1997). Thus, accurate imaging of metastatic cer- malized iodine concentration associated with
vical lymph nodes is crucial to detect clinically malignant lymph nodes (Liu et al. 2015; Tawfik
not palpable lymph nodes, to evaluate the extent et al. 2014). In this context it is important to con-
of lymph node metastases or to confirm a N0 sta- sider factors influencing iodine concentration,
tus in follow-up examinations. However, in daily such as intra- and inter-individual as well as
radiological routine reliable delineation and inter-scanner differences (Lennartz et  al. 2021;
detection of cervical lymph nodes may be diffi- Zopfs et al. 2020a, b). The current body of evi-
cult due to the proximity of anatomical structures dence is not yet sufficient to distinguish benign
in the neck region. Unfortunately, enlargement of from metastatic lymph nodes based on absolute
lymph nodes is neither sensitive nor specific iodine concentrations with high confidence
enough to facilitate an accurate classification of (Fig. 3).
lymph nodes as malignant, as up to 20% of lymph
nodes are reactively increased in size or are
hyperplastic (Tawfik et  al. 2014). Therefore, 2 Salivary Glands
imaging biomarkers allowing for a more precise
assessment of ambiguous lymph nodes are highly Sialolithiasis is the most frequent disease of the
desirable. In this regard, Tawfik et  al. demon- major salivary glands and CT can be helpful in
strated that DECT derived iodine concentration visualizing complications or facilitate treatment
significantly varies between healthy, inflamma- planning, especially if conventional intraoral
tory, and metastatic lymph nodes (Tawfik et  al. radiographs are not well assessable. While unen-
2014). Similarly, Liu et al. reported differences in hanced CT images are preferred to depict sialoli-
normalized iodine concentration and the slope of thiasis, contrast-enhanced CT enables a better
the spectral Hounsfield unit curve between assessment of glandular parenchyma and possible

a b

Fig. 3  Improved delineation of metastatic lymph node images at 40  keV (a) compared to conventional image
(right white arrow) and healthy appearing lymph nodes reconstructions (b)
(left white arrow) in low keV virtual monoenergetic
132 D. Zopfs

inflammatory complications. However, focal 4 Metal Artifact Reduction

spots of hyperenhancement or cross-sections of
vessels may mimic lithiasis, thus leading to false In daily routine, image quality of head and neck
positive findings (Pulickal et al. 2019; Rzymska-­ CT examinations is often severely hampered by
Grala et al. 2010). Regarding this matter, virtual artifacts arising from different forms of metallic
non-contrast (VNC) images reconstructed from hardware, especially dentures and dental
contrast-enhanced DECT have been reported to implants. This limitation is aggravated with
be instrumental for delineation of sialolithiasis demographic change, as more and more older
(Beland et al. 2019; Pulickal et al. 2019). Beland patients are undergoing CT examinations and the
et al. reported comparable accuracy in the detec- frequent presence of metallic dental hardware in
tion of sialolithiasis between true non-contrast these patients. Thus, the oral cavity and the max-
images and VNC images derived from rapid kVp illofacial region are particularly affected by arti-
switching DECT (Beland et al. 2019). Therefore, facts. Metal artifacts mainly arise for three
a single phase, contrast-enhanced DECT exami- reasons: (1) beam hardening artifacts, result from
nation protocol with reconstruction of VNC the absorption of low-energy photons of the poly-
images may allow omitting true non-contrast energetic X-ray beam, (2) photon starvation,
images which could lead to significant radiation which are caused by the complete absorption of
dose reduction (Beland et al. 2019; Pulickal et al. photons, and (3) scatter artifacts, which occur
2019). However, validation in studies with larger from major differences in attenuation between
patient collectives and different technical highly attenuating metallic implants and the adja-
approaches to DECT is needed prior to routine cent soft tissue. These phenomena result in dif-
clinical application. ferent kind of artifacts, i.e. hypo- and hyperdense
areas adjacent to the metallic structure, streaking
artifacts that can spread over the whole CT image
3 Inflammation and an increase in image noise (Große Hokamp
et al. 2018; Mori et al. 2013; Roele et al. 2017;
Peritonsillar abscesses represent the most fre- Zopfs et  al. 2020c). Especially the combination
quent deep infection in the head and neck region of these effects can result in strong interferences,
in adolescent patients. While clinical examination which may severely impair diagnostic assess-
is most important for diagnosis, CT is the main ment. In this regard, DECT has been found an
imaging modality chosen to visualize the extent efficient resource to reduce metal artifacts and
and localization of the abscess (Roele et al. 2017; improve subjective image assessment. Especially
Steyer 2002). Scholtz et  al. demonstrated an virtual monoenergetic images at high energy lev-
increased soft tissue enhancement and improved els have proven to be an powerful tool for artifact
delineation of inflammatory changes in patients reduction in various studies, irrespective of the
with peritonsillar abscesses using low-­ tube-­ technical approach to dual-energy CT (Bamberg
voltage 80 kVp dual-source DECT (Scholtz et al. et al. 2011; Cha et al. 2017; Große Hokamp et al.
2015). Furthermore, DECT derived iodine over- 2018; Lee et  al. 2012; Morsbach et  al. 2013;
lay maps might be helpful in depicting inhomoge- Stolzmann et  al. 2013; Tanaka et  al. 2013)
neous iodine distribution of salivary gland tissue (Fig. 5). For example, Stolzmann et al. and Große
in sialadenitis (Chawla et  al. 2017). Due to the Hokamp et  al. found a significant reduction of
very limited available studies regarding this topic metal artifacts from dental implants in VMI with
and the lacking literature evidence, the clinical higher keV while simultaneous improving diag-
application of DECT in this field remains explor- nostic assessment of the oral cavity (Große
atory. However, an increased contrast of the hyper Hokamp et al. 2018; Stolzmann et al. 2013). The
vascularized abscess rim and an improved delin- optimal keV level for artifact reduction was
eation of boundaries in low keV VMI and iodine stated at 108 ± 17 keV and 145 ± 15 keV, respec-
overlay maps seem suggestive (see Fig. 4). tively (Große Hokamp et  al. 2018; Stolzmann
Head and Neck Imaging 133

a b

Fig. 4  Nuchal abscess (white arrow) in conventional strong contrast enhancement of the hyper vascularized
image reconstructions (a), virtual monoenergetic images abscess rim is highlighted in 40 keV VMI and iodine over-
(VMI) at 40  keV (b) and iodine overlay maps (c). The lay maps

et al. 2013). Zhou et al. and Guggenberger et al. adjustment is necessary depending on the spe-
investigated the usefulness of high keV VMI to cific patient, type of implant, and artifact local-
reduce metal artifacts arising from spinal ization (Bamberg et  al. 2011; Große Hokamp
implants (Guggenberger et al. 2012; Roele et al. et al. 2018; Guggenberger et al. 2012; Laukamp
2017; Zhou et al. 2011). Similar to artifacts from et  al. 2019; Roele et  al. 2017; Stolzmann et  al.
dental artifacts, both studies found VMI at around 2013; Tanaka et  al. 2013; Zhou et  al. 2011).
130 keV best suited for improving image quality Recent studies suggested that a combination of
(Guggenberger et al. 2012; Zhou et al. 2011). dedicated metal artifact reduction algorithms
Altogether, the optimal VMI level for artifact (MAR) and virtual monoenergetic images may
reduction in the head and neck region seems to yield an additional benefit compared to the sole
range between 100 and 150  keV, yet individual use of one of the techniques (Große Hokamp
134 D. Zopfs

et  al. 2020; Guggenberger et  al. 2012; Tanaka 140 and 200  keV and MAR (Laukamp et  al.
et  al. 2013). Laukamp et  al. demonstrated that 2019). However, this combined approach holds
especially hyperdense artifacts can be effectively the risk of overcorrection and new artifacts
reduced using a combination of VMI between (Laukamp et al. 2019; Zopfs et al. 2020c).

a b

c d

Fig. 5  Contrast-enhanced scan of the oral cavity in are reduced with higher keV virtual monoenergetic
venous phase with significant artifacts arising from metal- images (70 keV [b], 100 keV [c], 130 keV [d], 170 keV
lic dental hardware. Hyper- and hypodense artifact streaks [e], and 200 keV [f]) compared to conventional images (a)
Head and Neck Imaging 135

e f

Fig. 5 (continued)

5 Thyroid and Parathyroid a multiphasic protocol (Forghani et  al. 2016;

Lesions Woisetschläger et al. 2020). In this regard, DECT
derived VNC reconstructions may reduce radiation
Thyroid nodules are a common finding in CT exposure. Leiva-­Salinas et al. reported a compara-
examinations of the head and neck, whereas inci- ble diagnostic accuracy of a monophasic DECT
dental thyroid cancer is rare. Their high prevalence protocol with reconstruction of VNC images com-
and the associated cost of patient workup renders a pared to a conventional biphasic protocol, resulting
reliable differentiation in CT highly desirable in a considerable dose reduction (Leiva-Salinas
(Youserm et al. 1997). VMI and iodine maps have et  al. 2016). Forghani et  al. found an additional
been investigated for distinguishing malignant value of different DECT reconstructions in the
from benign focal thyroid lesions in different stud- detection of parathyroid adenomas, such as iodine
ies (Gao et al. 2016; Lee et al. 2019; Li et al. 2012). concentration, SHUAC curves, and effective Z,
Yet, the body of evidence is relatively small. which might increase diagnostic accuracy, espe-
Parathyroid adenomas are the leading cause of cially in equivocal cases (Forghani et al. 2016).
primary hyperparathyroidism (Duan et  al. 2015).
As treatment is surgical, accurate pretherapeutic
assessment of size and location is important. 6 Angiography of the Head
Whereas ultrasonography of the neck and and Neck
Technetium-99m-Sestamibi scintigraphy of the
parathyroid comprise the most established imaging CT-Angiography of the head and neck is fre-
methods for characterization of parathyroid adeno- quently used to detect stenosed or occluded cer-
mas, various comparative studies described a supe- vical arteries, evaluate atherosclerotic plaques or
rior performance of multiphasic CT to evaluate vascular malformations (Anzidei et  al. 2012).
parathyroid adenomas. Currently, CT is especially Numerous studies demonstrated a boost of vessel
used if results of previous examinations are incon- contrast in CT-Angiographies of the head and
sistent, however, utilization of CT is limited due to neck in low keV VMI compared to conventional
the considerable radiation exposure resulting from images (see Fig. 6) (Leithner et al. 2018; Martin
136 D. Zopfs

a b c

d e

Fig. 6  Sagittal reconstructions of a CT-Angiography of images (VMI) at 70 keV (b), 60 keV (c), 50 keV (d), and
the head and neck with depiction of the common carotid 40  keV (e). Improved subjective image quality can be
artery and the proximal external and internal carotid artery appreciated in low keV VMI
in conventional images (a) and virtual monoenergetic
Head and Neck Imaging 137

et  al. 2017; Zopfs et  al. 2018). This increase in addressed on the capability of DECT to differen-
iodine attenuation facilitates an improvement of tiate between recurrent disease and benign post-
objective image quality, such as SNR and CNR treatment changes in patients with history of
accompanied with a subjective superior vessel treated head and neck cancer undergoing follow-
delineation and assessment (Riffel et  al. 2016; ­up examinations. In a study from Takumi et al.,
Zopfs et al. 2018). Most studies report ideal keV VMI at 40  keV and iodine concentration were
levels of 40–60  keV, varying with the technical both significantly higher in recurrent tumor com-
approach to DECT and examination protocols pared to post-therapeutic tissue (Takumi et  al.
(Albrecht et al. 2019; Leithner et al. 2018; Zopfs 2020). Similarly, several other studies found the
et al. 2018). Neuhaus et al. found low keV VMI spectral HU curve and iodine quantification help-
especially useful to assess arteries nearby the ful to distinguish and benign reactive changes
skull base and smaller arterial branches (Neuhaus after treatment (Roele et  al. 2017; Srinivasan
et  al. 2018). Furthermore, VMI at 40  keV were et  al. 2013; Yamauchi et  al. 2016). Thus, these
found to be useful for increasing vessel contrast DECT derived reconstructions might assist in the
in venous phase examinations of the neck (see diagnosis of recurrent head and neck cancer and
Fig. 7), reaching an image quality comparable to decrease the number of biopsies needed.
that of conventional angiographic CT scans Recently, an increasing number of studies
(Zopfs et al. 2018). This allows for an equivalent demonstrated that radiomics can be employed to
assessment of incidental carotid stenosis in generate an additional value from the wide range
venous phase staging examinations compared to of quantitative DECT data (Agarwal et al. 2020;
dedicated CT-Angiographies, thus possibly Forghani et al. 2019; Guo et al. 2020; Tomita et al.
decreasing the number of subsequent follow-up 2020). A study from Agarwal et al. indicated that
examinations to evaluate the extent of carotid the radiomics parameter medium texture entropy
artery stenosis. While few studies reported differ- is predictive of local control and laryngectomy
ences in the quantification of the degree of carotid free survival (Agarwal et al. 2020). Forghani et al.
artery stenosis in low keV, most studies found no demonstrated that multi-energy texture analysis,
significant differences compared to conventional i.e. on different VMI levels, allows for prediction
studies (Leithner et  al. 2018; Paul et  al. 2013; of cervical lymph node metastases (Forghani
Saba et  al. 2019; Zopfs et  al. 2018). However, et al. 2019). A retrospective study of Tomita et al.
calcified plaques of the carotid arteries and bone suggested that texture analysis of DECT derived
frequently hamper the assessment of cervical VMI might facilitate an differentiation of benign
arteries and the determination of stenosis grades and malignant thyroid nodules (Tomita et  al.
as blooming artifacts may lead to an overestima- 2020). Additionally, different radiomics features
tion of stenosis in conventional CT-Angiography. based on absolute iodine concentration correlated
In this regard, different studies demonstrated a significantly with tumor recurrence in a prelimi-
benefit of DECT based virtual calcium or bone nary study from Bahig et  al. (2019). However,
removal (Deng et  al. 2009; Kaemmerer et  al. inter-individual physiological and longitudinal
2016; Kamalian et al. 2017; Mannil et al. 2017). intra-individual changes of iodine concentrations
should be taken into account before providing
absolute cut-off values to distinguish malignant
7 Future Directions: and benign tissue or predict outcome, especially
Assessment of Therapy in studies with rather small patient collectives
Response and Prediction (Lennartz et al. 2021; Zopfs et al. 2020a, b).
of Recurrence In general, however, most of the studies deal-
ing with the significance of radiomics analysis
Whereas the vast majority of research has been are retrospective and limited in their sample size.
focused on the assessment of newly diagnosed Therefore, a basic practical clinical applicability
head and neck cancer so far, only a few studies is not given yet.
138 D. Zopfs

a b c

d e

Fig. 7  Sagittal reconstructions of the common carotid neck. Low keV virtual monoenergetic images at 40 keV
artery and the proximal internal carotid artery in a venous (b), 50 keV (c), 60 keV (d), and 70 keV (e) increase vessel
phase oncologic staging examination of the head and contrast compared to conventional images (a)
Head and Neck Imaging 139

Compliance with Ethical Standards of major salivary gland stones. Acta Radiol 60:1144–
1152. https://doi.org/10.1177/0284185118817906
Bray F, Ferlay J, Soerjomataram I et  al (2018) Global
Funding None.
cancer statistics 2018: GLOBOCAN estimates of
­incidence and mortality worldwide for 36 cancers in
Disclosure of Interests David Zopfs receives research 185 countries. CA Cancer J Clin 68:394–424. https://
support from Philips Healthcare. doi.org/10.3322/caac.21492
Cha J, Kim H-J, Kim ST et  al (2017) Dual-energy CT
with virtual monochromatic images and metal arti-
Ethical Approval  This article does not contain any stud-
fact reduction software for reducing metallic den-
ies with human participants performed by any of the
tal artifacts. Acta Radiol 58:1312–1319. https://doi.
authors.
org/10.1177/0284185117692174
Chawla A, Srinivasan S, Lim T-C et al (2017) Dual-energy
CT applications in salivary gland lesions. Br J Radiol
References 90. https://doi.org/10.1259/bjr.20160859
Deng K, Liu C, Ma R et al (2009) Clinical evaluation of
dual-energy bone removal in CT angiography of the
Agarwal JP, Sinha S, Goda JS et  al (2020) Tumor
head and neck: comparison with conventional bone-­
radiomic features complement clinico-radiological
subtraction CT angiography. Clin Radiol 64:534–541.
factors in predicting long-term local control and lar-
https://doi.org/10.1016/j.crad.2009.01.007
yngectomy free survival in locally advanced laryngo-­
Duan K, Gomez Hernandez K, Mete O (2015)
pharyngeal cancers. Br J Radiol 93:20190857. https://
Clinicopathological correlates of hyperparathyroid-
doi.org/10.1259/bjr.20190857
ism. J Clin Pathol 68:771–787. https://doi.org/10.1136/
Albrecht MH, Scholtz J-E, Kraft J et al (2015) Assessment
jclinpath-­2015-­203186
of an advanced monoenergetic reconstruction tech-
Forghani R (2019) An update on advanced dual-energy
nique in dual-energy computed tomography of head
CT for head and neck cancer imaging. Expert Rev
and neck cancer. Eur Radiol 25:2493–2501. https://
Anticancer Ther 19:633–644. https://doi.org/10.1080
doi.org/10.1007/s00330-­015-­3627-­1
/14737140.2019.1626234
Albrecht MH, Vogl TJ, Martin SS et  al (2019) Review
Forghani R, Levental M, Gupta R et al (2015) Different
of clinical applications for virtual monoenergetic
spectral hounsfield unit curve and high-energy virtual
dual-energy CT. Radiology 293:260–271. https://doi.
monochromatic image characteristics of squamous
org/10.1148/radiol.2019182297
cell carcinoma compared with nonossified thyroid
Anzidei M, Napoli A, Zaccagna F et al (2012) Diagnostic
cartilage. AJNR Am J Neuroradiol 36:1194–1200.
accuracy of colour Doppler ultrasonography, CT angi-
https://doi.org/10.3174/ajnr.A4253
ography and blood-pool-enhanced MR angiography in
Forghani R, Roskies M, Liu X et al (2016) Dual-energy
assessing carotid stenosis: a comparative study with
CT characteristics of parathyroid adenomas on 25-and
DSA in 170 patients. Radiol Med 117:54–71. https://
55-second 4D-CT acquisitions: preliminary experi-
doi.org/10.1007/s11547-­011-­0651-­3
ence. J Comput Assist Tomogr 40:806–814. https://
Axelsson L, Nyman J, Haugen-Cange H et  al (2017)
doi.org/10.1097/RCT.0000000000000442
Prognostic factors for head and neck cancer
Forghani R, Kelly H, Yu E et al (2017) Low-energy virtual
of unknown primary including the impact of
monochromatic dual-energy computed tomography
human papilloma virus infection. J Otolaryngol
images for the evaluation of head and neck squamous
Head Neck Surg 46:45. https://doi.org/10.1186/
cell carcinoma: a study of tumor visibility compared
s40463-­017-­0223-­1
with single-energy computed tomography and user
Bahig H, Lapointe A, Bedwani S et  al (2019) Dual-
acceptance. J Comput Assist Tomogr 41:565–571.
energy computed tomography for prediction of
https://doi.org/10.1097/RCT.0000000000000571
loco-regional recurrence after radiotherapy in lar-
Forghani R, Chatterjee A, Reinhold C et al (2019) Head
ynx and hypopharynx squamous cell carcinoma.
and neck squamous cell carcinoma: prediction of cer-
Eur J Radiol 110:1–6. https://doi.org/10.1016/j.
vical lymph node metastasis by dual-energy CT texture
ejrad.2018.11.005
analysis with machine learning. Eur Radiol 29:6172–
Bamberg F, Dierks A, Nikolaou K et  al (2011) Metal
6181. https://doi.org/10.1007/s00330-­019-­06159-­y
artifact reduction by dual energy computed tomog-
Gao S-Y, Zhang X-Y, Wei W et  al (2016) Identification
raphy using monoenergetic extrapolation. Eur
of benign and malignant thyroid nodules by in  vivo
Radiol 21:1424–1429. https://doi.org/10.1007/
iodine concentration measurement using single-source
s00330-­011-­2062-­1
dual energy CT: a retrospective diagnostic accuracy
Beland B, Levental M, Srinivasan A et al (2019) Practice
study. Medicine (Baltimore) 95:e4816. https://doi.
variations in salivary gland imaging and utility of virtual
org/10.1097/MD.0000000000004816
unenhanced dual energy CT images for the detection
140 D. Zopfs

Große Hokamp N, Laukamp KR, Lennartz S et al (2018) roid lesions. Head Neck 41:1024–1031. https://doi.
Artifact reduction from dental implants using virtual org/10.1002/hed.25524
monoenergetic reconstructions from novel spectral Leithner D, Mahmoudi S, Wichmann JL et  al (2018)
detector CT.  Eur J Radiol 104:136–142. https://doi. Evaluation of virtual monoenergetic imaging algorithms
org/10.1016/j.ejrad.2018.04.018 for dual-energy carotid and intracerebral CT angiogra-
Große Hokamp N, Maintz D, Shapira N et  al (2020) phy: effects on image quality, artefacts and diagnostic
Technical background of a novel detector-based performance for the detection of stenosis. Eur J Radiol
approach to dual-energy computed tomography. Diagn 99:111–117. https://doi.org/10.1016/j.ejrad.2017.12.024
Interv Radiol 26:68–71. https://doi.org/10.5152/ Leiva-Salinas C, Flors L, Durst CR et al (2016) Detection
dir.2019.19136 of parathyroid adenomas using a monophasic dual-­
Guggenberger R, Winklhofer S, Osterhoff G et al (2012) energy computed tomography acquisition: diag-
Metallic artefact reduction with monoenergetic dual-­ nostic performance and potential radiation dose
energy CT: systematic ex vivo evaluation of posterior reduction. Neuroradiology 58:1135–1141. https://doi.
spinal fusion implants from various vendors and dif- org/10.1007/s00234-­016-­1736-­4
ferent spine levels. Eur Radiol 22:2357–2364. https:// Lennartz S, Parakh A, Cao J et al (2021) Inter-scan and
doi.org/10.1007/s00330-­012-­2501-­7 inter-scanner variation of quantitative dual-energy
Guo R, Guo J, Zhang L et al (2020) CT-based radiomics CT: evaluation with three different scanner types. Eur
features in the prediction of thyroid cartilage inva- Radiol. https://doi.org/10.1007/s00330-­020-­07611-­0
sion from laryngeal and hypopharyngeal squamous Li M, Zheng X, Li J et  al (2012) Dual-energy com-
cell carcinoma. Cancer Imaging 20:81. https://doi. puted tomography imaging of thyroid nodule
org/10.1186/s40644-­020-­00359-­2 specimens: comparison with pathologic findings.
Kaemmerer N, Brand M, Hammon M et al (2016) Dual-­ Investig Radiol 47:58–64. https://doi.org/10.1097/
energy computed tomography angiography of the head RLI.0b013e318229fef3
and neck with single-source computed tomography: a Liu X, Ouyang D, Li H et al (2015) Papillary thyroid can-
new technical (split filter) approach for bone removal. cer: dual-energy spectral CT quantitative parameters
Investig Radiol 51:618–623. https://doi.org/10.1097/ for preoperative diagnosis of metastasis to the cervi-
RLI.0000000000000290 cal lymph nodes. Radiology 275:167–176. https://doi.
Kamalian S, Lev MH, Pomerantz SR (2017) Dual-energy org/10.1148/radiol.14140481
computed tomography angiography of the head Lohöfer FK, Kaissis GA, Köster FL et al (2018) Improved
and neck and related applications. Neuroimaging detection rates and treatment planning of head
Clin N Am 27:429–443. https://doi.org/10.1016/j. and neck cancer using dual-layer spectral CT.  Eur
nic.2017.04.009 Radiol 28:4925–4931. https://doi.org/10.1007/
Kuno H, Onaya H, Iwata R et  al (2012) Evaluation of s00330-­018-­5511-­2
cartilage invasion by laryngeal and hypopharyn- Magnano M, de Stefani A, Lerda W et al (1997) Prognostic
geal squamous cell carcinoma with dual-energy factors of cervical lymph node metastasis in head
CT. Radiology 265:488–496. https://doi.org/10.1148/ and neck squamous cell carcinoma. Tumori 83:
radiol.12111719 922–926
Kuno H, Onaya H, Fujii S et al (2014) Primary staging of Mannil M, Ramachandran J, Vittoria de Martini I et  al
laryngeal and hypopharyngeal cancer: CT, MR imag- (2017) Modified dual-energy algorithm for calci-
ing and dual-energy CT.  Eur J Radiol 83:e23–e35. fied plaque removal: evaluation in carotid computed
https://doi.org/10.1016/j.ejrad.2013.10.022 tomography angiography and comparison with digital
Kuno H, Sakamaki K, Fujii S et al (2018) Comparison of subtraction angiography. Investig Radiol 52:680–685.
MR imaging and dual-energy CT for the evaluation https://doi.org/10.1097/RLI.0000000000000391
of cartilage invasion by laryngeal and hypopharyngeal Martin SS, Albrecht MH, Wichmann JL et  al (2017)
squamous cell carcinoma. AJNR Am J Neuroradiol Value of a noise-optimized virtual monoenergetic
39:524–531. https://doi.org/10.3174/ajnr.A5530 reconstruction technique in dual-energy CT for
Laukamp KR, Zopfs D, Lennartz S et al (2019) Metal arti- planning of transcatheter aortic valve replacement.
facts in patients with large dental implants and bridges: Eur Radiol 27:705–714. https://doi.org/10.1007/
combination of metal artifact reduction algorithms and s00330-­016-­4422-­3
virtual monoenergetic images provides an approach to May MS, Wiesmueller M, Heiss R et al (2019) Comparison
handle even strongest artifacts. Eur Radiol 29:4228– of dual- and single-source dual-energy CT in head and
4238. https://doi.org/10.1007/s00330-­018-­5928-­7 neck imaging. Eur Radiol 29:4207–4214. https://doi.
Lee YH, Park KK, Song H-T et al (2012) Metal artefact org/10.1007/s00330-­018-­5762-­y
reduction in gemstone spectral imaging dual-energy Mori I, Machida Y, Osanai M et  al (2013) Photon star-
CT with and without metal artefact reduction software. vation artifacts of X-ray CT: their true cause and a
Eur Radiol 22:1331–1340. https://doi.org/10.1007/ solution. Radiol Phys Technol 6:130–141. https://doi.
s00330-­011-­2370-­5 org/10.1007/s12194-­012-­0179-­9
Lee DH, Lee YH, Seo HS et  al (2019) Dual-energy CT Morsbach F, Wurnig M, Kunz DM et  al (2013) Metal
iodine quantification for characterizing focal thy- artefact reduction from dental hardware in carotid
Head and Neck Imaging 141

CT angiography using iterative reconstructions. tions. Forensic Sci Med Pathol 9:327–332. https://doi.
Eur Radiol 23:2687–2694. https://doi.org/10.1007/ org/10.1007/s12024-­013-­9420-­z
s00330-­013-­2885-­z Takumi K, Hakamada H, Nagano H et al (2020) Usefulness
Neuhaus V, Große Hokamp N, Abdullayev N et  al of dual-layer spectral CT in follow-up examinations:
(2018) Comparison of virtual monoenergetic and diagnosing recurrent squamous cell carcinomas in the
polyenergetic images reconstructed from dual-layer head and neck. Jpn J Radiol. https://doi.org/10.1007/
detector CT angiography of the head and neck. Eur s11604-­020-­01071-­8
Radiol 28:1102–1110. https://doi.org/10.1007/ Tanaka R, Hayashi T, Ike M et  al (2013) Reduction of
s00330-­017-­5081-­8 dark-band-like metal artifacts caused by dental
Paul J, Tan MML, Farhang M et al (2013) Dual-energy CT implant bodies using hypothetical monoenergetic
spectral and energy weighted data sets: carotid steno- imaging after dual-energy computed tomography. Oral
sis and plaque detection. Acad Radiol 20:1144–1151. Surg Oral Med Oral Pathol Oral Radiol 115:833–838.
https://doi.org/10.1016/j.acra.2013.02.019 https://doi.org/10.1016/j.oooo.2013.03.014
Pérez-Lara A, Forghani R (2018) Spectral computed Tawfik AM, Razek AA, Kerl JM et al (2014) Comparison
tomography: technique and applications for head and of dual-energy CT-derived iodine content and iodine
neck cancer. Magn Reson Imaging Clin N Am 26:1– overlay of normal, inflammatory and metastatic
17. https://doi.org/10.1016/j.mric.2017.08.001 squamous cell carcinoma cervical lymph nodes.
Pulickal GG, Singh D, Lohan R et  al (2019) Dual-­ Eur Radiol 24:574–580. https://doi.org/10.1007/
source dual-energy CT in submandibular sialoli- s00330-­013-­3035-­3
thiasis: reliability and radiation burden. AJR Am J Toepker M, Czerny C, Ringl H et  al (2014) Can dual-­
Roentgenol 213:1291–1296. https://doi.org/10.2214/ energy CT improve the assessment of tumor margins
AJR.19.21299 in oral cancer? Oral Oncol 50:221–227. https://doi.
Riffel P, Haubenreisser H, Meyer M et al (2016) Carotid org/10.1016/j.oraloncology.2013.12.001
dual-energy CT angiography: evaluation of low keV Tomita H, Kuno H, Sekiya K et  al (2020) Quantitative
calculated monoenergetic datasets by means of a assessment of thyroid nodules using dual-energy com-
frequency-split approach for noise reduction at low puted tomography: iodine concentration measurement
keV levels. Eur J Radiol 85:720–725. https://doi. and multiparametric texture analysis for differentiating
org/10.1016/j.ejrad.2016.01.015 between malignant and benign lesions. Int J Endocrinol
Roele ED, Timmer VCML, Vaassen LAA et  al (2017) 2020:5484671. https://doi.org/10.1155/2020/5484671
Dual-energy CT in head and neck imaging. Curr Radiol Wichmann JL, Nöske E-M, Kraft J et  al (2014) Virtual
Rep 5:19. https://doi.org/10.1007/s40134-­017-­0213-­0 monoenergetic dual-energy computed tomography:
Rzymska-Grala I, Stopa Z, Grala B et al (2010) Salivary optimization of kiloelectron volt settings in head and
gland calculi - contemporary methods of imaging. Pol neck cancer. Investig Radiol 49:735–741. https://doi.
J Radiol 75:25–37 org/10.1097/RLI.0000000000000077
Saba L, Argioas GM, Lucatelli P et  al (2019) Variation Woisetschläger M, Gimm O, Johansson K et al (2020) Dual
of degree of stenosis quantification using differ- energy 4D-CT of parathyroid adenomas not clearly
ent energy level with dual energy CT scanner. localized by sestamibi scintigraphy and ultrasonogra-
Neuroradiology 61:285–291. https://doi.org/10.1007/ phy - a retrospective study. Eur J Radiol 124:108821.
s00234-­018-­2142-­x https://doi.org/10.1016/j.ejrad.2020.108821
Scholtz J-E, Hüsers K, Kaup M et  al (2015) Evaluation Yamauchi H, Buehler M, Goodsitt MM et  al (2016)
of image quality and dose reduction of 80 kVp neck Dual-energy CT-based differentiation of benign post-
computed tomography in patients with suspected peri- treatment changes from primary or recurrent malig-
tonsillar abscess. Clin Radiol 70:e67–e73. https://doi. nancy of the head and neck: comparison of spectral
org/10.1016/j.crad.2015.04.009 Hounsfield units at 40 and 70 keV and iodine concen-
Sheahan P (2014) Management of advanced laryngeal tration. AJR Am J Roentgenol 206:580–587. https://
cancer. Rambam Maimonides Med J 5:e0015. https:// doi.org/10.2214/AJR.15.14896
doi.org/10.5041/RMMJ.10149 Youserm DM, Huang T, Loevner LA et al (1997) Clinical
Srinivasan A, Parker RA, Manjunathan A et  al (2013) and economic impact of incidental thyroid lesions
Differentiation of benign and malignant neck found with CT and MR.  AJNR Am J Neuroradiol
pathologies: preliminary experience using spec- 18:1423–1428
tral computed tomography. J Comput Assist Zhou C, Zhao YE, Luo S et  al (2011) Monoenergetic
Tomogr 37:666–672. https://doi.org/10.1097/RCT. imaging of dual-energy CT reduces artifacts from
0b013e3182976365 implanted metal orthopedic devices in patients with
Steyer TE (2002) Peritonsillar abscess: diagnosis and factures. Acad Radiol 18:1252–1257. https://doi.
treatment. Am Fam Physician 65:93–96 org/10.1016/j.acra.2011.05.009
Stolzmann P, Winklhofer S, Schwendener N et al (2013) Zopfs D, Lennartz S, Laukamp K et al (2018) Improved
Monoenergetic computed tomography reconstructions depiction of atherosclerotic carotid artery stenosis in
reduce beam hardening artifacts from dental restora- virtual monoenergetic reconstructions of venous phase
142 D. Zopfs

dual-layer computed tomography in comparison to Zopfs D, Lennartz S, Pennig L et  al (2020c) Virtual
polyenergetic reconstructions. Eur J Radiol 100:36– monoenergetic images and post-processing algo-
42. https://doi.org/10.1016/j.ejrad.2018.01.008 rithms effectively reduce CT artifacts from intracra-
Zopfs D, Reimer RP, Sonnabend K et  al (2020a) nial aneurysm treatment. Sci Rep 10:6629. https://doi.
Intraindividual consistency of iodine concentration in org/10.1038/s41598-­020-­63574-­8
dual-energy computed tomography of the chest and Zopfs D, Große Hokamp N, Reimer R et  al (2021)
abdomen. Investig Radiol. https://doi.org/10.1097/ Value of spectral detector CT for pretherapeutic,
RLI.0000000000000724 locoregional assessment of esophageal cancer. Eur
Zopfs D, Graffe J, Reimer RP et al (2020b) Quantitative J Radiol 134:109423. https://doi.org/10.1016/j.
distribution of iodinated contrast media in body ejrad.2020.109423
computed tomography: data from a large refer-
ence cohort. Eur Radiol. https://doi.org/10.1007/
s00330-­020-­07298-­3
 Clinical Applications in Cardiac
Imaging

Basel Yacoub, Josua Decker, U. Joseph Schoepf,

Tilman Emrich, Jon F. Aldinger,
and Akos Varga-Szemes

Contents
1 History of Cardiac Imaging   144
2 Benefits and Applications of Cardiac CT Imaging   145
3 Functional Applications for Cardiac CT   146
4 Advantages of Spectral Imaging   146
4.1   ptimizing Image Quality 
O  146
4.2  Artifact Reduction   146
4.3  Reducing Contrast Media   146
4.4  Virtual Reconstructions to Lower Radiation Dose   147
5 Coronary Arteries   147
5.1  I mprovement of Imaging Quality and Artifact Reduction   147
5.2  General Improvements of Image Quality   147
5.3  Reducing Calcium Blooming Artifacts   148
5.4  Improved Visualization of Coronary Artery Stents   148
5.5  Improved Plaque Imaging   148
5.6  Calcium Scoring   149

B. Yacoub · U. J. Schoepf (*) · J. F. Aldinger ·

A. Varga-Szemes
Division of Cardiovascular Imaging, Department of
Radiology and Radiological Science, Medical
University of South Carolina, Charleston, SC, USA T. Emrich
e-mail: [email protected]; [email protected]; Division of Cardiovascular Imaging, Department of
[email protected]; [email protected] Radiology and Radiological Science, Medical
University of South Carolina, Charleston, SC, USA
J. Decker
Division of Cardiovascular Imaging, Department of Department of Diagnostic and Interventional
Radiology and Radiological Science, Medical Radiology, University Medical Center of the
University of South Carolina, Charleston, SC, USA Johannes Gutenberg University Mainz,
Mainz, Germany
Department of Diagnostic and Interventional
Radiology, University Hospital Augsburg, German Centre for Cardiovascular Research
Augsburg, Germany (DZHK), Partner Site Rhine-Main, Mainz, Germany
e-mail: [email protected] e-mail: [email protected]

© The Author(s), under exclusive license to Springer Nature Switzerland AG 2022 143
H. Alkadhi et al. (eds.), Spectral Imaging, Medical Radiology Diagnostic Imaging,
https://doi.org/10.1007/978-3-030-96285-2_10
144 B. Yacoub et al.

6 Myocardial Characterization   150

6.1  Myocardial Fibrosis   150
6.2  Iron Overload   150
6.3  Perfusion Imaging   151
6.4  Scar Imaging   152
7 Cardiac Valves   152
8 Cardiac Masses   153
9  utlook Cardiac CT 
O  153
9.1  Artificial Intelligence and Radiomics   153
9.2  Photon-Counting CT   154
9.3  Conclusion   155
References   155

Abstract 1 History of Cardiac Imaging

Cardiac computed tomography (CT) is an The value and potential of computed tomography
indispensable tool for evaluating cardiovascu- (CT) in cardiac imaging was realized soon after
lar structures. Its widespread use has increas- the first cardiovascular CT scans were performed
ingly gained momentum in the past decade as in the early 1980s. At that time, imaging options
results from large-scale trials became available, for the heart included angiography, radiography,
and it steadily found its way to become the early forms of magnetic resonance imaging
guidelines’ recommended first-line imaging (MRI), and echocardiography, the last of which
test for various cardiac conditions. Spectral was considered to be the reference standard.
imaging capabilities in cardiac CT first became These cutting-edge cardiovascular CT scans were
possible with the introduction of dual-­energy initially limited to diagnosing aortic dissections
computed tomography (DECT) scanners. and evaluating patency of coronary artery bypass
Cardiac CT examinations obtained using dual- grafts as it was evident that they had better diag-
energy techniques benefit from improved nostic performances for those purposes than
image quality, which enhances the diagnostic echocardiography. Within the following few
value of these scans. Additionally, DECT years, the use of cardiovascular CT expanded and
enables the generation of advanced postpro- found its way into clinical practice as an imaging
cessed image reconstructions that may be uti- modality for various cardiovascular abnormali-
lized to reduce the radiation and contrast media ties, including the evaluation of coronary arteries,
doses required in patients, thus making cardiac myocardial perfusion, ischemic cardiomyopathy,
CT scans safer overall. The latest emerging and intra-cardiac thrombi (Brundage and Lipton
advancement in CT imaging is photon-­counting 1982).
CT (PCCT) scanning that employs state-of-the- More so than other organ systems, cardiovas-
art photon detectors. This technology promises cular structures are particularly challenging to
higher spatial resolution and lower image noise, image, and therefore they are rarely, if ever, the
which are critical in visualizing the small struc- first organ system to be the center of investiga-
tures of the heart such as coronary arteries and tions by new imaging modalities and techniques.
stents. That said, the use of PCCT for cardiac There are several reasons for this distinguished
imaging is still under investigation with very status, and the severity of the resulting limitations
few research scans performed in human sub- vary by modality. The most prominent challenge
jects. As more validation studies are performed, to overcome is the heart’s perpetual movement
time will tell whether PCCT is set to become originating from myocardial contraction, motion
the next frontier of cardiac CT imaging. of cardiac valves, and respiration. There are three
Clinical Applications in Cardiac Imaging 145

possibilities to mitigate the impact of motion on spread popularity at many medical centers as
the quality of the resulting images. One would be the utilization of CT imaging continued to
to stop the movement of the heart, which is not climb. These scanners offer high temporal res-
performed in humans for obvious reasons. A sec- olution, necessary to image moving structures
ond option is to acquire images instantaneously, such as the myocardium and heart valves, and
as in echocardiography. The third method is to high spatial resolution, needed to differentiate
synchronize acquisition of the images to the small structures such as the coronary arteries.
rhythm of the heart, as in ECG-­ gated CT or The practicality and attractiveness for the use
MRI.  Another challenging attribute that makes of CT lies in the relative ease and speed in per-
imaging of the heart more problematic is its posi- forming examinations, making them particu-
tion. Being located deep within the thorax and larly useful for patients presenting to
protected by the rib cage, it is one of the two most emergency departments.
shielded organs in the human body, along with While no single modality offers a “one-stop-­
the brain. This attribute limits acoustic windows shop” for all cardiac imaging purposes, CT has
on ultrasonography, increases image noise on CT, become the guidelines’ recommended first-line
and reduces radiofrequency signal for MRI. All imaging test for various heart conditions.
these encountered difficulties are not prominent Coronary CT angiography (CCTA) is deemed a
when imaging other structures of the body. first-line test for stable chest pain as recom-
An ever-increasing demand for medical diag- mended by the American College of Cardiology/
nostic tools has fueled technological advances in American Heart Association guideline of 2012
the field of radiology, which in turn has greatly and European Society of Cardiology guideline of
expanded utilization of non-invasive cardiac 2019 (Knuuti et al. 2020; Fihn et al. 2012). CCTA
imaging. Today’s physicians have several is also a first-line investigation recommended in
­prominent options of imaging modalities for the patients presenting with acute chest pain, who are
cardiovascular system, such as CT, MRI, nuclear at low to intermediate risk of acute coronary syn-
stress testing, and echocardiography; each having drome (ACS), by a joint guideline published in
its own specific indications for appropriate use. 2010 by multiple medical societies spearheaded
This chapter will explore the use of dual-energy by the American College of Cardiology (Taylor
computed tomography (DECT), discuss benefits et al. 2010). This guideline also deems the use of
of multi-energy imaging, and provide an outlook cardiac CT appropriate for evaluation of systolic
on the future role and applications of photon-­ function, ventricular morphology, intra- and
counting computed tomography (PCCT) in car- extra-cardiac structures, adult congenital heart
diac imaging. diseases, as well as for preoperative assessment
prior to various cardiac procedures. The United
Kingdom’s National Institute for Health and Care
2 Benefits and Applications Excellence guideline published in 2017 also rec-
of Cardiac CT Imaging ommends CCTA as a first-line investigation for
all patients presenting with chest pain due to sus-
Cardiac CT has come a long way since its early pected coronary artery disease (CAD) (Moss
days and has grown significantly in complexity et al. 2017). Apart from being a tool to provide a
and clinical utility. Technical innovations have diagnosis, cardiac CT is also used to calculate
significantly reduced radiation doses associ- coronary artery calcium score and has also been
ated with CT imaging, and safety concerns pre- incorporated into guidelines for risk assessment
viously raised over exposure to ionizing of cardiovascular adverse events (Grundy et  al.
radiation have dwindled. Over the past decade, 2019).
advanced DECT scanners have gained wide-
146 B. Yacoub et al.

3 Functional Applications 2007). More advanced non-linear algorithms are

for Cardiac CT better able to optimize the blending process in
order to improve enhancement and reduce image
Cardiac CT imaging was predominantly used noise. This is done by selectively combining low-
solely for anatomical assessment, particularly in and high-energy image sets with varying ratios
evaluating the severity of CAD. Recent innova- for different regions of the CT dataset based on
tions in CT technology have increased cardiac attenuation.
CT’s potential applications. One of these ven-
tures has been a cardiac functional application of
CCTA to predict fractional flow reserve (FFR) 4.2 Artifact Reduction
by applying computational fluid dynamics. This
can be processed and evaluated from a standard Dual- and multi-energy imaging can assess mate-
CCTA without the need for additional image rial differential attenuation at each energy level,
acquisitions (Baumann et  al. 2021; Schwartz which enables the measurement of the fractions
et al. 2019). Another use for cardiac CT in func- of materials in each voxel. With this material-­
tional assessment of the heart has been for the specific information on hand, the attenuation of
measurement of blood perfusion before and after each voxel can be extrapolated to any desired
myocardial stress. This technique can provide energy level using complex algorithms to gener-
valuable information on the hemodynamic sig- ate virtual monoenergetic images (VMI). These
nificance of coronary artery stenoses (Cannaò images reduce beam hardening at higher energy
et  al. 2015). In addition to detecting coronary levels, which consequently reduces high attenua-
plaque and indicating its burden, CCTA can also tion artifacts such as metal artifacts from stents
provide detailed characterization of plaque mor- and blooming artifacts from coronary artery cal-
phology which can aid in cardiac risk assess- cification (Mangold et al. 2016a).
ment (Nerlekar et  al. 2018; Motoyama et  al.
2015).
4.3 Reducing Contrast Media

4 Advantages of Spectral In turn, low-energy VMI are able to substantially

Imaging increase contrast enhancement, which is espe-
cially useful in CT scans where the contrast
4.1 Optimizing Image Quality media bolus is suboptimally timed. This can
enable a reduction in the volume and concentra-
The capacity to acquire CT images at more than tion of the contrast media administered while still
one energy level and combining data obtained achieving adequate enhancement in the lumen of
from all energy levels enables benefits that would vessels and myocardium. However, these low-­
not be possible on single-energy acquisitions. energy VMI may suffer from an increase in image
Information from each energy level image set is noise as a tradeoff for higher contrast attenuation
then postprocessed to create blended CT images. (Albrecht et  al. 2019). Newer processing algo-
Basic postprocessing algorithms use linear blend- rithms known as VMI plus offer a solution for
ing and are based on blending image sets with a this limitation as they provide the ability to
specific ratio assigned to each energy level decompose data from each energy level. This
(Cavedon and Rudin 2015). The resulting images serves to combine the benefits of high contrast
will have characteristics similar to those that enhancement on low-energy image sets with the
would have been obtained at an energy level low noise information on high-energy image sets,
intermediate to those of the image sets used, in contrast enhanced acquisitions (Lenga et  al.
depending on ratios assigned (Johnson et  al. 2017).
Clinical Applications in Cardiac Imaging 147

4.4 Virtual Reconstructions pected CAD, a lot of effort has been placed into
to Lower Radiation Dose improving and optimizing the diagnostic quality
of its images. Using spectral imaging, the lynch-
With material decomposition information avail- pin of further improvements is the generation of
able through the analysis of element dependent VMI reconstructions. In short, information from
attenuation on multi-energy imaging, it is possi- multi-energy CT datasets is used to reconstruct
ble to create material-specific color overlay maps approximated VMI that represent CT images as
of the CT dataset. These can be used to generate they would be obtained using a true monoener-
iodine maps for measuring its concentration in getic X-ray beam at specific tube voltages (Yu
tissue, for example. It is also possible to remove et al. 2011). Different studies have demonstrated
the iodine overlay maps to generate virtual non- that when using VMI, overall image quality is
contrast (VNC) images which can potentially improved as compared to conventional 120 kVp
eliminate the need for a true non-­contrast (TNC) images or polychromatic CT (Matsumoto et  al.
acquisition, thus reducing both scan times and 2011; Pomerantz et al. 2013). Specific approaches
the patients’ exposure to radiation. These VNC in which spectral imaging, especially using VMI,
images have been shown to have excellent corre- can enhance the diagnostic value of CCTA are
lation with TNC images (Yamada et  al. 2014). described here.
Given its higher energy discriminating capabili-
ties, the use of PCCT can provide even more dis-
crete information on spectral attenuation than 5.2 General Improvements
DECT.  Theoretically, this can be utilized to of Image Quality
achieve more accurate material decomposition
(McCollough et al. 2015). Inadequate CCTA image quality may be caused
The major restricting factor in reducing radia- by various factors such as insufficient contrast
tion dose in scans acquired on single- or density within the coronary arteries due to
­dual-­energy CT is the resulting increase in image patient’s obesity, inaccurate scan timing, insuffi-
noise. This limitation may be ameliorated by the cient contrast volume or contrast extravasation.
physical capabilities of PCCT which enable it to The resulting diagnostic limitations due to poor
discriminate the low-amplitude signals causing image signal-to-noise ratio (SNR) and contrast-­
electronic image noise and to exclude them by to-­noise ratio (CNR) may potentially necessitate
adjusting the low-energy threshold. Thus, the use a repeat scan with additional contrast administra-
of PCCT results in improved image quality and tion and radiation exposure (Yan et  al. 2013).
higher contrast-to-noise ratios (Wang et al. 2012; Low-energy VMI at 40 keV have been shown to
Rajagopal et  al. 2020). This capacity serves to boost iodine attenuation in suboptimal vascular
increase the diagnostic quality on low-dose CT studies, improve both objective and subjective
scans, particularly in morbidly obese patients who image quality and obviate the need to repeat
require higher radiation doses to obtain diagnostic scans (Yu et al. 2011; Arendt et al. 2020; Grant
quality images (den Harder et al. 2016). et  al. 2014; Kalisz et  al. 2017). Furthermore,
using VMI generated at 40–50  keV showed
improved image quality, evaluability, and diag-
5 Coronary Arteries nostic accuracy compared to single-energy CCTA
(Albrecht et  al. 2016; Yi et  al. 2019; Andreini
5.1 Improvement of Imaging et al. 2015). The amplification of iodine contrast
Quality and Artifact may also be utilized in administration of smaller
Reduction volumes of contrast media. Two studies reported
equivalent coronary image quality in CCTA
With CCTA being established as the first-line test using only half of contrast media and amplifying
in the workup of patients presenting with sus- iodine contrast using VMI at 50–60  keV
148 B. Yacoub et al.

(Carrascosa et  al. 2015; Huang et  al. 2020). arteries limit diagnostic quality, particularly
Mangold et al. investigated the use of DECT in when needing to evaluate the stent lumen. This
obese patients and reported a routinely obtain- may restrict performing a full assessment which
able diagnostic image quality of CCTA (Mangold is important given that in-stent restenosis is
et al. 2016b). Ohta et al. described that coronary reported in about 5–10% of coronary artery stents
lumen in general is best assessed on 70 keV VMI, (Gogas et al. 2013). Again, the use of high-energy
which showed the overall best SNC and CNR VMI have been shown to improve image quality
(Ohta et al. 2017). by reducing noise and beam hardening artifacts
(Fig. 1) (Fuchs et al. 2013; Kuchenbecker et al.
2015; Pessis et  al. 2013; Secchi et  al. 2015; Yu
5.3  educing Calcium Blooming
R et al. 2012; Zou and Silver 2009). Several groups
Artifacts have reported reliable stent imaging with
improved in-lumen visibility using VMI gener-
Another advantage of multi-energy imaging is ated at 80–130  keV acquired using DECT as
the ability to reduce calcium blooming artifacts. compared to reconstructions from single-energy
Such artifacts may potentially cause an overesti- CT (Mangold et  al. 2016a; Boll et  al. 2008b;
mation of stenosis on CCTA and possibly lead to Hickethier et al. 2017; Stehli et al. 2015).
unnecessary invasive coronary angiography in
patients with suspected CAD (Yan et  al. 2013;
Brodoefel et  al. 2008; Cademartiri et  al. 2005; 5.5 Improved Plaque Imaging
Kruk et al. 2014; Zhang et al. 2008). Using high-­
energy VMI, earlier studies by Boll et  al. and It is an established fact that different histomor-
Scheffel et  al. reported a reduction in calcium phologic characteristics of coronary artery
blooming artifacts and improvement in lumen plaques have distinct prognostic implications
visualization when extensive calcifications were (Narula et al. 2013). Using CCTA, it may be pos-
present (Boll et al. 2008a; Scheffel et al. 2006). sible to identify several plaque features such as
Several following studies showed that calcium plaque burden, positive remodeling, napkin ring
blooming artifacts are significantly diminished sign, and small spotty calcifications which have
on VMI generated at 80–90 keV, decreasing the been associated with instability and high risk for
stenotic grading and increasing luminal dimen- ACS (Andreini et  al. 2020; Danad et  al. 2015;
sions which led to more accurate assessments of Ferencik et  al. 2018; Hoffmann et  al. 2006;
coronary artery stenosis (Foley et al. 2016; Kang Williams et al. 2019). Promising earlier ex vivo
et al. 2010; Scheske et al. 2013; Van Hedent et al. studies showed very good potential for DECT in
2018; Wang et al. 2011). A different approach in further differentiation of certain plaque features.
reducing calcium blooming artifacts involves Barreto et al. compared 80 and 140 kVp images
using calcium subtraction images generated to show changes in attenuation of densely calci-
using spectral CT data. These have also been fied and fibrocalcific plaques (Barreto et  al.
shown to improve coronary lumen visualization 2008). The ability to discriminate between lipid-­
and diagnostic performance in patients with rich and fibrous plaques using both 80 and 140
heavily calcified lesions (De Santis et  al. 2018; kVp images has been reported by Tanami et al.
Yunaga et al. 2017). (Tanami et al. 2010). Zachrisson et al. described
better discrimination of soft tissues occurring in
plaques with DECT (Zachrisson et  al. 2010).
5.4 Improved Visualization Furthermore, Obaid et al. showed improved dif-
of Coronary Artery Stents ferentiation of necrotic core and fibrous plaque in
ex vivo arteries using DECT that was not, how-
Similar to calcium blooming artifacts, high atten- ever, translated to in vivo imaging due to reduced
uation artifacts from metallic stents in coronary image quality (Obaid et  al. 2014). A different
Clinical Applications in Cardiac Imaging 149

Fig. 1 Curved
multiplanar a b
reconstruction (MPR)
from (a) 120 keV
reconstruction generated
from DECT and (b)
SECT. VMI
reconstructions on
DECT reduce image
noise and beam
hardening artifacts in
imaging of coronary
stents and improve the
overall visibility of the
lumen which is
important to assess for
in-stent restenosis

approach by Haghighi et al. showed the potential However, such findings are yet to be applied in
of DECT to investigate the composition of non-­ practice for clinical assessments.
calcified plaques using electron density and
effective atomic numbers (Haghighi et al. 2015).
VMI reconstructions at varying energy levels 5.6 Calcium Scoring
may also provide increased accuracy in diagnos-
ing stenoses in regard to plaque composition. For Using multi-energy CT data, VNC images can be
example, Stehli et al. reported that 90 keV VMI generated from contrast enhanced CT scans by
provided the best luminal evaluation when calci- subtracting the component of the iodine attenua-
fied and mixed plaques are present (Stehli et al. tion from the CT attenuation number (Fig.  2)
2016). More recent literature also reported that (McCollough et  al. 2015). These images had
spectral CT imaging can further improve CNR gained a special interest since they have potential
and the differentiation of plaque components on to reduce both radiation dose and scan time and
CCTA, which aids in more accurate assessment to eliminate the need for pre-contrast scans of the
of plaque vulnerability (Boussel et  al. 2014; heart when assessing coronary artery calcium
Mandal et  al. 2018; Symons et  al. 2018a). (CAC) burden (Yamada et  al. 2014; Kay 2020;
150 B. Yacoub et al.

a b c

Fig. 2 (a) Contrast enhanced, (b) virtual non-contrast of TNC and the calcifications can be easily visualized
(VNC), and (c) true non-contrast (TNC) acquisitions from without necessitating an additional non-enhanced CT
a patient with coronary artery calcifications in the acquisition
LAD. The quality of the VNC image is comparable to that

Kim et al. 2009). Different studies showed high values (Nacif et  al. 2012, 2013; Kurita et  al.
correlation between TNC and VNC images, how- 2016). Through material decomposition capabili-
ever, all of them reported lower CAC values ties of multi-energy imaging, which allow the
when reading VNC images, which led to a more differentiation of iodine attenuation at more than
limited clinical applicability (Yamada et al. 2014; one energy level, it is possible to quantify iodine
Fuchs et  al. 2014; Schwarz et  al. 2012; Song content and distribution and to use it as a surro-
et  al. 2016). A recent study by Nadjiri et  al. gate marker for blood volume (Fig. 3). This has
reported that VNC underestimated CAC volume led to more investigations assessing ECV values
and plaque density and that CAC scores obtained derived from a single delayed phase DECT
from VNC are approximately half of those from acquisition with processing of VNC images. Van
TNC images (Nadjiri et al. 2018). The results of Assen et  al. demonstrated the feasibility of this
these studies demonstrate that there is potential approach and its ability to differentiate diseased
to reliably calculate CAC scores using VNC from healthy myocardium (van Assen et  al.
images, but further validation is required before 2019a). Furthermore, Abadia et  al. quantified
they can be implemented in clinical practice. myocardial ECV using iodine maps without VNC
or TNC images and established the cutoff values
that may be used to identify diseased tissue.
6 Myocardial Characterization (Abadia et al. 2020).

6.1 Myocardial Fibrosis

6.2 Iron Overload
Myocardial extracellular volume (ECV) fraction
is increased as a result of myocardial remodeling Myocardial iron deposition is caused by iron
and subsequent fibrosis, which are hallmarks of overload states such as hemochromatosis, certain
several cardiomyopathies. Conventionally, MRI hematologic diseases or frequent blood transfu-
has been used to evaluate ECV with a both true sions. These form another spectrum of conditions
non-contrast and delayed contrast phase acquisi- for which DECT may be a fast and reliable diag-
tions (Schoepf 2019). Given the wider availabil- nostic tool. Although the gold standard for diag-
ity of CT, its shorter examination times, and the nosing iron overload in organs is tissue biopsy,
shortfalls of MRI in patients with metal implants, imaging with MRI has been widely utilized in
several groups have examined and found good clinical practice due to its non-invasive nature
correlation between CT and MRI derived ECV (Chu et  al. 2012). An early feasibility study by
Clinical Applications in Cardiac Imaging 151

a b

Fig. 3 (a) An illustration of an iodine map obtained from polar map, generated by a fully automated method, show-
a delayed contrast-enhanced DECT scan showing manu- ing a regional elevation in ECV readings (arrows) that
ally performed segmentations of the left ventricular and indicate myocardial scarring from cardiomyopathy
blood pool (b) Myocardial extracellular volume (ECV)

Hazirolan et  al. on cardiac DECT performed in select appropriate treatments and interventions
thalassemia patients showed good correlation for (Pijls et  al. 2010). There are two technical
HU values from the septal muscles with corre- approaches for analyzing myocardial perfusion
sponding T2* values in MRI.  Later phantom on cardiac CT using rest and stress images. The
studies evaluating iron content in porcine cardiac first and more basic approach is the static “single-­
tissue and tube phantoms also found very high shot” scan with image acquisition performed in a
correlations for measurements performed on narrow temporal window within the early first-­
DECT and MRI scans (Tsai et al. 2014; Ibrahim pass arterial phase. The second approach is a
and Bowman 2014). Conversely, a more recent dynamic one that uses several consecutive acqui-
study by Ma et al. assessing iron overload with sitions throughout the cardiac cycle and captures
DECT and MRI in patients with histories of the first pass of contrast during wash-in and
myelodysplastic syndrome or aplastic anemia wash-out phases which makes quantitative analy-
failed to show a good correlation in myocardial sis of myocardial blood flow possible.
iron content measurements. This is despite find- Dynamic CT for myocardial perfusion has
ing a strong correlation in liver iron content mea- been shown to have a clear advantage over static
surements obtained from both modalities (Ma CT and shows high diagnostic accuracy in detect-
et al. 2020). ing ischemia when compared to MRI and SPECT
(Bamberg et  al. 2014; Sørgaard et  al. 2016;
Caruso et al. 2016). The added benefit for DECT
6.3 Perfusion Imaging over SECT in perfusion imaging is its ability to
evaluate the distribution of blood using iodine
Adding functional assessment of the heart in maps, which more accurately reflect its content,
addition to the anatomical evaluation that is tradi- rather than through the measurement of contrast
tionally reported on every cardiac CT scan pro- attenuation on CT images which is utilized in
vides a more comprehensive representation of SECT (Schoepf 2019). This is supported by the
cardiac physiology. It can provide valuable infor- results of Arnoldi et. el. that examined the accu-
mation on the hemodynamic relevance of coro- racy for detection of myocardial perfusion defi-
nary stenosis and allow physicians to timely cits on SECT and DECT scans and concluded
152 B. Yacoub et al.

that iodine maps from DECT showed better cor- that these images had a mean CNR of 9.14 with a
relation with SPECT findings in diagnosing myo- sensitivity and specificity of 92% and 98%,
cardial hypoperfusion (Arnoldi et  al. 2011). As respectively, in detecting the myocardial delayed
such, iodine concentration has the potential to enhancement that represents scar tissue (Ohta
differentiate between ischemic and normal myo- et al. 2018).
cardium (van Assen et al. 2019b).

7 Cardiac Valves
6.4 Scar Imaging
Cardiac CT provides a comprehensive modality
Delayed gadolinium enhancement on CMR has for non-invasive anatomical examination of the
been the go-to imaging test for evaluating myo- cardiac valves. This makes it an optimal tool for
cardial scaring given its higher sensitivity and pre-procedural planning in patients scheduled to
CNR compared to delayed enhanced SECT undergo transcatheter aortic valve replacement
(Nieman et  al. 2008). Yet, spectral imaging CT (TAVR). CT examinations for this purpose
using VMI reconstructions can slightly improve involve measurements of geometrical dimensions
the performance of CT for that use (Fig.  4). of the aortic valve, its degree of calcification and
Sandort et al. conducted a DECT study on human its distance from the coronary ostia. These values
and canine subjects to evaluate CNR values in are necessary to appropriately select prosthesis
infarcted myocardial tissue. They concluded that model and size, which is of paramount impor-
VMI plus reconstructions at 40  keV improved tance in improving post-operative outcomes
infarct delineation in canine subjects, with histol- (Binder et  al. 2013; Tops et  al. 2008). Multi-­
ogy as a reference standard, by up to 25% com- energy imaging has been thoroughly evaluated in
pared to conventional VMI. These reconstructions this particular patient group and its potential and
also provided higher CNR compared to conven- added value in improving the CT protocols has
tional VMI and linearly blended images, yet their been examined.
values did not exceed 4.2. In contrast, CNR val- The value of noise-optimized VMI reconstruc-
ues from delayed gadolinium enhancement on tions was assessed in the prospective Spectral CT
MRI are generally greater than 10 (Sandfort et al. Assessment Prior to TAVR (SPECTACULAR)
2017). Undeterred by underwhelming results for study which showed the feasibility of using VMI
a role for VMI plus in scar imaging, Ohta et al. with administering lower contrast media doses
evaluated the use of iodine density images against while ensuring a comprehensive evaluation of
MRI in detecting and classifying myocardial vessel access and aortic root dimensions (Cavallo
scarring patients with heart failure. They reported et  al. 2020). Reducing the risk of acute kidney

a b c

Fig. 4 (a) Phase sensitive inversion recovery (PSIR) the inferior ventricular wall is observed on all three
sequence on delayed enhancement MRI, (b) iodine map modalities. The defect is illustrated as hyperenhancement
from DECT, and (c) single-photon emission computed in (a), reduction of iodine content in (b), and reduction in
tomography (SPECT). Perfusion abnormality (arrows) in radiotracer uptake in (c)
Clinical Applications in Cardiac Imaging 153

injury is particularly desired in TAVR candidates standard transesophageal echocardiography

as renal dysfunction is present in about half of (TEE) as a reference, and reported each of the
this patient group (Faggiano et  al. 2012). The sensitivity, specificity, negative predictive value,
image quality of noise-optimized VMI plus and positive predicted value to exceed 95% (Hur
reconstructions was evaluated in a study by et  al. 2012). As such, DECT can provide an
Martin et al. where it was compared to traditional excellent alternative for diagnosis given that TEE
VMI reconstructions as well as standard linearly is often uncomfortable to patients and requires
blended images. It concluded that these advanced specialized skills to perform it and interpret its
reconstructions significantly increased CNR and results. Additionally, material composition capa-
SNR over the others (Martin et al. 2017). bilities of DECT are useful in differentiating
Given that metallic aortic valves demonstrate between cardiac masses. Results from Hong et al.
moderate to severe high attenuation artifact, illustrate that iodine concentrations on DECT
assessment for post-procedural complications were significantly higher in cardiac myxomas
such as valve thrombosis or paravalvular leaks compared to thrombi, while attenuation values in
may become complicated on CT (Suchá et  al. Hounsfield units showed no difference between
2015). For this purpose, Schwartz et  al. con- the two (Hong et al. 2014).
ducted a study on 80 patients with aortic valve
replacement to quantify metal artifact and
reported that high keV VMI offered up to 17.2% 9 Outlook Cardiac CT
artifact reduction when compared with standard
SECT (Schwartz et al. 2020). Finally, dual source 9.1 Artificial Intelligence
CT scanners may also be operated in single-­ and Radiomics
energy mode at low-tube-voltage acquisitions to
safely and effectively evaluate TAVR candidates Being in a medical specialty that is heavily reli-
using a lower radiation dose (Felmly et al. 2017). ant on technology, radiologists are often among
the pioneering physicians who openly embrace
and incorporate new innovations into their prac-
8 Cardiac Masses tice. As with other sectors of industry, the rise of
artificial intelligence (AI) promises to revolution-
Intra-cardiac masses such as lipomas, fibromas, ize many workplaces, including departments of
myxomas, and thrombi are very rare entities radiology. AI has found its way into cardiac CT
found in less than 0.1% of autopsy studies and workflows and has been used for applications
are often assessed by echocardiography as a first-­ such as the automated detection of coronary
line diagnostic modality. Not infrequently, artery calcium, measurement of myocardial per-
another confirmatory imaging modality would be fusion, and measurement of CT derived FFR
requested when echocardiography is inconclu- (Monti et  al. 2020; Fischer et  al. 2020; Tesche
sive due to poor acoustic windows or artifacts and Gray 2020). Besides its use in detection or
(Mankad and Herrmann 2016). CT offers a high-­ measurement of findings, AI has also been uti-
resolution volumetric evaluation of the heart and lized for optimizing scanning protocols for indi-
can explore a cardiac tumor’s relationship with vidual patient requirements and for improving
adjacent structures such as chambers, arteries, CT image quality and reducing image noise
valves, myocardium or epicardium. It can also (Wang et al. 2019; Sharma et al. 2020). Radiomics
offer a comprehensive four-dimensional anatom- is another discipline whose potential in cardiac
ical assessment in planning for surgical resection CT is also being explored. This developing field
of cardiac tumors if indicated (Young et al. 2019). aims to utilize spatial and texture information
Hur et  al. evaluated the diagnostic performance from every voxel on a CT image for use in quan-
of cardiac DECT in detecting left atrial append- titative analysis in order to identify biomarkers
age thrombi, using the findings from the gold-­ that can expand diagnostic and predictive abili-
154 B. Yacoub et al.

ties based on imaging (Kolossváry et al. 2018). A ing factor when trying to reduce radiation dose in
recent study by Kolossváry et  al. demonstrated CT imaging and the optimized noise properties
that cluster analysis of radiomic features of coro- of PCCT images may be the key to overcome that
nary plaque on CCTA differentiated between limitation. Given the physical abilities of detec-
patients with traditional cardiovascular risk fac- tors in PCCT, they have to ability to significantly
tors and those with nontraditional risk factors decrease spectral overlap. This in turn leads to
such as cocaine use and HIV. This enables mor- improved CNR compared to DECT which further
phologic phenotyping of CAD and provides more enhances its spectral imaging capabilities such as
insight into the pathogenesis of CAD (Kolossváry material decomposition (Sandfort et al. 2020).
et al. 2021). Such new developments hold great First pre-clinical PCCT studies reported
promise for use in cardiac CT and will continue improved SNR in iodine enhanced scans that
to modernize the practice of radiology in the may translate to further reductions in the contrast
coming years and may even be the driving force doses necessary in CT imaging (Leng et  al.
for a new era of personalized precision medicine 2017). Increased spatial resolution and multi-­
in management of cardiac disease (Schoepf and material composition bear the possibility to fur-
Emrich 2021). ther improve imaging of the coronary arteries and
to aid in reducing calcium blooming artifacts
(Holmes et al. 2020; Leng et al. 2018). Increased
9.2 Photon-Counting CT spatial resolution has also been shown to improve
visualization of coronary artery stents and stent-­
A principal development of PCCT over DECT lumen evaluation, which allows for non-invasive
lies in its photon detector systems used. The assessment of in-stent restenosis (Mannil et  al.
advanced physical properties of detectors in 2018; Bratke et  al. 2020; Sigovan et  al. 2019;
PCCT scanners enable them to have higher dose Symons et al. 2018b). Another promising field of
efficiency and be smaller in size. This in turn PCCT is the assessment of CAC score. Having a
vastly improves spatial resolution and quality of CAC score of zero has been described as a
the resulting images when compared side-by-side 15-year warranty period against mortality in
with those obtained from DECT scans that use asymptomatic patients (Valenti et  al. 2015).
identical tube voltage, tube current, and recon- Because of this, it is important to identify patients
struction parameters (Mannil et al. 2018). Thus, with ultra-low calcium scores and further dis-
PCCT leads to overall increased diagnostic abili- criminate between patients with zero and near
ties and confidence when assessing miniscule zero CAC scores that may still be at risk (Hsieh
structures such as small coronary plaques or the et  al. 2020; Senoner et  al. 2020). First in  vivo
distal coronary arteries. experiences by Symons et al. showed that PCCT
At the time of writing of this chapter, the field can significantly improve CAC scoring and might
of cardiac PCCT is still in its infancy. The num- even reduce radiation dose while maintaining
ber of published articles evaluating the heart diagnostic quality (Symons et  al. 2019). It was
using this novel technology remains in the single shown that PCCT has the potential to accurately
digits with only one article describing the use of quantify the mass and density of CAC lesions
PCCT for imaging of the heart human volunteers. which might quickly translate into its use in clini-
Hence the information on its use in cardiac imag- cal practice, and to obsolescence of the currently
ing derives from pre-clinical phantom studies, as necessary TNC cardiac CT scans (Juntunen et al.
well as from extrapolating experiences with dual-­ 2020). Also, further discrimination of calcified
energy spectral imaging. The major technical lesions by their density, shape and distribution
advantages for use of PCCT lie in its increased could enhance risk assessment models and lead
spatial resolution and lower noise levels leading to more personalized management plans of
to better quality images. Image noise is a restrict- patients (Sandfort et al. 2020).
Clinical Applications in Cardiac Imaging 155

9.3 Conclusion ate healthy from cardiomyopathic myocardium using

dual-source dual-energy CT.  J Cardiovasc Comput
Tomogr 14(2):162–167. https://doi.org/10.1016/j.
Cardiac CT is a fundamental tool for imaging of the jcct.2019.09.008
heart. It is the guideline-recommended first-­line test Albrecht MH, Trommer J, Wichmann JL, Scholtz J-E,
in a variety of cardiac conditions, plays an essential Martin SS, Lehnert T, Vogl TJ, Bodelle B (2016)
Comprehensive comparison of virtual monoener-
role in planning of cardiac procedures, and is uti-
getic and linearly blended reconstruction techniques
lized in cardiovascular risk stratification of patients. in third-generation dual-source dual-energy com-
It has gradually developed from a static and purely puted tomography angiography of the thorax and
anatomical image modality and is now able to pro- abdomen. Investig Radiol 51(9):582–590. https://doi.
org/10.1097/RLI.0000000000000272
vide us with functional information on the heart.
Albrecht MH, Vogl TJ, Martin SS, Nance JW, Duguay
Additionally, the rise of spectral imaging capabili- TM, Wichmann JL, De Cecco CN, Varga-Szemes A,
ties in CT with dual-­energy scanners has enabled van Assen M, Tesche C, Schoepf UJ (2019) Review
the differentiation of materials based on atomic of clinical applications for virtual monoenergetic dual-­
energy CT.  Radiology 293(2):260–271. https://doi.
numbers, as well as the development of complex org/10.1148/radiol.2019182297
post-processing image reconstructions algorithms. Andreini D, Pontone G, Mushtaq S, Bertella E, Conte
These may be used to make CT a safer option for E, Segurini C, Giovannardi M, Baggiano A, Annoni
patients by reducing the contrast and radiation doses A, Formenti A, Petullà M, Beltrama V, Volpato V,
Bartorelli AL, Trabattoni D, Fiorentini C, Pepi M
necessary. More recent advancements in photon
­ (2015) Diagnostic accuracy of rapid kilovolt peak–
detector technologies of CT systems have led to the switching dual-energy CT coronary angiography in
introduction of PCCT scanners that have a promis- patients with a high calcium score. JACC Cardiovasc
ing potential for cardiac imaging due to their higher Imaging 8(6):746–748. https://doi.org/10.1016/j.
jcmg.2014.10.013
spatial resolution and the advantages that they carry Andreini D, Magnoni M, Conte E, Masson S, Mushtaq
for improved visualization of the coronary arteries. S, Berti S, Canestrari M, Casolo G, Gabrielli D,
The refinement of these CT systems as well as Latini R, Marraccini P, Moccetti T, Modena MG,
research on their use in human subjects is still ongo- Pontone G, Gorini M, Maggioni AP, Maseri A (2020)
Coronary plaque features on CTA can identify patients
ing. Based on findings over the next few years, we at increased risk of cardiovascular events. JACC
may be witnessing the steady surge of PCCT to Cardiovasc Imaging 13(8):1704–1717. https://doi.
become the mainstream of cardiac CT imaging. org/10.1016/j.jcmg.2019.06.019
Arendt CT, Czwikla R, Lenga L, Wichmann JL, Albrecht
MH, Booz C, Martin SS, Leithner D, Tischendorf P,
Compliance with Ethical Standards
Blandino A, Vogl TJ, D’Angelo T (2020) Improved
coronary artery contrast enhancement using noise-­
Disclosure of Interests  U. Joseph Schoepf has received optimised virtual monoenergetic imaging from
institutional research support, consulting fees, and/or dual-source dual-energy computed tomography.
speaker honoraria from Bayer, Bracco, Elucid, GE, Eur J Radiol 122:108666. https://doi.org/10.1016/j.
Guerbet, HeartFlow Inc., Keya Medical, and Siemens. ejrad.2019.108666
Tilman Emrich has received a speaker fee and travel sup- Arnoldi E, Lee YS, Ruzsics B, Weininger M, Spears
port from Siemens. Akos Varga-Szemes receives institu- JR, Rowley CP, Chiaramida SA, Costello P, Reiser
tional research support from Siemens and is a consultant MF, Schoepf UJ (2011) CT detection of myocardial
for Bayer and Elucid Bioimaging. The other authors have blood volume deficits: dual-energy CT compared
no potential conflict of interest to disclose. with single-energy CT spectra. J Cardiovasc Comput
Tomogr 5(6):421–429. https://doi.org/10.1016/j.
jcct.2011.10.007
Funding This book chapter did not receive any grant
Bamberg F, Marcus RP, Becker A, Hildebrandt K, Bauner
from funding agencies in the public, commercial, or not-
K, Schwarz F, Greif M, von Ziegler F, Bischoff B,
for-profit sectors.
Becker HC, Johnson TR, Reiser MF, Nikolaou K,
Theisen D (2014) Dynamic myocardial CT perfu-
sion imaging for evaluation of myocardial ischemia
References as determined by MR imaging. JACC Cardiovasc
Imaging 7(3):267–277. https://doi.org/10.1016/j.
jcmg.2013.06.008
Abadia AF, van Assen M, Martin SS, Vingiani V, Griffith
Barreto M, Schoenhagen P, Nair A, Amatangelo S, Milite
LP, Giovagnoli DA, Bauer MJ, Schoepf UJ (2020)
M, Obuchowski NA, Lieber ML, Halliburton SS
Myocardial extracellular volume fraction to differenti-
(2008) Potential of dual-energy computed tomography
156 B. Yacoub et al.

to characterize atherosclerotic plaque: ex vivo assess- multislice computed tomography: observations in

ment of human coronary arteries in comparison to his- an ex vivo model of coronary computed tomography
tology. J Cardiovasc Comput Tomogr 2(4):234–242. angiography. Eur Radiol 15(7):1426–1431. https://doi.
https://doi.org/10.1016/j.jcct.2008.05.146 org/10.1007/s00330-­005-­2697-­x
Baumann S, Kaeder F, Schoepf UJ, Golden JW, Kryeziu Cannaò PM, Schoepf UJ, Muscogiuri G, Wichmann JL,
P, Tesche C, Renker M, Jannsen S, Weiss C, Hetjens Fuller SR, Secchi F, Varga-Szemes A, De Cecco CN
S, Schoenberg SO, Borggrefe M, Akin I, Lossnitzer D, (2015) Technical prerequisites and imaging protocols
Overhoff D (2021) Prognostic value of coronary com- for dynamic and dual energy myocardial perfusion
puted tomography angiography-derived morphologic imaging. Eur J Radiol 84(12):2401–2410. https://doi.
and quantitative plaque markers using semiautomated org/10.1016/j.ejrad.2015.02.010
plaque software. J Thorac Imaging 36(2):108–115. Carrascosa P, Leipsic JA, Capunay C, Deviggiano A,
https://doi.org/10.1097/rti.0000000000000509 Vallejos J, Goldsmit A, Rodriguez-Granillo GA (2015)
Binder RK, Webb JG, Willson AB, Urena M, Hansson Monochromatic image reconstruction by dual energy
NC, Norgaard BL, Pibarot P, Barbanti M, Larose E, imaging allows half iodine load computed tomogra-
Freeman M, Dumont E, Thompson C, Wheeler M, phy coronary angiography. Eur J Radiol 84(10):1915–
Moss RR, Yang TH, Pasian S, Hague CJ, Nguyen G, 1920. https://doi.org/10.1016/j.ejrad.2015.06.019
Raju R, Toggweiler S, Min JK, Wood DA, Rodés-­ Caruso D, Eid M, Schoepf UJ, Jin KN, Varga-Szemes
Cabau J, Leipsic J (2013) The impact of integration A, Tesche C, Mangold S, Spandorfer A, Laghi A, De
of a multidetector computed tomography annulus Cecco CN (2016) Dynamic CT myocardial perfusion
area sizing algorithm on outcomes of transcatheter imaging. Eur J Radiol 85(10):1893–1899. https://doi.
aortic valve replacement: a prospective, multicenter, org/10.1016/j.ejrad.2016.07.017
­controlled trial. J Am Coll Cardiol 62(5):431–438. Cavallo AU, Patterson AJ, Thomas R, Alaiti MA, Attizzani
https://doi.org/10.1016/j.jacc.2013.04.036 GF, Laukamp K, Große Hokamp N, Bezerra H,
Boll DT, Merkle EM, Paulson EK, Mirza RA, Fleiter Gilkeson R, Rajagopalan S (2020) Low dose contrast
TR (2008a) Calcified vascular plaque specimens: CT for transcatheter aortic valve replacement assess-
assessment with cardiac dual-energy multidetector ment: results from the prospective SPECTACULAR
CT in anthropomorphically moving heart phantom. study (spectral CT assessment prior to TAVR). J
Radiology 249(1):119–126. https://doi.org/10.1148/ Cardiovasc Comput Tomogr 14(1):68–74. https://doi.
radiol.2483071576 org/10.1016/j.jcct.2019.06.015
Boll DT, Merkle EM, Paulson EK, Fleiter TR (2008b) Cavedon C, Rudin S (2015) Cardiovascular and neuro-
Coronary stent patency: dual-energy multidetector vascular imaging: physics and technology. CRC Press,
CT assessment in a pilot study with anthropomorphic Boca Raton
phantom. Radiology 247(3):687–695. https://doi. Chu WC, Au WY, Lam WW (2012) MRI of cardiac iron
org/10.1148/radiol.2473070849 overload. J Magn Reson Imaging 36(5):1052–1059.
Boussel L, Coulon P, Thran A, Roessl E, Martens G, https://doi.org/10.1002/jmri.23628
Sigovan M, Douek P (2014) Photon counting spectral Danad I, Fayad ZA, Willemink MJ, Min JK (2015)
CT component analysis of coronary artery athero- New applications of cardiac computed tomography.
sclerotic plaque samples. BJR 87(1040):20130798. JACC Cardiovasc Imaging 8(6):710–723. https://doi.
https://doi.org/10.1259/bjr.20130798 org/10.1016/j.jcmg.2015.03.005
Bratke G, Hickethier T, Bar-Ness D, Bunck AC, Maintz D, De Santis D, Jin KN, Schoepf UJ, Grant KL, De Cecco
Pahn G, Coulon P, Si-Mohamed S, Douek P, Sigovan CN, Nance JW, Vogl TJ, Laghi A, Albrecht MH (2018)
M (2020) Spectral photon-counting computed tomog- Heavily calcified coronary arteries: advanced calcium
raphy for coronary stent imaging: evaluation of the subtraction improves luminal visualization and diag-
potential clinical impact for the delineation of in-stent nostic confidence in dual-energy coronary computed
restenosis. Investig Radiol 55(2):61–67. https://doi. tomography angiography. Investig Radiol 53(2):103–
org/10.1097/RLI.0000000000000610 109. https://doi.org/10.1097/RLI.0000000000000416
Brodoefel H, Burgstahler C, Tsiflikas I, Reimann A, den Harder AM, Willemink MJ, de Jong PA, Schilham
Schroeder S, Claussen CD, Heuschmid M, Kopp AF AM, Rajiah P, Takx RA, Leiner T (2016) New hori-
(2008) Dual-source CT: effect of heart rate, heart zons in cardiac CT.  Clin Radiol 71(8):758–767.
rate variability, and calcification on image quality https://doi.org/10.1016/j.crad.2016.01.022
and diagnostic accuracy. Radiology 247(2):346–355. Faggiano P, Frattini S, Zilioli V, Rossi A, Nistri S, Dini
https://doi.org/10.1148/radiol.2472070906 FL, Lorusso R, Tomasi C, Cas LD (2012) Prevalence
Brundage BH, Lipton MJ (1982) The emergence of com- of comorbidities and associated cardiac diseases
puted tomography as a cardiovascular diagnostic in patients with valve aortic stenosis. Potential
technique. Am Heart J 103(2):313–316. https://doi. implications for the decision-making process. Int
org/10.1016/0002-­8703(82)90517-­8 J Cardiol 159(2):94–99. https://doi.org/10.1016/j.
Cademartiri F, Mollet NR, Runza G, Bruining N, Hamers ijcard.2011.02.026
R, Somers P, Knaapen M, Verheye S, Midiri M, Krestin Felmly LM, De Cecco CN, Schoepf UJ, Varga-Szemes
GP, de Feyter PJ (2005) Influence of intracoronary A, Mangold S, McQuiston AD, Litwin SE, Bayer
attenuation on coronary plaque measurements using RR 2nd, Vogl TJ, Wichmann JL (2017) Low contrast
Clinical Applications in Cardiac Imaging 157

medium-volume third-generation dual-source com- Imaging 30(7):1399–1405. https://doi.org/10.1007/

puted tomography angiography for transcatheter aortic s10554-­014-­0474-­0
valve replacement planning. Eur Radiol 27(5):1944– Gogas BD, Garcia-Garcia HM, Onuma Y, Muramatsu T,
1953. https://doi.org/10.1007/s00330-­016-­4537-­6 Farooq V, Bourantas CV, Serruys PW (2013) Edge
Ferencik M, Mayrhofer T, Bittner DO, Emami H, Puchner vascular response after percutaneous coronary inter-
SB, Lu MT, Meyersohn NM, Ivanov AV, Adami EC, vention. J Am Coll Cardiol Intv 6(3):211–221. https://
Patel MR, Mark DB, Udelson JE, Lee KL, Douglas doi.org/10.1016/j.jcin.2013.01.132
PS, Hoffmann U (2018) Use of high-risk coronary Grant KL, Flohr TG, Krauss B, Sedlmair M, Thomas C,
atherosclerotic plaque detection for risk stratifica- Schmidt B (2014) Assessment of an advanced image-­
tion of patients with stable chest pain: a secondary based technique to calculate virtual monoenergetic
analysis of the PROMISE randomized clinical trial. computed tomographic images from a dual-energy
JAMA Cardiol 3(2):144. https://doi.org/10.1001/ examination to improve contrast-to-noise ratio in
jamacardio.2017.4973 examinations using iodinated contrast media. Investig
Fihn SD, Gardin JM, Abrams J, Berra K, Blankenship Radiol 49(9):586–592. https://doi.org/10.1097/
JC, Dallas AP, Douglas PS, Foody JM, Gerber TC, RLI.0000000000000060
Hinderliter AL, King SB 3rd, Kligfield PD, Krumholz Grundy SM, Stone NJ, Bailey AL, Beam C, Birtcher
HM, Kwong RY, Lim MJ, Linderbaum JA, Mack MJ, KK, Blumenthal RS, Braun LT, de Ferranti S, Faiella-­
Munger MA, Prager RL, Sabik JF, Shaw LJ, Sikkema Tommasino J, Forman DE, Goldberg R, Heidenreich
JD, Smith CR Jr, Smith SC Jr, Spertus JA, Williams PA, Hlatky MA, Jones DW, Lloyd-Jones D, Lopez-­
SV, Anderson JL (2012) 2012 ACCF/AHA/ACP/ Pajares N, Ndumele CE, Orringer CE, Peralta
AATS/PCNA/SCAI/STS guideline for the ­diagnosis CA, Saseen JJ, Smith SC Jr, Sperling L, Virani SS,
and management of patients with stable ischemic Yeboah J (2019) 2018 AHA/ACC/AACVPR/AAPA/
heart disease: a report of the American College of ABC/ACPM/ADA/AGS/APhA/ASPC/NLA/PCNA
Cardiology Foundation/American Heart Association guideline on the management of blood cholesterol:
task force on practice guidelines, and the American executive summary: a report of the American College
College of Physicians, American Association for of Cardiology/American Heart Association Task
Thoracic Surgery, Preventive Cardiovascular Nurses Force on Clinical Practice Guidelines. J Am Coll
Association, Society for Cardiovascular Angiography Cardiol 73(24):3168–3209. https://doi.org/10.1016/j.
and Interventions, and Society of Thoracic Surgeons. jacc.2018.11.002
Circulation 126(25):e354–e471. https://doi. Haghighi RR, Chatterjee S, Tabin M, Sharma S, Jagia P,
org/10.1161/CIR.0b013e318277d6a0 Ray R, Singh RP, Yadav R, Sharma M, Krishna K,
Fischer AM, Eid M, De Cecco CN, Gulsun MA, van Vani VC, Lakshmi R, Mandal SR, Kumar P, Arava
Assen M, Nance JW, Sahbaee P, De Santis D, Bauer S (2015) DECT evaluation of noncalcified coronary
MJ, Jacobs BE, Varga-Szemes A, Kabakus IM, artery plaque. Med Phys 42(10):5945–5954. https://
Sharma P, Jackson LJ, Schoepf UJ (2020) Accuracy doi.org/10.1118/1.4929935
of an artificial intelligence deep learning algorithm Hickethier T, Baeßler B, Kroeger JR, Doerner J,
implementing a recurrent neural network with long Pahn G, Maintz D, Michels G, Bunck AC (2017)
short-term memory for the automated detection of cal- Monoenergetic reconstructions for imaging of coro-
cified plaques from coronary computed tomography nary artery stents using spectral detector CT: in-vitro
angiography. J Thorac Imaging 35:S49–S57. https:// experience and comparison to conventional images. J
doi.org/10.1097/rti.0000000000000491 Cardiovasc Comput Tomogr 11(1):33–39. https://doi.
Foley WD, Shuman WP, Siegel MJ, Sahani DV, Boll org/10.1016/j.jcct.2016.12.005
DT, Bolus DN, De Cecco CN, Kaza RK, Morgan Hoffmann U, Moselewski F, Nieman K, Jang I-K, Ferencik
DE, Schoepf UJ, Vrtiska TJ, Yeh BM, Berland LL M, Rahman AM, Cury RC, Abbara S, Joneidi-Jafari
(2016) White paper of the Society of Computed Body H, Achenbach S, Brady TJ (2006) Noninvasive assess-
Tomography and Magnetic Resonance on dual-energy ment of plaque morphology and composition in culprit
CT, part 2: radiation dose and iodine sensitivity. J and stable lesions in acute coronary syndrome and sta-
Comput Assist Tomogr 40(6):846–850. https://doi. ble lesions in stable angina by multidetector computed
org/10.1097/RCT.0000000000000539 tomography. J Am Coll Cardiol 47(8):1655–1662.
Fuchs TA, Stehli J, Fiechter M, Dougoud S, Gebhard C, https://doi.org/10.1016/j.jacc.2006.01.041
Ghadri JR, Husmann L, Gaemperli O, Kaufmann PA Holmes T, Ulzheimer S, Pourmorteza A (2020) Dose-­
(2013) First experience with monochromatic coronary efficient ultra-high-resolution imaging of calcified
computed tomography angiography from a 64-slice coronary artery stenoses with photon-counting Ct. J
CT scanner with gemstone spectral imaging (GSI). J Cardiovasc Comput Tomogr 14(3):S69. https://doi.
Cardiovasc Comput Tomogr 7(1):25–31. https://doi. org/10.1016/j.jcct.2020.06.133
org/10.1016/j.jcct.2013.01.004 Hong YJ, Hur J, Kim YJ, Lee HJ, Hong SR, Suh YJ, Kim
Fuchs TA, Stehli J, Dougoud S, Sah B-R, Bull S, Clerc HY, Lee JW, Choi BW (2014) Dual-energy cardiac
OF, Possner M, Buechel RR, Gaemperli O, Kaufmann computed tomography for differentiating cardiac
PA (2014) Coronary artery calcium quantification myxoma from thrombus. Int J Cardiovasc Imaging
from contrast enhanced CT using gemstone spectral 30(Suppl 2):121–128. https://doi.org/10.1007/
imaging and material decomposition. Int J Cardiovasc s10554-­014-­0490-­0
158 B. Yacoub et al.

Hsieh SS, Leng S, Rajendran K, Tao S, McCollough Kolossváry M, Kellermayer M, Merkely B, Maurovich-­
CH (2020) Photon counting CT: clinical applica- Horvat P (2018) Cardiac computed tomography
tions and future developments. IEEE Trans Radiat radiomics: a comprehensive review on radiomic tech-
Plasma Med Sci:1–1. https://doi.org/10.1109/ niques. J Thorac Imaging 33(1):26–34. https://doi.
TRPMS.2020.3020212 org/10.1097/rti.0000000000000268
Huang X, Gao S, Ma Y, Lu X, Jia Z, Hou Y (2020) The Kolossváry M, Gerstenblith G, Bluemke DA, Fishman
optimal monoenergetic spectral image level of coro- EK, Mandler RN, Kickler TS, Chen S, Bhatia S, Lai S,
nary computed tomography (CT) angiography on a Lai H (2021) Contribution of risk factors to the devel-
dual-layer spectral detector CT with half-dose con- opment of coronary atherosclerosis as confirmed via
trast media. Quant Imaging Med Surg 10(3):592–603. coronary CT angiography: a longitudinal radiomics-­
https://doi.org/10.21037/qims.2020.02.17 based study. Radiology 299(1):97–106. https://doi.
Hur J, Kim YJ, Lee HJ, Nam JE, Hong YJ, Kim HY, org/10.1148/radiol.2021203179
Lee JW, Choi BW (2012) Cardioembolic stroke: Kruk M, Noll D, Achenbach S, Mintz GS, Pręgowski
dual-energy cardiac CT for differentiation of left J, Kaczmarska E, Kryczka K, Pracoń R, Dzielińska
atrial appendage thrombus and circulatory stasis. Z, Śleszycka J, Witkowski A, Demkow M, Rużyłło
Radiology 263(3):688–695. https://doi.org/10.1148/ W, Kępka C (2014) Impact of coronary artery cal-
radiol.12111691 cium characteristics on accuracy of CT angiography.
Ibrahim E-SH, Bowman AW (2014) Evaluation JACC Cardiovasc Imaging 7(1):49–58. https://doi.
of iron overload: dual-energy computed org/10.1016/j.jcmg.2013.07.013
tomography versus magnetic resonance imaging.
­ Kuchenbecker S, Faby S, Sawall S, Lell M, Kachelrieß
J Cardiovasc Magn Reson 16(1):O92. https://doi. M (2015) Dual energy CT: how well can pseudo-­
org/10.1186/1532-­429X-­16-­S1-­O92 monochromatic imaging reduce metal artifacts?:
Johnson TR, Krauss B, Sedlmair M, Grasruck M, Bruder DECT: can monoenergetic imaging remove metal
H, Morhard D, Fink C, Weckbach S, Lenhard M, artifacts? Med Phys 42(2):1023–1036. https://doi.
Schmidt B, Flohr T, Reiser MF, Becker CR (2007) org/10.1118/1.4905106
Material differentiation by dual energy CT: initial Kurita Y, Kitagawa K, Kurobe Y, Nakamori S, Nakajima H,
experience. Eur Radiol 17(6):1510–1517. https://doi. Dohi K, Ito M, Sakuma H (2016) Estimation of myo-
org/10.1007/s00330-­006-­0517-­6 cardial extracellular volume fraction with cardiac CT in
Juntunen MAK, Inkinen SI, Ketola JH, Kotiaho A, subjects without clinical coronary artery disease: a fea-
Kauppinen M, Winkler A, Nieminen MT (2020) sibility study. J Cardiovasc Comput Tomogr 10(3):237–
Framework for photon counting quantitative material 241. https://doi.org/10.1016/j.jcct.2016.02.001
decomposition. IEEE Trans Med Imaging 39(1):35– Leng S, Zhou W, Yu Z, Halaweish A, Krauss B, Schmidt
47. https://doi.org/10.1109/TMI.2019.2914370 B, Yu L, Kappler S, McCollough C (2017) Spectral
Kalisz K, Halliburton S, Abbara S, Leipsic JA, performance of a whole-body research photon count-
Albrecht MH, Schoepf UJ, Rajiah P (2017) Update ing detector CT: quantitative accuracy in derived
on cardiovascular applications of multienergy image sets. Phys Med Biol 62(17):7216–7232. https://
CT.  RadioGraphics 37(7):1955–1974. https://doi. doi.org/10.1088/1361-­6560/aa8103
org/10.1148/rg.2017170100 Leng S, Rajendran K, Gong H, Zhou W, Halaweish
Kang DK, Schoepf UJ, Bastarrika G, Nance JW, Abro JA, AF, Henning A, Kappler S, Baer M, Fletcher JG,
Ruzsics B (2010) Dual-energy computed tomography McCollough CH (2018) 150-μm spatial resolution
for integrative imaging of coronary artery disease: using photon-counting detector computed tomography
principles and clinical applications. Semin Ultrasound technology: technical performance and first patient
CT MRI 31(4):276–291. https://doi.org/10.1053/j. images. Investig Radiol 53(11):655–662. https://doi.
sult.2010.05.004 org/10.1097/RLI.0000000000000488
Kay FU (2020) Dual-energy CT and coronary imaging. Lenga L, Albrecht MH, Othman AE, Martin SS, Leithner
Cardiovasc Diagn Ther 10(4):1090–1107. https://doi. D, D’Angelo T, Arendt C, Scholtz JE, De Cecco
org/10.21037/cdt.2020.04.04 CN, Schoepf UJ, Vogl TJ, Wichmann JL (2017)
Kim KP, Einstein AJ, Berrington de González A Monoenergetic dual-energy computed tomographic
(2009) Coronary artery calcification screening: imaging: cardiothoracic applications. J Thorac
estimated radiation dose and cancer risk. Arch Imaging 32(3):151–158. https://doi.org/10.1097/
Intern Med 169(13):1188. https://doi.org/10.1001/ rti.0000000000000259
archinternmed.2009.162 Ma Q, Hu J, Yang W, Hou Y (2020) Dual-layer detector
Knuuti J, Wijns W, Saraste A, Capodanno D, Barbato spectral CT versus magnetic resonance imaging for the
E, Funck-Brentano C, Prescott E, Storey RF, Deaton assessment of iron overload in myelodysplastic syn-
C, Cuisset T, Agewall S, Dickstein K, Edvardsen T, dromes and aplastic anemia. Jpn J Radiol 38(4):374–
Escaned J, Gersh BJ, Svitil P, Gilard M, Hasdai D, 381. https://doi.org/10.1007/s11604-­020-­00921-­9
Hatala R, Mahfoud F, Masip J, Muneretto C, Valgimigli Mandal SR, Bharati A, Haghighi RR, Arava S, Ray R,
M, Achenbach S, Bax JJ (2020) 2019 ESC guidelines Jagia P, Sharma S, Chatterjee S, Tabin M, Sharma
for the diagnosis and management of chronic coronary M, Sharma S, Kumar P (2018) Non-invasive charac-
syndromes. Eur Heart J 41(3):407–477. https://doi. terization of coronary artery atherosclerotic plaque
org/10.1093/eurheartj/ehz425 using dual energy CT: explanation in ex-vivo sam-
Clinical Applications in Cardiac Imaging 159

ples. Phys Med 45:52–58. https://doi.org/10.1016/j. Nacif MS, Kawel N, Lee JJ, Chen X, Yao J, Zavodni
ejmp.2017.12.006 A, Sibley CT, Lima JA, Liu S, Bluemke DA (2012)
Mangold S, Cannaó PM, Schoepf UJ, Wichmann JL, Interstitial myocardial fibrosis assessed as extracel-
Canstein C, Fuller SR, Muscogiuri G, Varga-Szemes lular volume fraction with low-radiation-dose car-
A, Nikolaou K, De Cecco CN (2016a) Impact of an diac CT.  Radiology 264(3):876–883. https://doi.
advanced image-based monoenergetic reconstruc- org/10.1148/radiol.12112458
tion algorithm on coronary stent visualization using Nacif MS, Liu Y, Yao J, Liu S, Sibley CT, Summers RM,
third generation dual-source dual-energy CT: a phan- Bluemke DA (2013) 3D left ventricular extracel-
tom study. Eur Radiol 26(6):1871–1878. https://doi. lular volume fraction by low-radiation dose cardiac
org/10.1007/s00330-­015-­3997-­4 CT: assessment of interstitial myocardial fibrosis. J
Mangold S, Wichmann JL, Schoepf UJ, Litwin SE, Cardiovasc Comput Tomogr 7(1):51–57. https://doi.
Canstein C, Varga-Szemes A, Muscogiuri G, Fuller org/10.1016/j.jcct.2012.10.010
SR, Stubenrauch AC, Nikolaou K, De Cecco CN Nadjiri J, Kaissis G, Meurer F, Weis F, Laugwitz K-L,
(2016b) Coronary CT angiography in obese patients Straeter AS, Muenzel D, Noël PB, Rummeny EJ,
using 3rd generation dual-source CT: effect of body Rasper M (2018) Accuracy of calcium scoring cal-
mass index on image quality. Eur Radiol 26(9):2937– culated from contrast-enhanced coronary computed
2946. https://doi.org/10.1007/s00330-­015-­4161-­x tomography angiography using a dual-layer spectral
Mankad R, Herrmann J (2016) Cardiac tumors: echo CT: a comparison of calcium scoring from real and
assessment. Echo Res Pract 3(4):R65–r77. https://doi. virtual non-contrast data. PLoS One 13(12):e0208588.
org/10.1530/erp-­16-­0035 https://doi.org/10.1371/journal.pone.0208588
Mannil M, Hickethier T, von Spiczak J, Baer M, Henning Narula J, Nakano M, Virmani R, Kolodgie FD, Petersen
A, Hertel M, Schmidt B, Flohr T, Maintz D, Alkadhi R, Newcomb R, Malik S, Fuster V, Finn AV (2013)
H (2018) Photon-counting CT: high-resolution imag- Histopathologic characteristics of atherosclerotic
ing of coronary stents. Investig Radiol 53(3):143–149. coronary disease and implications of the findings for
https://doi.org/10.1097/RLI.0000000000000420 the invasive and noninvasive detection of vulnerable
Martin SS, Albrecht MH, Wichmann JL, Hüsers K, plaques. J Am Coll Cardiol 61(10):1041–1051. https://
Scholtz JE, Booz C, Bodelle B, Bauer RW, Metzger doi.org/10.1016/j.jacc.2012.10.054
SC, Vogl TJ, Lehnert T (2017) Value of a noise-­ Nerlekar N, Ha FJ, Cheshire C, Rashid H, Cameron JD,
optimized virtual monoenergetic reconstruction tech- Wong DT, Seneviratne S, Brown AJ (2018) Computed
nique in dual-energy CT for planning of transcatheter tomographic coronary angiography-derived plaque
aortic valve replacement. Eur Radiol 27(2):705–714. characteristics predict major adverse cardiovascu-
https://doi.org/10.1007/s00330-­016-­4422-­3 lar events: a systematic review and meta-analysis.
Matsumoto K, Jinzaki M, Tanami Y, Ueno A, Yamada M, Circ Cardiovasc Imaging 11(1):e006973. https://doi.
Kuribayashi S (2011) Virtual monochromatic spectral org/10.1161/circimaging.117.006973
imaging with fast kilovoltage switching: improved Nieman K, Shapiro MD, Ferencik M, Nomura CH,
image quality as compared with that obtained with Abbara S, Hoffmann U, Gold HK, Jang IK, Brady
conventional 120-kVp CT.  Radiology 259(1):257– TJ, Cury RC (2008) Reperfused myocardial infarc-
262. https://doi.org/10.1148/radiol.11100978 tion: contrast-enhanced 64-Section CT in comparison
McCollough CH, Leng S, Yu L, Fletcher JG (2015) Dual- to MR imaging. Radiology 247(1):49–56. https://doi.
and multi-energy CT: principles, technical approaches, org/10.1148/radiol.2471070332
and clinical applications. Radiology 276(3):637–653. Obaid DR, Calvert PA, Gopalan D, Parker RA, West NEJ,
https://doi.org/10.1148/radiol.2015142631 Goddard M, Rudd JHF, Bennett MR (2014) Dual-­
Monti CB, Codari M, van Assen M, De Cecco CN, energy computed tomography imaging to determine
Vliegenthart R (2020) Machine learning and deep neu- atherosclerotic plaque composition: a prospective
ral networks applications in computed tomography for study with tissue validation. J Cardiovasc Comput
coronary artery disease and myocardial perfusion. J Tomogr 8(3):230–237. https://doi.org/10.1016/j.
Thorac Imaging 35:S58–S65. https://doi.org/10.1097/ jcct.2014.04.007
rti.0000000000000490 Ohta Y, Kitao S, Watanabe T, Kishimoto J, Yamamoto
Moss AJ, Williams MC, Newby DE, Nicol ED (2017) K, Ogawa T (2017) Evaluation of image quality of
The updated NICE guidelines: cardiac CT as the first-­ coronary artery plaque with rapid kVp-switching
line test for coronary artery disease. Curr Cardiovasc dual-energy CT.  Clin Imaging 43:42–49. https://doi.
Imaging Rep 10(5):15. https://doi.org/10.1007/ org/10.1016/j.clinimag.2017.01.014
s12410-­017-­9412-­6 Ohta Y, Kitao S, Yunaga H, Fujii S, Mukai N, Yamamoto
Motoyama S, Ito H, Sarai M, Kondo T, Kawai H, K, Ogawa T (2018) Myocardial delayed enhance-
Nagahara Y, Harigaya H, Kan S, Anno H, Takahashi ment CT for the evaluation of heart failure: compari-
H, Naruse H, Ishii J, Hecht H, Shaw LJ, Ozaki Y, son to MRI.  Radiology 288(3):682–691. https://doi.
Narula J (2015) Plaque characterization by coronary org/10.1148/radiol.2018172523
computed tomography angiography and the likelihood Pessis E, Campagna R, Sverzut J-M, Bach F, Rodallec
of acute coronary events in mid-term follow-up. J Am M, Guerini H, Feydy A, Drapé J-L (2013) Virtual
Coll Cardiol 66(4):337–346. https://doi.org/10.1016/j. monochromatic spectral imaging with fast kilo-
jacc.2015.05.069 voltage switching: reduction of metal artifacts at
160 B. Yacoub et al.

CT.  Radiographics 33(2):573–583. https://doi. chest pain. J Thorac Imaging 34(1):12–17. https://doi.
org/10.1148/rg.332125124 org/10.1097/rti.0000000000000369
Pijls NH, Fearon WF, Tonino PA, Siebert U, Ikeno F, Schwartz FR, Tailor T, Gaca JG, Kiefer T, Harrison K,
Bornschein B, van’t Veer M, Klauss V, Manoharan G, Hughes GC, Ramirez-Giraldo JC, Marin D, Hurwitz
Engstrøm T, Oldroyd KG, Ver Lee PN, MacCarthy PA, LM (2020) Impact of dual energy cardiac CT for
De Bruyne B (2010) Fractional flow reserve versus metal artefact reduction post aortic valve replacement.
angiography for guiding percutaneous coronary inter- Eur J Radiol 129:109135. https://doi.org/10.1016/j.
vention in patients with multivessel coronary artery ejrad.2020.109135
disease: 2-year follow-up of the FAME (Fractional Schwarz F, Nance JW, Ruzsics B, Bastarrika G, Sterzik A,
Flow Reserve Versus Angiography for Multivessel Schoepf UJ (2012) Quantification of coronary artery
Evaluation) study. J Am Coll Cardiol 56(3):177–184. calcium on the basis of dual-energy coronary CT
https://doi.org/10.1016/j.jacc.2010.04.012 angiography. Radiology 264(3):700–707. https://doi.
Pomerantz SR, Kamalian S, Zhang D, Gupta R, Rapalino org/10.1148/radiol.12112455
O, Sahani DV, Lev MH (2013) Virtual monochromatic Secchi F, De Cecco CN, Spearman JV, Silverman JR,
reconstruction of dual-energy unenhanced head CT Ebersberger U, Sardanelli F, Schoepf UJ (2015)
at 65–75 keV maximizes image quality compared Monoenergetic extrapolation of cardiac dual energy
with conventional polychromatic CT.  Radiology CT for artifact reduction. Acta Radiol 56(4):413–418.
266(1):318–325. https://doi.org/10.1148/ https://doi.org/10.1177/0284185114527867
radiol.12111604 Senoner T, Plank F, Beyer C, Langer C, Birkl K, Steinkohl
Rajagopal JR, Farhadi F, Solomon J, Sahbaee P, F, Widmann G, Barbieri F, Adukauskaite A, Friedrich
Saboury B, Pritchard WF, Jones EC, Samei E (2020) G, Dichtl W, Feuchtner GM (2020) Does coronary cal-
Comparison of low dose performance of photon-­ cium score zero reliably rule out coronary artery dis-
counting and energy integrating CT.  Acad Radiol. ease in low-to-intermediate risk patients? A coronary
https://doi.org/10.1016/j.acra.2020.07.033 CTA study. J Cardiovasc Comput Tomogr 14(2):155–
Sandfort V, Palanisamy S, Symons R, Pourmorteza A, 161. https://doi.org/10.1016/j.jcct.2019.09.009
Ahlman MA, Rice K, Thomas T, Davies-Venn C, Krauss Sharma P, Suehling M, Flohr T, Comaniciu D (2020)
B, Kwan A, Pandey A, Zimmerman SL, Bluemke DA Artificial intelligence in diagnostic imaging: sta-
(2017) Optimized energy of spectral CT for infarct tus quo, challenges, and future opportunities. J
imaging: experimental validation with human valida- Thorac Imaging 35(Suppl 1):S11–s16. https://doi.
tion. J Cardiovasc Comput Tomogr 11(3):171–178. org/10.1097/rti.0000000000000499
https://doi.org/10.1016/j.jcct.2017.02.003 Sigovan M, Si-Mohamed S, Bar-Ness D, Mitchell J,
Sandfort V, Persson M, Pourmorteza A, Noël PB, Langlois J-B, Coulon P, Roessl E, Blevis I, Rokni M,
Fleischmann D, Willemink MJ (2020) Spectral Rioufol G, Douek P, Boussel L (2019) Feasibility of
photon-counting CT in cardiovascular imaging. J improving vascular imaging in the presence of metallic
Cardiovasc Comput Tomogr. https://doi.org/10.1016/j. stents using spectral photon counting CT and K-edge
jcct.2020.12.005 imaging. Sci Rep 9(1):19850. https://doi.org/10.1038/
Scheffel H, Alkadhi H, Plass A, Vachenauer R, Desbiolles s41598-­019-­56427-­6
L, Gaemperli O, Schepis T, Frauenfelder T, Schertler Song I, Yi JG, Park JH, Kim SM, Lee KS, Chung MJ
T, Husmann L, Grunenfelder J, Genoni M, Kaufmann (2016) Virtual non-contrast CT using dual-energy
PA, Marincek B, Leschka S (2006) Accuracy of dual-­ spectral CT: feasibility of coronary artery calcium
source CT coronary angiography: first experience in a scoring. Korean J Radiol 17(3):321. https://doi.
high pre-test probability population without heart rate org/10.3348/kjr.2016.17.3.321
control. Eur Radiol 16(12):2739–2747. https://doi. Sørgaard MH, Kofoed KF, Linde JJ, George RT, Rochitte
org/10.1007/s00330-­006-­0474-­0 CE, Feuchtner G, Lima JA, Abdulla J (2016) Diagnostic
Scheske JA, O’Brien JM, Earls JP, Min JK, LaBounty accuracy of static CT perfusion for the detection of
TM, Cury RC, Lee T-Y, So A, Hague CJ, Al-Hassan D, myocardial ischemia. A systematic review and meta-­
Kuriyabashi S, Dowe DA, Leipsic JA (2013) Coronary analysis. J Cardiovasc Comput Tomogr 10(6):450–
artery imaging with single-source rapid kilovolt peak– 457. https://doi.org/10.1016/j.jcct.2016.09.003
switching dual-energy CT.  Radiology 268(3):702– Stehli J, Fuchs TA, Singer A, Bull S, Clerc OF, Possner M,
709. https://doi.org/10.1148/radiol.13121901 Gaemperli O, Buechel RR, Kaufmann PA (2015) First
Schoepf UJ (2019) CT of the heart. Humana Press, Totowa experience with single-source, dual-energy CCTA for
Schoepf UJ, Emrich T (2021) A brave new world: toward monochromatic stent imaging. Eur Heart J Cardiovasc
precision phenotyping and understanding of coro- Imaging 16(5):507–512. https://doi.org/10.1093/ehjci/
nary artery disease using radiomics plaque analysis. jeu282
Radiology 299(1):107–108. https://doi.org/10.1148/ Stehli J, Clerc OF, Fuchs TA, Possner M, Gräni C, Benz
radiol.2021204456 DC, Buechel RR, Kaufmann PA (2016) Impact of
Schwartz FR, Koweek LM, Nørgaard BL (2019) Current monochromatic coronary computed tomography
evidence in cardiothoracic imaging: computed angiography from single-source dual-energy CT on
tomography-­derived fractional flow reserve in stable coronary stenosis quantification. J Cardiovasc Comput
Clinical Applications in Cardiac Imaging 161

Tomogr 10(2):135–140. https://doi.org/10.1016/j. Tsai YS, Chen JS, Wang CK, Lu CH, Cheng CN, Kuo
jcct.2015.12.008 CS, Liu YS, Tsai HM (2014) Quantitative assessment
Suchá D, Symersky P, Tanis W, Mali WP, Leiner T, van of iron in heart and liver phantoms using dual-energy
Herwerden LA, Budde RP (2015) Multimodality computed tomography. Exp Ther Med 8(3):907–912.
imaging assessment of prosthetic heart valves. https://doi.org/10.3892/etm.2014.1813
Circ Cardiovasc Imaging 8(9):e003703. https://doi. Valenti V, ó Hartaigh B, Heo R, Cho I, Schulman-
org/10.1161/circimaging.115.003703 Marcus J, Gransar H, Truong QA, Shaw LJ, Knapper
Symons R, Choi Y, Cork TE, Ahlman MA, Mallek M, J, Kelkar AA, Sandesara P, Lin FY, Sciarretta S,
Bluemke DA, Sandfort V (2018a) Optimized energy of Chang H-J, Callister TQ, Min JK (2015) A 15-year
spectral coronary CT angiography for coronary plaque warranty period for asymptomatic individuals with-
detection and quantification. J Cardiovasc Comput out coronary artery calcium. JACC Cardiovasc
Tomogr 12(2):108–114. https://doi.org/10.1016/j. Imaging 8(8):900–909. https://doi.org/10.1016/j.
jcct.2018.01.006 jcmg.2015.01.025
Symons R, De Bruecker Y, Roosen J, Van Camp L, Cork van Assen M, De Cecco CN, Sahbaee P, Eid MH, Griffith
TE, Kappler S, Ulzheimer S, Sandfort V, Bluemke LP, Bauer MJ, Savage RH, Varga-Szemes A, Oudkerk
DA, Pourmorteza A (2018b) Quarter-millimeter M, Vliegenthart R, Schoepf UJ (2019a) Feasibility of
spectral coronary stent imaging with photon-­ extracellular volume quantification using dual-energy
counting CT: initial experience. J Cardiovasc Comput CT.  J Cardiovasc Comput Tomogr 13(1):81–84.
Tomogr 12(6):509–515. https://doi.org/10.1016/j. https://doi.org/10.1016/j.jcct.2018.10.011
jcct.2018.10.008 van Assen M, Lavra F, Schoepf UJ, Jacobs BE, Williams
Symons R, Sandfort V, Mallek M, Ulzheimer S, BT, Thompson ZM, Varga-Szemes A, Ruzsics B,
Pourmorteza A (2019) Coronary artery calcium scor- Oudkerk M, Vliegenthart R, De Cecco CN (2019b)
ing with photon-counting CT: first in  vivo human Iodine quantification based on rest / stress perfusion
experience. Int J Cardiovasc Imaging 35(4):733–739. dual energy CT to differentiate ischemic, infarcted
https://doi.org/10.1007/s10554-­018-­1499-­6 and normal myocardium. Eur J Radiol 112:136–143.
Tanami Y, Ikeda E, Jinzaki M, Satoh K, Nishiwaki Y, https://doi.org/10.1016/j.ejrad.2019.01.017
Yamada M, Okada Y, Kuribayashi S (2010) Computed Van Hedent S, Große Hokamp N, Kessner R, Gilkeson R,
tomographic attenuation value of coronary atheroscle- Ros PR, Gupta A (2018) Effect of virtual monoener-
rotic plaques with different tube voltage: an ex  vivo getic images from spectral detector computed tomog-
study. J Comput Assist Tomogr 34(1):58–63. https:// raphy on coronary calcium blooming. J Comput Assist
doi.org/10.1097/RCT.0b013e3181b66c41 Tomogr 42(6):912–918. https://doi.org/10.1097/
Taylor AJ, Cerqueira M, Hodgson JM, Mark D, Min J, RCT.0000000000000811
O’Gara P, Rubin GD (2010) ACCF/SCCT/ACR/AHA/ Wang R, Yu W, Wang Y, He Y, Yang L, Bi T, Jiao J, Wang
ASE/ASNC/NASCI/SCAI/SCMR 2010 appropriate Q, Chi L, Yu Y, Zhang Z (2011) Incremental value
use criteria for cardiac computed tomography. A report of dual-energy CT to coronary CT angiography for
of the American College of Cardiology Foundation the detection of significant coronary stenosis: com-
Appropriate Use Criteria Task Force, the Society of parison with quantitative coronary angiography and
Cardiovascular Computed Tomography, the American single photon emission computed tomography. Int
College of Radiology, the American Heart Association, J Cardiovasc Imaging 27(5):647–656. https://doi.
the American Society of Echocardiography, the org/10.1007/s10554-­011-­9881-­7
American Society of Nuclear Cardiology, the North Wang X, Zamyatin A, Shi D (2012) Dose reduction poten-
American Society for Cardiovascular Imaging, tial with photon counting computed tomography. In:
the Society for Cardiovascular Angiography and Medical imaging 2012: physics of medical imag-
Interventions, and the Society for Cardiovascular ing. International Society for Optics and Photonics,
Magnetic Resonance. J Cardiovasc Comput Tomogr p 831349
4(6):407.e401–433. https://doi.org/10.1016/j. Wang Y, Yu M, Wang M, Wang Y, Kong L, Yi Y, Wang
jcct.2010.11.001 M, Li Y, Jin Z (2019) Application of artificial intel-
Tesche C, Gray HN (2020) Machine learning and deep ligence–based image optimization for computed
neural networks applications in coronary flow assess- tomography angiography of the aorta with low tube
ment: the case of computed tomography fractional voltage and reduced contrast medium volume. J Thorac
flow reserve. J Thorac Imaging 35:S66–S71. https:// Imaging 34(6):393–399. https://doi.org/10.1097/
doi.org/10.1097/rti.0000000000000483 rti.0000000000000438
Tops LF, Wood DA, Delgado V, Schuijf JD, Mayo JR, Williams MC, Moss AJ, Dweck M, Adamson PD, Alam
Pasupati S, Lamers FP, van der Wall EE, Schalij MJ, S, Hunter A, Shah ASV, Pawade T, Weir-McCall JR,
Webb JG, Bax JJ (2008) Noninvasive evaluation of Roditi G, van Beek EJR, Newby DE, Nicol ED (2019)
the aortic root with multislice computed tomography Coronary artery plaque characteristics associated with
implications for transcatheter aortic valve replace- adverse outcomes in the SCOT-HEART study. J Am
ment. JACC Cardiovasc Imaging 1(3):321–330. Coll Cardiol 73(3):291–301. https://doi.org/10.1016/j.
https://doi.org/10.1016/j.jcmg.2007.12.006 jacc.2018.10.066
162 B. Yacoub et al.

Yamada Y, Jinzaki M, Okamura T, Yamada M, Tanami dual-energy CT. Med Phys 38(12):6371–6379. https://
Y, Abe T, Kuribayashi S (2014) Feasibility of coro- doi.org/10.1118/1.3658568
nary artery calcium scoring on virtual unenhanced Yu L, Leng S, McCollough CH (2012) Dual-energy CT–
images derived from single-source fast kVp-­ based monochromatic imaging. Am J Roentgenol
switching dual-energy coronary CT angiography. J 199(5_supplement):S9–S15. https://doi.org/10.2214/
Cardiovasc Comput Tomogr 8(5):391–400. https:// AJR.12.9121
doi.org/10.1016/j.jcct.2014.08.005 Yunaga H, Ohta Y, Kaetsu Y, Kitao S, Watanabe T, Furuse
Yan RT, Miller JM, Rochitte CE, Dewey M, Niinuma Y, Yamamoto K, Ogawa T (2017) Diagnostic per-
H, Clouse ME, Vavere AL, Brinker J, Lima JAC, formance of calcification-suppressed coronary CT
Arbab-Zadeh A (2013) Predictors of inaccurate coro- angiography using rapid kilovolt-switching dual-­
nary arterial stenosis assessment by CT angiography. energy CT.  Eur Radiol 27(7):2794–2801. https://doi.
JACC Cardiovasc Imaging 6(9):963–972. https://doi. org/10.1007/s00330-­016-­4675-­x
org/10.1016/j.jcmg.2013.02.011 Zachrisson H, Engström E, Engvall J, Wigström L,
Yi Y, Zhao X-M, Wu R-Z, Wang Y, Vembar M, Jin Z-Y, Smedby Ö, Persson A (2010) Soft tissue discrimina-
Wang Y-N (2019) Low dose and low contrast medium tion ex vivo by dual energy computed tomography. Eur
coronary CT angiography using dual-layer spectral J Radiol 75(2):e124–e128. https://doi.org/10.1016/j.
detector CT: low keV versus low kVp. Int Heart J ejrad.2010.02.001
60(3):608–617. https://doi.org/10.1536/ihj.18-­340 Zhang S, Levin DC, Halpern EJ, Fischman D, Savage M,
Young PM, Foley TA, Araoz PA, Williamson EE Walinsky P (2008) Accuracy of MDCT in assessing
(2019) Computed tomography imaging of cardiac the degree of stenosis caused by calcified coronary
masses. Radiol Clin N Am 57(1):75–84. https://doi. artery plaques. Am J Roentgenol 191(6):1676–1683.
org/10.1016/j.rcl.2018.08.002 https://doi.org/10.2214/AJR.07.4026
Yu L, Christner JA, Leng S, Wang J, Fletcher JG, Zou Y, Silver MD (2009) Elimination of blooming arti-
McCollough CH (2011) Virtual monochromatic imag- facts off stents by dual energy CT. In: Samei E, Hsieh J
ing in dual-source dual-energy CT: radiation dose and (eds) SPIE medical imaging, 2009, Lake Buena Vista,
image quality: monochromatic imaging in dual-source FL, vol 72581X. https://doi.org/10.1117/12.811696
 Dual-Energy CT Angiography

Matthias Stefan May and Armin Muttke

Contents
1 Dual-Energy CT Angiography   163
1.1  Image Contrast   163
1.2  Contrast Agent   165
1.3  Virtual Unenhanced   166
1.4  Plaques   167
1.5  Clinical Applications   169
References   175

1 Dual-Energy CT Angiography general technical considerations of Dual-Energy

CT Angiography and then evaluate the respective
Successful CT imaging of the vascular system clinical applications.
requires contrast media injection. High vessel
contrast is vital to obtain high diagnostic image
quality. Besides optimized injection protocols, 1.1 Image Contrast
the energy level of the X-ray source has a high
impact on the imaging results. A significant limi- The image quality from DE examinations is
tation of CT angiography is the limited evaluabil- superior to conventional acquisitions at 120 kV
ity in highly calcified plaque burden. tube voltage potential if the low and the high
Unfortunately, low photon energy levels further energy dataset are blended to obtain polychro-
decrease the evaluability of the vessel lumen in matic images (PI). Non-linear blending tech-
precisely these patients because of increased niques provide the best results with the highest
blooming artifacts. Dual-energy (DE) is a prom- contrast to noise ratios. Modified sigmoid func-
ising tool to overcome this dilemma by simulta- tions provide the optimum contrast from this
neously addressing image contrast and material raw data (Li et  al. 2014). Ahead of blending
decomposition. This chapter will first review the techniques, extrapolating the simulated energy
levels with virtual monoenergetic image recon-
structions (VMI) can further increase image
M. S. May (*) · A. Muttke contrast. The high vessel contrast from low-
Department of Radiology, University Hospital energy reconstructions can achieve two primary
Erlangen, Erlangen, Germany goals: to improve the diagnostic performance
e-mail: [email protected]

© The Author(s), under exclusive license to Springer Nature Switzerland AG 2022 163
H. Alkadhi et al. (eds.), Spectral Imaging, Medical Radiology Diagnostic Imaging,
https://doi.org/10.1007/978-3-030-96285-2_11
164 M. S. May and A. Muttke

or to reduce the contrast agent dose. Also, a However, the additional flexibility to apply
combination of both effects can be helpful, various other reconstruction types to DE data
depending on the clinical situation and the favors DE acquisitions over SE techniques. In
intention for imaging. The lowest available some studies, moderately low monoenergetic
VMI levels provide the highest vessel contrast reconstructions between 50 and 70 keV had bet-
(Fig.  1). Forty kiloelectron Volt (keV) is the ter subjective image quality than 40 keV despite
value that approaches the K-edge of iodine the lower objective image quality in CNR (Martin
closest (Yoshida et al. 2020). Vessel reproduc- et al. 2017a). So, the implementation of very low
tion in these images and contrast to noise ratio (40 keV) or moderately low reconstructions (50–
(CNR) is substantially superior to the PI and 70 keV) in the clinical routine setting seems more
conventional Single Energy (SE) acquisitions at a matter of individual taste than evidence-based
120 kVp (Murphy et  al. 2019). The objective medicine. Early reconstruction algorithms suf-
vessel image quality is also superior to low tube fered from a substantial increase in image noise,
voltage acquisitions, but subjective and overall especially in the very low VMI. The next genera-
image quality is favorable in the low kV images tion, the so-called advanced monoenergetic
(Yoshida et al. 2020). Therefore, low tube volt- reconstructions (VMI+), solved these drawbacks
ages could always be a good selection if DE is using frequency selective techniques. These
unavailable. advanced algorithms do not affect the image con-

Fig. 1  Visualization of dual-energy CT angiographies: Very low virtual monoenergetic reconstructions (40 keV)
maximum intensity projections (MIP) suffer from bone provide the best vessel contrast. Improved quality of
and calcified plaque superposition. Dual-energy can plaque removal techniques (PR) helps to achieve high
reduce the error rates of automated bone removal (aBR) diagnostic accuracies (right superficial femoral artery ste-
compared to conventional techniques. Therefore, the nosis in this case) but often require manual adaptations
required time for manual post-processing to obtain over- before a comprehensive demonstration in MIP or volume
view images with bone removal (BR) is also reduced. rendering techniques (VRT) is possible
Dual-Energy CT Angiography 165

trast but significantly reduce the image noise. ment of smaller vessel branches in obese
This effect substantially improves the CNR in the patients. Therefore, a high power capacity of
40 keV images (Riffel et al. 2016). the X-ray tube is required to obtain DE datasets
Conventional window settings from standard in a large collective. This need is also essential
SE acquisitions are inappropriate for the ves- to increase the radiation dose efficiency by
sels’ very high attenuation values in these con- reduction of the tube voltage. Savings up to
trast optimized VMI, with up to far more than 46% at constant CNR are possible by voltage
1000 HU. Some studies evaluated the subjective reduction of the main tube from 100 kV to 80
and calculated optimum window setting kV in Dual Source scanners (Li et  al. 2017).
(D’Angelo et  al. 2018, 2020; Caruso et  al. Automated attenuation-­based kV pair selection
2017). The suggested values differed between may be an attractive technique for future scan-
different vessel regions and injection protocols. ner generations to improve the individualiza-
With mean attenuation values around 1000 HU tion of vascular DE protocols (Renapurkar
in the carotids, width values (W) around 1600 et  al. 2017). Another way to overcome image
HU, and level values (L) around 550 HU seemed noise problems in DE angiography is the com-
most appropriate. This combination was higher bination with iterative reconstructions. The
than recommended evaluating the pulmonary effects are comparable to the impact on SE data
arteries (mean attenuation around 650 HU, sug- (Lee et al. 2016).
gested W/L = 1100/400 HU). However, two
independent studies about the abdominal arter-
ies suggested different window levels, such as 1.2 Contrast Agent
W/L = 1400/450 HU for a mean attenuation
around 700 HU and W/L = 1600/700 for a mean Generally, the lowest dose of contrast medium
attenuation of 1100 HU. This roughly copies the consistent with a diagnostic result for all patients
recommendations in the literature for attenua- should be favored (van der Molen et  al. 2018).
tion-adapted adjustments of the window settings The limit to not exceed the volume injection in
for conventional acquisition techniques (Saba mL above the threefold glomerular filtration rate
and Mallarin 2009). Thus, presets could support may be challenging in patients with impaired
the implementation into the routine workflow kidney function. DE can help to provide diagnos-
and adapt to the local technique of contrast tic vascular image quality, also in very low con-
injection and timing. trast agent dose protocols. The range of reported
The image quality of VMI reconstructions reduction rates is vast (28–75%) (Xin et al. 2015;
seems to be relatively stable to different acqui- Meier et al. 2016, 2017). Different study designs
sition techniques. All methods to generate DE can explain this discrepancy. Rather conservative
datasets follow a common trend, independent trials reported a simultaneously increased objec-
from the vendor and the scanner generation. tive image quality. Some of the more aggressive
Different acquisition speeds and tube power studies used unequal intensities of iterative
mainly account for potential limitations and reconstructions to obtain similar objective image
differences. Time gaps between the high- and quality at matched radiation dose (Shuman et al.
the low-energy exposure may induce artifacts 2017). Lowest reported total volume is 15 mL
based on the movement of structures, like the and the lowest total iodine dose is 6 g for imaging
heart or the digestive tract, in the meantime. of the pulmonary arteries, and 28 mL or 9.8 g for
These artifacts are most pronounced in the very the Aorta (Meier et  al. 2016, 2017; Hou et  al.
low VMI and can be neglected in the PI (Eichler 2017; Carrascosa et  al. 2014). The authors rec-
et  al. 2020; May et  al. 2019a). Luckily, the ommend adapting the bolus dynamics in these
impact on vessel evaluation is shallow in most protocols by diluting the contrast agent using
regions beyond the ascending Aorta. Increased dual head injectors or choosing contrast agents
image noise may hamper especially the assess- with lower iodine concentrations. The combina-
166 M. S. May and A. Muttke

tion of 60% contrast dose reduction and VMI at scope of CT angiograms, and virtual unenhanced
50–60 keV was reported to best match a 120 kV vessels and tissues on the other side. The use of
SE acquisition in a systematic dilution series over these virtual unenhanced images to replace
different collectives (Carrascosa et  al. 2014). actual unenhanced images may help to reduce
Overall, most clinical trials outweighed the the radiation dose for the patient but is discussed
expectations from phantom trials (40% reduc- controversially in the literature. Some authors
tion), probably due to an additional effect of the reported significant differences in objective and
systemic circulation, dilution processes, and subjective image quality compared to the refer-
bolus dynamics in  vivo. On the other side, DE ence (Lehti et al. 2018). Especially the very high
acquisitions can help overcome these uncertain- iodine concentrations in the large vessels pose
ties of unpredictable image contrast levels in a problems. The virtual unenhanced datasets con-
clinical collective. Individual retrospective selec- tain significantly increased attenuation values
tion of the VMI level, based on the mean attenu- and significantly increased image noise in these
ation values, could help to standardize the image locations (Pinho et  al. 2013). Therefore, late
quality in the respective collective (Hou et  al. arterial or portal-venous phase images are more
2017). appropriate for these reconstructions in multi-
Gadolinium-containing contrast agents may phase studies than the arteriograms (Lehti et al.
serve as an alternative in case of contraindica- 2019). The problem is also more pronounced in
tions for iodine-containing contrast agents. Few the thoracic aorta, and pulsation artifacts may
authors described this feasibility in one animal further hamper the evaluation. On the contrary,
trial and one single translational clinical trial. PI the image quality of virtual unenhanced data in
with 150% and 250% gadopentetate dimeglu- the abdominal aorta seems suitable for a clinical
mine injection in rabbits provided a sound diagnosis. So, patients prone to radiation dose
reproduction of healthy and obstructed pulmo- can benefit from reduced phase acquisitions
nary vessels. The clinical examinations used (Shaida et al. 2012).
single-­dose injections (Zhang et  al. 2011).
Standard reconstructions could not provide 1.3.1 Bones
diagnostic image quality in the aortic system in One of MR angiography’s primary advantages
this setting in humans, but VMI allowed for over CT angiography is the easily obtained maxi-
diagnostic data. Like with iodine, 40 keV has mum intensity projection overview of the exam-
the highest impact on vessel contrast and is best ined vessels. These are very useful for
suitable for patients undergoing iodine-free demonstrations and quick identification of the
contrast-enhanced CT angiography. This result regions requiring intensified thin-slice workup.
is astonishing regarding the K-edge of Threshold-based bone removal of SE CT data
Gadolinium at 50.2 keV. Here, the extrapolation requires intensive manual corrections and, there-
algorithm in VMI detaches the image contrast fore, high post-processing times. Error rates of
from the physical principles of X-ray absorption automated bone removal algorithms without
(Nadjiri et al. 2018). manual corrections are high. Literature reports
that up to more than 90% of the captured bones
may be missed (Morhard et al. 2009). DE prom-
1.3 Virtual Unenhanced ises to overcome this problem because the mate-
rial decomposition technique can specifically
DE data also allows for material decomposition. detect calcium. Automated bone removal recon-
This technique can separate the iodine-­containing structions are available and tested for a variety of
contrast agents from the tissues. The resulting regions. All of the studies found substantially
images display iodine concentration quantita- reduced missing rates of bone detection ranging
tively on one side, which is not so much in the from 0 to 46% (Fig.  1) (Sommer et  al. 2009;
Dual-Energy CT Angiography 167

Meyer et al. 2008). Most problems occur with the tion (Jin et al. 2017). The overall underestima-
ribcage (error rate 46%) and the patella (error tion of the proper lumen area can still be up to
rate 25%). Also, the head and neck region remains 50% in stenotic segments. However, these
challenging, with reported error rates of 12% results are much better than those from SE eval-
(Morhard et al. 2009). Supportive metal artifact uation of stenosis by calcified plaques, severely
reconstructions before automated DE bone suffering from blooming artifacts (Li et  al.
removal are beneficial in around 60% of all cases 2020). A toxic combination of conditions with
and should be used routinely for this region small vessels, poor lumen contrast, and severe
(Kaemmerer et al. 2016). Primary error rates for calcifications is often present below the knee.
all other regions are below 5%. The resulting Poor diagnostic results for this body region have
time effort for manual corrections is also signifi- been described (Meyer et  al. 2008). Notably,
cantly lower than with the threshold method and algorithms for calcified plaque removal
can be consistently around 2–4 min, compared to improved over time, with the most recently pub-
4–6 min. The resulting vessel contrast from DE lished diagnostic accuracy up to 96.5%. The
studies was simultaneously better than in the SE increased specificity and negative predictive
acquisitions (Schulz et  al. 2012). However, in value contribute to the superiority over conven-
severe vessel calcifications, new vessel irregular- tional evaluation without plaque removal
ities may occur from the DE bone removal algo- (Fig. 1) (de Santis et al. 2019). Non-­significant
rithms (Yamamoto et  al. 2009). This artifact differences to invasive digital subtraction angi-
should be known and respected during post-­ ography are possible. The immediate success of
processing and interpretation. Higher tube volt- automated plaque removal can be around
ages during acquisition may reduce these artificial 40–75%. Still, about one-fourth of all datasets
vessel erosions in Dual Source DE examinations can suffer from insufficiently high residual cal-
of the supra-aortic arteries (Korn et  al. 2015). cifications, limiting the overview visualization
This principle especially accounts for the lower as maximum intensity projection. There are no
segments, the internal carotid artery, and the V3/4 reported negative influences of the plaque
segments of the vertebral artery, which remains removal algorithms on the overall image quality
problematic. (Mannil et al. 2017).
Another critical issue in plaque imaging is the
influence of different VMI reconstructions. It is
1.4 Plaques well known that lower tube voltages in SE acqui-
sitions come along with increased vessel con-
Calcified plaques pose another main difficulty of trast, but with the drawback of higher blooming
overview images derived from CT angiography artifacts (Grimes et  al. 2015). Low keV VMI
compared to MR.  Automated calcified plaque from DE examinations of the carotids copied
removal by material decomposition is required to this effect in an evaluation with fixed window
derive a true luminogram and avoid unevaluable levels (width 850 HU, level 300 HU). The mean
segments. Systematic phantom trials found that effect was as high as a 30% increase of the grade
in vessels with a diameter equal to or above 5 of diameter stenosis in 66 keV VMI compared to
mm, DE with automated plaque removal recon- 86 keV (Fig. 2) (Saba et al. 2019). Conversely, in
struction can provide reliable results. Vessels an evaluation with adjustments of the window
below 2  mm may be deleted entirely from the level according to the reader’s discretion for
images in case of severe calcifications. Other each case, no influence on the total plaque bur-
negatively influencing factors for plaque removal den, the vessel diameter, or the reported grade of
are low iodine concentration in the lumen and stenosis was found (Bai et al. 2020). This finding
high density of the plaques. further emphasizes the need for adjustments of
Interestingly, the calcium plaque thickness the window level for the interpretation of VMI. A
does not influence the vessel lumen visualiza- considerable additional effect is an improved
168 M. S. May and A. Muttke

Fig. 2  Flexibility of dual-energy reconstructions: bone virtual monoenergetic reconstructions allow for reduced
removal (BR) and plaque removal (PR) support a compre- blooming artifacts (86 keV) or increased vessel contrast
hensive demonstration of carotid stenosis using volume (40 keV)
rendering techniques (VRT). Different energy levels of

contrast in low VMI that can support the differ- sible by calculation of the mean atomic number
entiation of the plaque composition, such as a (Z). Low Z-values inversely correlate with the
fibrous cap, large lipid core, and intraplaque areas of fibro-fatty components in vulnerable
hemorrhage. Also, it is easier to delineate non- plaques derived from virtual histology intravas-
calcified plaque margins from the surrounding cular ultrasound (Shinohara et  al. 2015).
adipose tissue and delineate calcified plaque Therefore, DE provides a high convenience for
margins from the contrast agent-filled lumen at subjective plaque evaluation in general and an
the same time (Reynoso et  al. 2017). objective quantification tool to identify patients
Quantification of the plaque c­ omposition is fea- at risk simultaneously (Fig. 3).
Dual-Energy CT Angiography 169

Fig. 3  Plaque imaging with dual-energy: different calci- contrast of the lumen stenosis and the delineation of the
fied and non-calcified plaque composition areas can be plaque from the surrounding adipose tissue are increased.
better differentiated by increased contrast from low virtual Z-value and electron density (Rho/Z) maps allow for
monoenergetic reconstructions (40 keV) compared to the plaque quantification
reference polychromatic images (PI). Simultaneously the

1.5 Clinical Applications Therefore, contrast improvements are often

required, and retrospective options are frequently
1.5.1 Pulmonary Angiography welcome. More than 90% of retrospectively
Pulmonary angiography is performed frequently selected, non-diagnostic pulmonary angiograms
in the emergency setting and remains one of the with attenuation values below 200 HU could be
radiology enigmas. No matter how well one pre- salvaged by VMI reconstructions compared to
pares the examination, how intense one calcu- the PI (Ghandour et al. 2018). In particular, the
lates circulation times, or how much contrast one sensitivity increases substantially from 80% to
injects, sometimes the vessel contrast is inferior. 92% and adds to the increased diagnostic perfor-
Venous capacity, collateral flow, cardiac func- mance. The specificity is as high as 94–97% in
tion, intravascular volume, and last but not that situation (Bae et  al. 2018). Other studies
least—alternating intra-thoracic pressures by reported the sensitivity and specificity for detect-
forced respiration may influence the bolus ing pulmonary embolism in portal-venous phase
dynamics in pulmonary angiograms. Abrupt images as high as 90/100%. The subjectively
breathing commands and well-behaved patients selected monoenergetic level was 48.5 keV (Foti
taking deep breaths can have devastating effects, et  al. 2021). This approach appears to be an
even if the bolus is detected automatically in the attractive bail-out strategy in oncologic staging
target vessel. This failure becomes incredibly dis- patients, where incidental pulmonary embolism
ruptive for the evaluation of the smaller vessels. is as frequent as 4% and overlooked in about one-­
170 M. S. May and A. Muttke

Fig. 4  Enhanced iodine contrast: The bilateral pulmonary embolism in portal-venous phase staging CT could easily be
overlooked. Low kiloelectron voltage reconstructions, 40 keV in this case, help to delineate the thrombi

fourth of cases (Fig. 4) (Gladish et al. 2006). The (Meyer et  al. 2018). In contrast, radiation dose
impact of DE, mainly due to the detection of reduction by more than 50% is possible for preg-
smaller subsegmental embolism, was indicated to nant and postpartum women. The rate of images
double the detection rate (Uhrig et al. 2018). The with limited or non-diagnostic quality was simul-
best visualization of the relatively small bron- taneously reduced from 37% to 10% using DE
chial arteries was found at 63 keV in pulmonary (McDermott et  al. 2018). Notably, the positive
angiograms with bolus detection in the descend- effect on the diagnostic image quality is not lim-
ing Aorta (Ma et  al. 2016a). However, also in ited to radiologists. The higher contrast levels of
regular pulmonary embolism examinations, low VMI improve the accuracy of computer-aided
VMI seem to support the reader. The best diag- detection (CAD) algorithms, especially in insuf-
nostic performance (>96%) was reported for 60 ficient contrast levels with the standard recon-
keV images, while 40 keV had the highest sub- structions (Kröger et  al. 2017). In a systematic
jective diagnostic confidence (Sauter et al. 2020). evaluation, between 40 and 80 keV, intermediate
More is probably not always better in DE, even if low energies of 65 keV appear to be the best
the differences were only subtle. Advanced VMI+ trade-off between high sensitivity (84%) and a
are potentially able to overcome this non-linear low rate of false positives (12%) (Ma et al. 2019).
pattern by further lowering the image noise. This compromise may underline the potential of
Some authors found that this technique could DE images to adapt the image quality to the indi-
lower artifacts in the pulmonary arteries, but vidual. Future concepts certainly need to com-
structured performance evaluation compared to bine all available technologies, from physics over
the preceding generation is missing in the litera- medical history to artificial intelligence, to
ture (Meier et  al. 2015). The increased contrast achieve the best results to support clinicians and
allows for substantial contrast agent dose reduc- improve the patients’ outcomes (Huang et  al.
tions and comes along with only a moderate 2020).
increase of image noise. Therefore, substantial
radiation dose reductions are also feasible with 1.5.2 Carotid Angiography
this advanced reconstruction technique (Petritsch Angiography of the carotid arteries is critical to
et al. 2017). A hundred percent diagnostic accu- read due to two disturbing factors: calcified
racy was found for a protocol with 83% iodine plaques and bones. The calcifications in the
dose reduction (5.4 g instead of 32 g) in a collec- carotid sinus and the proximal internal carotid
tive with acute or chronic renal insufficiency artery can be vast and dense. Lumen evaluation
Dual-Energy CT Angiography 171

can be challenging in these cases on one side. carotid artery (Zopfs et al. 2018). The advanced
However, there is strong evidence in the literature methods for bone and plaque removal addition-
about the carotid plaque composition as a risk ally promise to ease up the cumbersome report-
factor on the other side. Radiologists should ing process in the vicinity of dense bones or
therefore love and hate carotid plaques at the severe vessel wall calcifications. Sensitivity and
same time. Hypodense plaques are associated specificity of 100% and 92% to detect hemody-
with embolic cerebrovascular stroke and are namically relevant (>70%) stenosis can be
therefore considered vulnerable (Mikail et  al. achieved with combined bone and hard plaque
2021). Microcalcifications are known to be asso- removal (Fig.  2) (Uotani et  al. 2009)
ciated with plaque instability. Paradoxically, Overestimation of the grade of stenosis in maxi-
macrocalcifications seem to stabilize the plaque mum intensity projections of the carotid bifurca-
(Montanaro et al. 2021). CT angiography is the tion after plaque and bone removal was only 6%
only method to assess the plaque composition compared to digital subtraction angiography, but
reliably. DE has proven a good sensitivity of a considerable amount of pseudo-occlusions
100% for detecting calcification, 89% for mixed should be expected (Korn et al. 2011). In general,
plaques, and 85% for low-density fatty plaques vessel reproduction below the skull base seems
correlated with histopathology (Das et al. 2009). superior in DE bone removal techniques com-
Radiologists should be aware that the historically pared to conventional image registration bone
established threshold levels (130 HU) for assess- subtraction. However, disadvantages for the
ing the calcified plaque volume may be inappro- intra-osseous and intracranial sections were
priate in DE reconstructions (Agatston et  al. found (Lell et al. 2009). Stenosis grade overesti-
1990). A comparative study with unenhanced SE mation especially happens in the course of the
acquisitions reported that the conventional calcu- internal carotid artery through the skull base and
lation methods may yield underestimations and in the V3/4 segments of the vertebral artery.
that 180 HU could provide higher consistency Hence, it remains always recommended to review
(Watanabe et  al. 2011). Moreover, VMI can these areas in MPR mode using the images with-
increase the intraplaque contrast and allow a bet- out calcium removal techniques (Buerke et  al.
ter differentiation (Fig. 3) (Reynoso et al. 2017). 2009). Especially the ophthalmic artery was
Studies, for example, found an increasing nega- mentioned in the literature to be usually elimi-
tive correlation of plaque density with the extent nated (Lell et al. 2010).
of leukoaraiosis using DE (Saba et  al. 2017).
Some also speculated that contrast enhancement 1.5.3 Aortic Angiography
of these critical plaques could be assessed by DE A wide variety of indications exists for CT angi-
as well. In general, lower energies can improve ography of the aorta. Acute aortic syndromes and
the objective and subjective image quality of the traumatic injuries of the aorta have a different
cervical, petrous and intracranial vessels com- scope than aneurysm evaluation, treatment plan-
pared to PI. The vascular contrast increases and ning, and surveillance. The aortic root often suf-
significantly improves the assessment of arteries fers from motion artifacts in non-gated
close to the skull base and small arterial branches. examinations, posing a diagnostic problem in the
The lowest energy levels provide the best results emergency setting (Qanadli et  al. 1999). ECG-­
(Neuhaus et  al. 2018). However, the diagnostic gated DE examinations can provide motion-free
performance seems to remain unaffected by VMI images of the coronary arteries. The diagnostic
alone (Leithner et  al. 2018). Objective image image quality in a frequency-controlled collec-
quality comparable to the standard PI from arteri- tive was comparable to SE acquisitions (98.4%
ally triggered examinations is even feasible in vs. 99.1% diagnostic segments), simultaneously
venous phase acquisitions. The slightly increas- increasing CNR (Kerl et al. 2011). Unfortunately,
ing blooming artifacts at lower energy levels do no study has evaluated the DE image quality of
not affect the grade of stenosis in the internal the aortic root in the emergency setting so far,
172 M. S. May and A. Muttke

where higher and probably irregular heart rates maximum extent of this effect. The very high
are frequent. Recently, the visualization of intra- vessel contrast in minimum energy VMI is objec-
mural hematomas of the thoracic aorta by a tively also superior to SE examinations with the
custom-­made two-material decomposition algo- lowest tube voltages of 70 kV. However, the sub-
rithm to obtain dark blood images provided jective overall image quality and also the detec-
appealing results. Subjective image scores were tion rate of endoleaks were slightly inferior
substantially outranging the actual non-contrast (Beeres et  al. 2016). A similar discrepancy was
images for delineation of the healthy vessel wall also reported between the highest objective
as well as for hematomas (Rotzinger et al. 2020). image quality with 40 keV VMI and the highest
Trials that aimed to use DE three-material subjective image quality with 60 keV VMI in the
decomposition techniques for radiation dose pre-interventional workup for transarterial valve
reduction by replacing an unenhanced acquisi- replacements (TAVR). The main reason for this
tion in patients with an acute aortic syndrome difference was the altered image noise texture
reported only limited success for the image qual- (Martin et al. 2017a). Therefore, maximum ves-
ity of virtual non-contrast reconstructions (Lehti sel contrast seems to be not the only limiting fac-
et al. 2018). Disturbing aspects are an increased tor for evaluating the large vessels, and relatively
image noise, incomplete iodine removal from the moderate energy levels may be appropriate in the
vessel lumen with higher attenuation values, clinical setting, depending on the individual taste
elimination of calcifications or stent material, of image impression.
reduced overall subjective image quality, and
motion artifacts in the aortic root. 1.5.4 Abdominal Angiography
Other authors reported the feasibility of vir- The caliber of the abdominal arteries is quickly
tual unenhanced images to differentiate calcifica- decreasing after their point of origin in the
tions from endoleaks in patients after endovascular abdominal aorta. The lowest energy VMI have
aortic replacement (EVAR) (Godoy et al. 2010). the most pronounced advantages for the subjec-
However, the indicated error rate of calcification tive delineation of the small abdominal branches
subtraction in virtual unenhanced reconstructions (Fig. 5) (Albrecht et al. 2016a). Also, a system-
in this scenario was 70% (Sommer et al. 2010). atic review of the liver arteries found the best
Luckily, the detection rate will be unaffected by results at 40 keV (Marin et  al. 2015). These
this technique and will range from 96 to 100% if results were only limited in obese patients, where
applied on venous or late phase acquisitions image noise exceeded the diagnostically accept-
(Buffa et  al. 2014). Some authors recommend able level. The PI reconstructions did not suffer
using a split bolus approach with 40% injection from this limitation. In general, DE can also help
of the total volume 35 s earlier to allow contrast to reduce radiation dose and maintain high image
agent accumulation in the endoleak sac (Javor quality in obese collectives compared to conven-
et al. 2017). The respective radiation dose reduc- tional SE examinations (Liu et al. 2016). Studies
tion is around 40% compared to a biphasic proto- without VMI proved a relevant impact of DE on
col and 60% compared to a triphasic protocol. the hepatic vessel representation compared to
VMI+ reconstructions can obtain the best results conventional 120 kV acquisitions. The delinea-
for the sometimes subtle contrast of endoleaks at tion of the right gastric artery, the vessels of the
40 keV. This advanced reconstruction algorithm liver segment IV, and potential extrahepatic non-­
also has a measurable positive effect on the detec- target vessels for the workup before radioemboli-
tion rate compared to conventional VMI and PI zation is easier with PI (Altenbernd et al. 2015).
(Martin et al. 2017b). The further increase of the Like for the aorta, the best objective image qual-
CNR can be 50–100% compared to classic VMI, ity can be expected in the minimum energy VMI
mainly delivered by image noise reduction. The datasets, too. The CNR can be increased by fac-
lowest energy levels (40/50 keV) and the rela- tor 3 compared to PI (Albrecht et  al. 2016b).
tively high energy levels (100 keV) achieve the However, moderately decreased energy levels,
Dual-Energy CT Angiography 173

Fig. 5  Increased precision: virtual non-contrast (VN) iodine map. The lowest virtual monoenergetic reconstruc-
reconstructions from dual-energy acquisitions of the tions (40 keV) and curved maximum intensity projections
abdomen are comparable to true non-contrast (TN) (cMIP) helped to detect and locate the bleeding and the
images. Additionally, the slight Iodine extravasation from feeding vessel for immediate embolization therapy
a lumbar artery, in this case, is explicitly detected in the (Angio)

like 60–70 keV, provide the abdomen's best over- optimum kiloelectron voltage reconstructions are
all subjective vessel image quality (Gao et  al. used (mean 53 keV). The optimum energy level
2016). Especially motion artifacts from the small for evaluating the tiny branches of the inferior
intestine can further reduce the image quality at epigastric artery before deep-inferior-epigastric-­
40 keV (Eichler et  al. 2020). Several studies perforator (DIEP) flap transplantation was 63
examined the potential of VMI to reduce contrast keV in another study (Gao et al. 2016). Notably,
agent dose in the abdomen. Values between 15 DE was the modality with the highest diagnostic
and 25% were reported without affecting the ves- detection rate of the tiny anterior spinal artery in
sel reproduction of the renal arteries and the abdominal angiography (82%) when compared
splanchnic arteries (Liu et  al. 2014; He et  al. to SE CT angiography (81%), MR angiography
2015). A study about simultaneous arteriovenous (75%), and invasive catheter angiography (60%)
CT angiography of the kidneys was even able to (Abdelbaky et al. 2019).
demonstrate a positive impact on kidney function Abdominal vessel evaluation is also often
in a large collective (Zhou et al. 2017). Patients required in the emergency setting. I assume it
with abnormal renal function, defined as esti- happens to every radiologist at least once in his
mated glomerular filtration rate between 30 and career that he thinks about intestinal ischemia
90 mL/min 1.73  m2, had a significantly lower after the scan of a patient with an acute abdomen
incidence of contrast-induced nephropathy in the was finalized in the portal-venous phase only. In
group with the lower contrast agent concentra- precisely this situation, you would give your
tion. No measurable difference was found for kingdom for a DE exam. VMI can provide con-
patients without reduced renal function before trast values in the superior mesenteric artery from
CT.  The image quality was comparable if the portal-venous acquisitions that can exceed the
174 M. S. May and A. Muttke

contrast values from arterial series (Hickethier role here. The drop in diagnostic accuracy below
et al. 2018). Embolic occlusions should therefore the knee was determined to be around 10% com-
be fully evaluable. Care must be taken during the pared to lesions above in PI images from DE
evaluation of stenosis grade by atherosclerotic acquisitions (Schabel et al. 2015). Other authors
lesions because overestimations can occur at 40 reported even lower accuracies for MIP-only
keV.  Another emergency to trade kingdoms is evaluations with abysmal results for the pedal
acute bleedings, especially if an unenhanced arteries. The respective accuracy drop can be as
examination was forgotten in the examination high as 20–30% for that region, but the results for
protocol. Luckily, the diagnostic performance of bypass segments were comparable to above the
virtual non-contrast reconstructions was non-­ knee (Kau et  al. 2011). Especially the rate of
inferior to actual non-contrast examinations false positives increases in the lower leg
before contrast-enhanced acquisitions in an (Brockmann et al. 2009). Therefore it was recom-
­evaluation in the acute emergency setting (Fig. 5) mended to use the lowest available VMI+ recon-
(Sun et  al. 2015). Moreover, bleeding detection struction for this area, which provides increased
has a higher precision with VMI reconstructions diagnostic accuracy over VMI and PI (Wichmann
from portal-venous examinations than with the et al. 2016). Some authors also claimed diagnos-
respective PI (area under the curve 0.96 vs. 0.82) tic accuracies above 90% by this technique for
(Martin et al. 2017c). stenosis and occlusion of the lower leg. The direct
comparison with conventional CT was also sig-
1.5.5 Peripheral Angiography nificantly superior (Jia et  al. 2020). The lowest
Imaging of the peripheral arteries has to deal with VMI also substantially improved the evaluation
two different challenges. First, the vessel diame- of the very small peroneal artery perforators
ter in the proximal sections is often larger than before autologous transplantation in head and
3  mm and relatively easy to visualize. Some neck surgery. Manual vessel segmentation and
authors found 100% sensitivity and specificity multiplanar unfolding have the potential to sup-
for high-grade stenosis (>75%) in the iliac arter- port this benefit further (May et  al. 2019b). A
ies (Schabel et al. 2015). However, solid calcifi- transfer of this fundamental principle of maxi-
cations or stent material can lead to an mum contrast optimization for evaluating small
overestimated degree of stenosis and limit diag- peripheral arteries to other regions is possible.
nostic accuracy. A trade-off between high iodine The subclavian artery branches and the arterial
contrast and blooming artifacts seems appropri- supply to the nipple-areola-complex before
ate for these vessels. It seems that 90 keV VMI nipple-­sparing mastectomy in case of breast can-
are appropriate for stent lumen visualization, cer are just one example in the literature (Zeng
while 40 keV VMI provide the best contrast and et al. 2020).
100 keV VMI the most extensive in-stent lumen Workflow is another crucial issue in CT angi-
reproduction (Zhang et  al. 2020). VMI+ come ography of the peripheral arteries. The long scan
along with slightly lower optimum energy results. range and the transversal physics of CT scanners
Here, moderate energies of 70–80 keV provide ensure high diagnostic image quality for evaluat-
the best in-stent lumen reproduction directly ing the vessel diameter in general. However, it is
compared with the gold standard from invasive time-consuming to evaluate a large stack of thin-­
catheter digital subtraction angiography slice axial images. Unfortunately, the appropriate-
(Mangold et al. 2016). These VMI+ significantly ness for coronal overview images is very limited,
improved the diagnostic accuracy compared to PI especially if compared with MR angiography,
(96% vs. 89%). because of the superposition of bones and calci-
Second, the small vessels in the periphery of fied plaques. Thus, especially workflow issues
the forearm and the lower leg can be challenging need to be addressed in CT of the peripheral ves-
to capture on one side and difficult to evaluate on sels. Evidence exists in the literature that imple-
the other side. Image contrast plays an important menting bone and plaque removal algorithms
Dual-Energy CT Angiography 175

based on DE acquisitions can provide coronal 2020). Also, adrenal vein delineation, before
overview images with high sensitivities (94–98%, catheter-based adrenal vein sampling in patients
Fig. 1) (Klink et al. 2017). The required time for with primary aldosteronism, benefits from high
post-processing DE datasets seems affordable CNR and subjective image quality (Nakayama
(1–2 min) and roughly 50% below the duration of et  al. 2020). The detection rate was 100% and
conventional techniques (Brockmann et al. 2009). improved compared to the 92% reported in the
There was no relevant difference in the literature literature (Onozawa et  al. 2016). The sampling
than for conventional MPR, MIP, curved MPR, success rate with DE was 95% and comparable to
and automated assessment, making it a reasonable the results from other studies. Contrast agent
method for rapid assessment and demonstration. concentration in a portal-venous system is even
However, specificity remains the Achilles' heel of more insufficient. Early studies found that non-­
this technique (67–75%) and can be below 50% in linear blending in PI increases the rate of patients
case of severe calcifications. Therefore, additional with visible 5th to 6th generation branches of the
detailed evaluation of questionable sections in portal vein from 36% to 76% (Wang et al. 2013).
MPR mode is recommended in a two-part work- These images were also better than VMI for a
flow to achieve acceptable specificities (92%) combined evaluation of the portal and hepatic
(Kosmala et al. 2020). Also, a direct comparison veins (Wang et  al. 2014). However, low-energy
of two commercially available products claimed a VMI+ were recommended in deplorable contrast
strong dependence of the diagnostic accuracy conditions, such as late phase imaging or cir-
from the vessel segmentation and bone removal rhotic patients, to visualize the venous systems in
algorithm (Kostrzewa et  al. 2016). Future soft- the liver (Schabel et  al. 2014). Also, it was
ware may hopefully further improve the workflow reported that low VMI (50–60 keV) could help to
and diagnostic performance. reduce the contrast agent dose by 25–52% (Han
et al. 2019; Ma et al. 2016b). In general, enhanced
1.5.6 Venous System iodine contrast in the venous system can help
Venography often suffers from poor contrast due obtain diagnostic image quality in conventionally
to the free diffusion of contrast agents across the unevaluable cases. Especially the retrospective,
blood–tissue barrier and low Iodine concentra- spontaneous application of contrast-enhancing
tion in the venous backflow, especially compared algorithms, like VMI+, to portal venous or late
to arteriography. So it is hardly surprising that phase datasets can direct the radiologists’ deci-
some studies in the literature targeted the contrast sion and increase their confidence level.
improvements by low kiloelectron voltage VMI
from DE acquisitions (Kulkarni et al. 2012). For Compliance with Ethical Standards
example, 40–50 keV VMI provide the best CNR
in the pelvic veins, and the diagnostic perfor- Disclosure of Interests  All authors declare they have no
mance and confidence for detection of deep conflict of interest.
venous thrombosis is best in these images
(Tanoue et al. 2020). This advantage is of particu- Ethical Approval  This article does not contain any stud-
ies with human participants performed by any of the
lar interest because the documented prevalence authors.
of isolated pelvic thrombosis, thus thrombosis
that CT exclusively sees, is as low as 0.1–0.7%
(Reichert et  al. 2011). The respective radiation
dose is considerably high (5–10 mSv). Therefore, References
if performed at all, the radiation dose efficiency
Abdelbaky M, Zafar MA, Saeyeldin A et al (2019) Routine
should be as high as possible. Other authors anterior spinal artery visualization prior to descending
reported enhanced visualization for the superfi- and thoracoabdominal aneurysm repair: high detec-
cial veins of the legs before coronary artery tion success. J Card Surg 34(12):1563–1568. https://
bypass graft surgery by VMI+ (Schmoee et  al. doi.org/10.1111/jocs.14310
176 M. S. May and A. Muttke

Agatston AS, Janowitz WR, Hildner FJ, Zusmer NR, Caruso D, Parinella AH, Schoepf UJ et  al (2017)
Viamonte M, Detrano R (1990) Quantification of cor- Optimization of window settings for standard
onary artery calcium using ultrafast computed tomog- and advanced virtual monoenergetic imaging in
raphy. J Am Coll Cardiol 15(4):827–832. https://doi. abdominal dual-energy CT angiography. Abdom
org/10.1016/0735-­1097(90)90282-­t Radiol 42(3):772–780. https://doi.org/10.1007/
Albrecht MH, Trommer J, Wichmann JL et  al (2016a) s00261-­016-­0963-­9
Comprehensive comparison of virtual monoener- D’Angelo T, Bucher AM, Lenga L et  al (2018)
getic and linearly blended reconstruction techniques Optimisation of window settings for traditional and
in third-generation dual-source dual-energy com- noise-optimised virtual monoenergetic imaging in
puted tomography angiography of the thorax and dual-energy computed tomography pulmonary angi-
abdomen. Investig Radiol 51(9):582–590. https://doi. ography. Eur Radiol 28(4):1393–1401. https://doi.
org/10.1097/RLI.0000000000000272 org/10.1007/s00330-­017-­5059-­6
Albrecht MH, Scholtz J-E, Hüsers K et  al (2016b) D’Angelo T, Lenga L, Arendt CT et  al (2020) Carotid
Advanced image-based virtual monoenergetic dual-­ and cerebrovascular dual-energy computed tomog-
energy CT angiography of the abdomen: optimization raphy angiography: optimization of window settings
of kiloelectron volt settings to improve image contrast. for virtual monoenergetic imaging reconstruction.
Eur Radiol 26(6):1863–1870. https://doi.org/10.1007/ Eur J Radiol 130:109166. https://doi.org/10.1016/j.
s00330-­015-­3970-­2 ejrad.2020.109166
Altenbernd J-C, Stein I, Wetter A et  al (2015) Impact Das M, Braunschweig T, Mühlenbruch G et  al (2009)
of dual-energy CT prior to radioembolization Carotid plaque analysis: comparison of dual-source
(RE). Acta Radiol 56(11):1293–1299. https://doi. computed tomography (CT) findings and histopatho-
org/10.1177/0284185114558973 logical correlation. Eur J Vasc Endovasc Surg 38(1):14–
Bae K, Jeon KN, Cho SB et al (2018) Improved opacifi- 19. https://doi.org/10.1016/j.ejvs.2009.03.013
cation of a suboptimally enhanced pulmonary artery de Santis D, de Cecco CN, Schoepf UJ et  al (2019)
in chest CT: experience using a dual-layer detector Modified calcium subtraction in dual-energy CT
spectral CT.  AJR Am J Roentgenol 210(4):734–741. angiography of the lower extremity runoff: impact
https://doi.org/10.2214/AJR.17.18537 on diagnostic accuracy for stenosis detection. Eur
Bai X, Gao P, Zhang D et  al (2020) Plaque burden Radiol 29(9):4783–4793. https://doi.org/10.1007/
assessment and attenuation measurement of carotid s00330-­019-­06032-­y
atherosclerotic plaque using virtual monoenergetic Eichler M, May M, Wiesmueller M et  al (2020)
images in comparison to conventional polyenergetic Single source split filter dual energy: image qual-
images from dual-layer spectral detector CT.  Eur ity and liver lesion detection in abdominal CT.  Eur
J Radiol 132:109302. https://doi.org/10.1016/j. J Radiol 126:108913. https://doi.org/10.1016/j.
ejrad.2020.109302 ejrad.2020.108913
Beeres M, Trommer J, Frellesen C et al (2016) Evaluation Foti G, Silva R, Faccioli N et  al (2021) Identification
of different keV-settings in dual-energy CT angi- of pulmonary embolism: diagnostic accuracy
ography of the aorta using advanced image-based of venous-­ phase dual-energy CT in compari-
virtual monoenergetic imaging. Int J Cardiovasc son to pulmonary arteries CT angiography. Eur
Imaging 32(1):137–144. https://doi.org/10.1007/ Radiol 31(4):1923–1931. https://doi.org/10.1007/
s10554-­015-­0728-­5 s00330-­020-­07286-­7
Brockmann C, Jochum S, Sadick M et  al (2009) Dual-­ Gao Z, Meng D, Lu H, Yao B, Huang N, Ye Z (2016)
energy CT angiography in peripheral arterial occlusive Utility of dual-energy spectral CT and low-iodine con-
disease. Cardiovasc Intervent Radiol 32(4):630–637. trast medium in DIEP angiography. Int J Clin Pract
https://doi.org/10.1007/s00270-­008-­9491-­5 70(9):64–71. https://doi.org/10.1111/ijcp.12855
Buerke B, Wittkamp G, Seifarth H, Heindel W, Kloska Ghandour A, Sher A, Rassouli N, Dhanantwari A,
SP (2009) Dual-energy CTA with bone removal for Rajiah P (2018) Evaluation of virtual monoenergetic
transcranial arteries: intraindividual comparison images on pulmonary vasculature using the dual-­
with standard CTA without bone removal and TOF-­ layer detector-based spectral computed tomography.
MRA.  Acad Radiol 16(11):1348–1355. https://doi. J Comput Assist Tomogr 42(6):858–865. https://doi.
org/10.1016/j.acra.2009.05.007 org/10.1097/RCT.0000000000000748
Buffa V, Solazzo A, D'Auria V et  al (2014) Dual-­ Gladish GW, Du Choe H, Marom EM, Sabloff BS,
source dual-energy CT: dose reduction after endo- Broemeling LD, Munden RF (2006) Incidental pul-
vascular abdominal aortic aneurysm repair. Radiol monary emboli in oncology patients: prevalence, CT
Med 119(12):934–941. https://doi.org/10.1007/ evaluation, and natural history. Radiology 240(1):246–
s11547-­014-­0420-­1 255. https://doi.org/10.1148/radiol.2401051129
Carrascosa P, Capunay C, Rodriguez-Granillo GA, Godoy MCB, Naidich DP, Marchiori E et  al (2010)
Deviggiano A, Vallejos J, Leipsic JA (2014) Substantial Single-acquisition dual-energy multidetector com-
iodine volume load reduction in CT angiography with puted tomography: analysis of vascular enhancement
dual-energy imaging: insights from a pilot randomized and postprocessing techniques for evaluating the tho-
study. Int J Cardiovasc Imaging 30(8):1613–1620. racic aorta. J Comput Assist Tomogr 34(5):670–677.
https://doi.org/10.1007/s10554-­014-­0501-­1 https://doi.org/10.1097/RCT.0b013e3181e10627
Dual-Energy CT Angiography 177

Grimes J, Duan X, Yu L et al (2015) The influence of focal Radiol 21(3):530–537. https://doi.org/10.1007/
spot blooming on high-contrast spatial resolution in s00330-­010-­1954-­9
CT imaging. Med Phys 42(10):6011–6020. https://doi. Klink T, Wilhelm T, Roth C, Heverhagen JT (2017)
org/10.1118/1.4931053 Dual Energy CTA bei Patienten mit symptomatischer
Han D, Chen X, Lei Y et al (2019) Iodine load reduction pAVK: Studie über die diagnostische Genauigkeit und
in dual-energy spectral CT portal venography with low limitierende Faktoren. Röfo 189(5):441–452. https://
energy images combined with adaptive statistical iter- doi.org/10.1055/s-­0043-­101526
ative reconstruction. Br J Radiol 92(1100):20180414. Korn A, Bender B, Thomas C et  al (2011) Dual energy
https://doi.org/10.1259/bjr.20180414 CTA of the carotid bifurcation: advantage of plaque
He J, Wang Q, Ma X, Sun Z (2015) Dual-energy CT subtraction for assessment of grade of the stenosis and
angiography of abdomen with routine concentration morphology. Eur J Radiol 80(2):120–125. https://doi.
contrast agent in comparison with conventional single-­ org/10.1016/j.ejrad.2010.08.028
energy CT with high concentration contrast agent. Eur Korn A, Bender B, Schabel C et  al (2015) Dual-source
J Radiol 84(2):221–227. https://doi.org/10.1016/j. dual-energy CT angiography of the supra-aortic arter-
ejrad.2014.11.025 ies with tin filter: impact of tube voltage selection.
Hickethier T, Byrtus J, Hauger M et al (2018) Utilization Acad Radiol 22(6):708–713. https://doi.org/10.1016/j.
of virtual mono-energetic images (MonoE) derived acra.2015.01.016
from a dual-layer spectral detector CT (SDCT) for Kosmala A, Weng AM, Schmid A et  al (2020) Dual-­
the assessment of abdominal arteries in venous con- energy CT angiography in peripheral arterial occlu-
trast phase scans. Eur J Radiol 99:28–33. https://doi. sive disease: diagnostic accuracy of different image
org/10.1016/j.ejrad.2017.12.007 reconstruction approaches. Acad Radiol. https://doi.
Hou P, Feng X, Liu J et  al (2017) Iterative reconstruc- org/10.1016/j.acra.2020.10.028
tion in single-source dual-energy CT angiography: Kostrzewa M, Rathmann N, Hesser J, Huck K, Schönberg
feasibility of low and ultra-low volume contrast
­ SO, Diehl SJ (2016) automated vessel segmentation in
medium protocols. Br J Radiol 90(1075):20160506. dual energy computed tomography data of the pelvis
https://doi.org/10.1259/bjr.20160506 and lower extremities. In Vivo 30(5):651–655
Huang S-C, Pareek A, Zamanian R, Banerjee I, Lungren Kröger JR, Hickethier T, Pahn G, Gerhardt F, Maintz
MP (2020) Multimodal fusion with deep neural net- D, Bunck AC (2017) Influence of spectral detector
works for leveraging CT imaging and electronic health CT based monoenergetic images on the computer-­
record: a case-study in pulmonary embolism detec- aided detection of pulmonary artery embolism. Eur
tion. Sci Rep 10(1):22147. https://doi.org/10.1038/ J Radiol 95:242–248. https://doi.org/10.1016/j.
s41598-­020-­78888-­w ejrad.2017.08.034
Javor D, Wressnegger A, Unterhumer S et  al (2017) Kulkarni NM, Sahani DV, Desai GS, Kalva SP (2012)
Endoleak detection using single-acquisition split-­ Indirect computed tomography venography of
bolus dual-energy computer tomography (DECT). the lower extremities using single-source dual-
Eur Radiol 27(4):1622–1630. https://doi.org/10.1007/ energy computed tomography: advantage of low-­
s00330-­016-­4480-­6 kiloelectron volt monochromatic images. J Vasc Interv
Jia X, Li X, Li J et al (2020) Improving diagnostic accu- Radiol 23(7):879–886. https://doi.org/10.1016/j.
racy for arteries of lower extremities with dual-energy jvir.2012.04.012
spectral CT imaging. Eur J Radiol 128:109061. https:// Lee JW, Lee G, Lee NK et  al (2016) Effectiveness of
doi.org/10.1016/j.ejrad.2020.109061 adaptive statistical iterative reconstruction for 64-slice
Jin KN, Chung JW, Park E-A, Lee W (2017) Dual-energy dual-energy computed tomography pulmonary angiog-
computed tomography angiography: virtual calci- raphy in patients with a reduced iodine load: compari-
fied plaque subtraction in a vascular phantom. Acta son with standard computed tomography pulmonary
Radiol Open 6(7):2058460117717765. https://doi. angiography. J Comput Assist Tomogr 40(5):777–783.
org/10.1177/2058460117717765 https://doi.org/10.1097/RCT.0000000000000443
Kaemmerer N, Brand M, Hammon M et al (2016) Dual-­ Lehti L, Söderberg M, Höglund P, Nyman U, Gottsäter
energy computed tomography angiography of the A, Wassélius J (2018) Reliability of virtual
head and neck with single-source computed tomogra- non-contrast computed tomography angiogra-
phy: a new technical (split filter) approach for bone phy: comparing it with the real deal. Acta Radiol
removal. Investig Radiol 51(10):618–623. https://doi. Open 7(7-8):2058460118790115. https://doi.
org/10.1097/RLI.0000000000000290 org/10.1177/2058460118790115
Kau T, Eicher W, Reiterer C et al (2011) Dual-energy CT Lehti L, Söderberg M, Höglund P, Wassélius J (2019)
angiography in peripheral arterial occlusive disease-­ Comparing arterial- and venous-phase acquisition
accuracy of maximum intensity projections in clinical for optimization of virtual noncontrast images from
routine and subgroup analysis. Eur Radiol 21(8):1677– dual-energy computed tomography angiography. J
1686. https://doi.org/10.1007/s00330-­011-­2099-­1 Comput Assist Tomogr 43(5):770–774. https://doi.
Kerl JM, Bauer RW, Maurer TB et  al (2011) Dose lev- org/10.1097/RCT.0000000000000903
els at coronary CT angiography--a comparison of Leithner D, Mahmoudi S, Wichmann JL et  al (2018)
dual energy-, dual source- and 16-slice CT.  Eur Evaluation of virtual monoenergetic imaging algo-
178 M. S. May and A. Muttke

rithms for dual-energy carotid and intracerebral CT in lower extremity run-off CT angiography. Eur
angiography: effects on image quality, artefacts and Radiol 26(12):4380–4389. https://doi.org/10.1007/
diagnostic performance for the detection of stenosis. s00330-­016-­4304-­8
Eur J Radiol 99:111–117. https://doi.org/10.1016/j. Mannil M, Ramachandran J, Vittoria de Martini I et  al
ejrad.2017.12.024 (2017) Modified dual-energy algorithm for calci-
Lell MM, Kramer M, Klotz E, Villablanca P, Ruehm SG fied plaque removal: evaluation in carotid computed
(2009) Carotid computed tomography angiography tomography angiography and comparison with digital
with automated bone suppression: a comparative subtraction angiography. Investig Radiol 52(11):680–
study between dual energy and bone subtraction tech- 685. https://doi.org/10.1097/RLI.0000000000000391
niques. Investig Radiol 44(6):322–328. https://doi. Marin D, Caywood DT, Mileto A et  al (2015) Dual-­
org/10.1097/RLI.0b013e31819e8ad9 energy multidetector-row computed tomography of
Lell MM, Hinkmann F, Nkenke E et  al (2010) Dual the hepatic arterial system: optimization of energy
energy CTA of the supraaortic arteries: technical and material-specific reconstruction techniques. J
improvements with a novel dual source CT system. Comput Assist Tomogr 39(5):721–729. https://doi.
Eur J Radiol 76(2):6–12. https://doi.org/10.1016/j. org/10.1097/RCT.0000000000000259
ejrad.2009.09.022 Martin SS, Albrecht MH, Wichmann JL et  al (2017a)
Li S, Wang C, Jiang X, Xu G (2014) Effects of dual-­ Value of a noise-optimized virtual monoenergetic
energy CT with non-linear blending on abdominal CT reconstruction technique in dual-energy CT for
angiography. Korean J Radiol 15(4):430–438. https:// planning of transcatheter aortic valve replacement.
doi.org/10.3348/kjr.2014.15.4.430 Eur Radiol 27(2):705–714. https://doi.org/10.1007/
Li X, Chen GZ, Zhao YE et  al (2017) Radiation opti- s00330-­016-­4422-­3
mized dual-source dual-energy computed tomography Martin SS, Wichmann JL, Weyer H et al (2017b) Endoleaks
pulmonary angiography: intra-individual and inter-­ after endovascular aortic aneurysm repair: Improved
individual comparison. Acad Radiol 24(1):13–21. detection with noise-optimized virtual monoenergetic
https://doi.org/10.1016/j.acra.2016.09.009 dual-energy CT. Eur J Radiol 94:125–132. https://doi.
Li Z, Leng S, Halaweish AF et  al (2020) Overcoming org/10.1016/j.ejrad.2017.06.017
calcium blooming and improving the quantification Martin SS, Wichmann JL, Scholtz J-E et  al (2017c)
accuracy of percent area luminal stenosis by material Noise-optimized virtual monoenergetic dual-energy
decomposition of multi-energy computed tomogra- CT improves diagnostic accuracy for the detection of
phy datasets. J Med Imaging 7(5):53501. https://doi. active arterial bleeding of the abdomen. J Vasc Interv
org/10.1117/1.JMI.7.5.053501 Radiol 28(9):1257–1266. https://doi.org/10.1016/j.
Liu J, Lv P-J, Wu R et al (2014) Aortic dual-energy CT jvir.2017.06.011
angiography with low contrast medium injection May MS, Wiesmueller M, Heiss R et  al (2019a)
rate. J Xray Sci Technol 22(5):689–696. https://doi. Comparison of dual- and single-source dual-energy
org/10.3233/XST-­140454 CT in head and neck imaging. Eur Radiol 29(8):4207–
Liu Y, Liu A, Liu L et  al (2016) Feasibility of spectral 4214. https://doi.org/10.1007/s00330-­018-­5762-­y
imaging with low-concentration contrast medium May MS, Wetzl M, Wust W, et  al (2019b) Advanced
in abdominal CT angiography of obese patients. Int visualization of peroneal artery perforators prior to
J Clin Pract 70(9):37–43. https://doi.org/10.1111/ autologous transplantation in head and neck surgery
ijcp.12856 by dual-energy CT and multiplanar vessel unfolding.
Ma G, He T, Yu Y, Duan H, Yang C (2016a) Improving image Radiological Society of North America. RC612-09
quality of bronchial arteries with virtual monochro- McDermott S, Otrakji A, Flores EJ, Kalra MK, Shepard
matic spectral CT images. PLoS One 11(3):e0150985. J-AO, Digumarthy SR (2018) Should dual-energy
https://doi.org/10.1371/journal.pone.0150985 computed tomography pulmonary angiography
Ma C-L, Chen X-X, Lei Y-X et al (2016b) Clinical value replace single-energy computed tomography pul-
of dual-energy spectral imaging with adaptive sta- monary angiography in pregnant and postpartum
tistical iterative reconstruction for reducing contrast patients? J Comput Assist Tomogr 42(1):25–32.
medium dose in CT portal venography: in compari- https://doi.org/10.1097/RCT.0000000000000655
son with standard 120-kVp imaging protocol. Br J Meier A, Wurnig M, Desbiolles L, Leschka S, Frauenfelder
Radiol 89(1062):20151022. https://doi.org/10.1259/ T, Alkadhi H (2015) Advanced virtual monoenergetic
bjr.20151022 images: improving the contrast of dual-energy CT pul-
Ma G, Dou Y, Dang S et al (2019) Influence of monoener- monary angiography. Clin Radiol 70(11):1244–1251.
getic images at different energy levels in dual-energy https://doi.org/10.1016/j.crad.2015.06.094
spectral CT on the accuracy of computer-aided detec- Meier A, Higashigaito K, Martini K et  al (2016)
tion for pulmonary embolism. Acad Radiol 26(7):967– Dual energy CT pulmonary angiography with 6g
973. https://doi.org/10.1016/j.acra.2018.09.007 Iodine-A propensity score-matched study. PLoS One
Mangold S, Cecco CN, Schoepf UJ et al (2016) A noise-­ 11(12):e0167214. https://doi.org/10.1371/journal.
optimized virtual monochromatic reconstruction pone.0167214
algorithm improves stent visualization and diag- Meier A, Higashigaito K, Martini K et  al (2017)
nostic accuracy for detection of in-stent re-stenosis Correction: dual energy CT pulmonary angiography
Dual-Energy CT Angiography 179

with 6g Iodine-A propensity score-matched study. Jpn J Radiol 34(9):611–619. https://doi.org/10.1007/

PLoS One 12(2):e0172251. https://doi.org/10.1371/ s11604-­016-­0564-­0
journal.pone.0172251 Petritsch B, Kosmala A, Gassenmaier T et  al (2017)
Meyer BC, Werncke T, Hopfenmüller W, Raatschen Diagnostik der akuten Lungenarterienembolie:
HJ, Wolf K-J, Albrecht T (2008) Dual energy CT of Vergleich von single-source CT und Dritt-­
peripheral arteries: effect of automatic bone and plaque generation dual-source CT unter Einsatz eines
removal on image quality and grading of stenoses. Eur dual-energy protokolls  – Bildqualität und
J Radiol 68(3):414–422. https://doi.org/10.1016/j. Strahlenexposition. Röfo 189(6):527–536. https://doi.
ejrad.2008.09.016 org/10.1055/s-­0043-­103089
Meyer M, Haubenreisser H, Schabel C et  al (2018) CT Pinho DF, Kulkarni NM, Krishnaraj A, Kalva SP, Sahani
pulmonary angiography in patients with acute or DV (2013) Initial experience with single-source
chronic renal insufficiency: evaluation of a low dose dual-energy CT abdominal angiography and com-
contrast material protocol. Sci Rep 8(1):1995. https:// parison with single-energy CT angiography: image
doi.org/10.1038/s41598-­018-­20254-­y quality, enhancement, diagnosis and radiation dose.
Mikail N, Meseguer E, Lavallée P et al (2021) Evaluation Eur Radiol 23(2):351–359. https://doi.org/10.1007/
of non-stenotic carotid atherosclerotic plaques with s00330-­012-­2624-­x
combined FDG-PET imaging and CT angiogra- Qanadli SD, El Hajjam M, Mesurolle B et  al (1999)
phy in patients with ischemic stroke of unknown Motion artifacts of the aorta simulating aortic dissec-
origin. J Nucl Cardiol. https://doi.org/10.1007/ tion on spiral CT. J Comput Assist Tomogr 23(1):1–6.
s12350-­020-­02511-­8 https://doi.org/10.1097/00004728-­199901000-­00001
Montanaro M, Scimeca M, Anemona L et  al (2021) Reichert M, Henzler T, Krissak R et  al (2011) Venous
The paradox effect of calcification in carotid thromboembolism: additional diagnostic value
­atherosclerosis: microcalcification is correlated with and radiation dose of pelvic CT venography in
plaque instability. Int J Mol Sci 22(1):395. https://doi. patients with suspected pulmonary embolism. Eur
org/10.3390/ijms22010395 J Radiol 80(1):50–53. https://doi.org/10.1016/j.
Morhard D, Fink C, Graser A, Reiser MF, Becker C, ejrad.2010.12.101
Johnson TRC (2009) Cervical and cranial com- Renapurkar RD, Primak A, Azok J et  al (2017)
puted tomographic angiography with automated Attenuation-based kV pair selection in dual source
bone removal: dual energy computed tomography dual energy computed tomography angiography of
versus standard computed tomography. Investig the chest: impact on radiation dose and image quality.
Radiol 44(5):293–297. https://doi.org/10.1097/ Eur Radiol 27(8):3283–3289. https://doi.org/10.1007/
RLI.0b013e31819b6fba s00330-­016-­4714-­7
Murphy A, Cheng J, Pratap J, Redman R, Coucher Reynoso E, Rodriguez-Granillo GA, Capunay C,
J (2019) Dual-energy computed tomography Deviggiano A, Meli F, Carrascosa P (2017) Spectral
pulmonary angiography: comparison of vessel signal density of carotid plaque using dual-energy
enhancement between linear blended and vir- computed tomography. J Neuroimaging 27(5):511–
tual monoenergetic reconstruction techniques. J 516. https://doi.org/10.1111/jon.12440
Med Imaging Radiat Sci 50(1):62–67. https://doi. Riffel P, Haubenreisser H, Meyer M et al (2016) Carotid
org/10.1016/j.jmir.2018.10.009 dual-energy CT angiography: evaluation of low keV
Nadjiri J, Pfeiffer D, Straeter AS et  al (2018) Spectral calculated monoenergetic datasets by means of a
computed tomography angiography with a frequency-split approach for noise reduction at low
gadolinium-­based contrast agent: first clinical imag- keV levels. Eur J Radiol 85(4):720–725. https://doi.
ing results in cardiovascular applications. J Thorac org/10.1016/j.ejrad.2016.01.015
Imaging 33(4):246–253. https://doi.org/10.1097/ Rotzinger DC, Si-Mohamed SA, Shapira N, Douek PC,
RTI.0000000000000335 Meuli RA, Boussel L (2020) “Dark-blood” dual-­
Nakayama K, Shimohira M, Nakagawa M et  al (2020) energy computed tomography angiography for tho-
Advanced monoenergetic reconstruction technique racic aortic wall imaging. Eur Radiol 30(1):425–431.
in dual-energy computed tomography for evalua- https://doi.org/10.1007/s00330-­019-­06336-­z
tion of vascular anatomy before adrenal vein sam- Saba L, Mallarin G (2009) Window settings for the study
pling. Acta Radiol 61(2):282–288. https://doi. of calcified carotid plaques with multidetector CT
org/10.1177/0284185119860226 angiography. AJNR Am J Neuroradiol 30(7):1445–
Neuhaus V, Große Hokamp N, Abdullayev N et  al 1450. https://doi.org/10.3174/ajnr.A1509
(2018) Comparison of virtual monoenergetic and Saba L, Sanfilippo R, Balestrieri A et  al (2017)
polyenergetic images reconstructed from dual-layer Relationship between carotid computed tomogra-
detector CT angiography of the head and neck. Eur phy dual-energy and brain leukoaraiosis. J Stroke
Radiol 28(3):1102–1110. https://doi.org/10.1007/ Cerebrovasc Dis 26(8):1824–1830. https://doi.
s00330-­017-­5081-­8 org/10.1016/j.jstrokecerebrovasdis.2017.04.016
Onozawa S, Murata S, Yamaguchi H et al (2016) Can an Saba L, Argioas GM, Lucatelli P, Lavra F, Suri JS,
enhanced thin-slice computed tomography delineate Wintermark M (2019) Variation of degree of stenosis
the right adrenal vein and improve the success rate? quantification using different energy level with dual
180 M. S. May and A. Muttke

energy CT scanner. Neuroradiology 61(3):285–291. bleeding: image quality, radiation dose and diagnostic
https://doi.org/10.1007/s00234-­018-­2142-­x performance. Eur J Radiol 84(5):884–891. https://doi.
Sauter AP, Shapira N, Kopp FK et al (2020) CTPA with org/10.1016/j.ejrad.2015.01.013
a conventional CT at 100 kVp vs. a spectral-detector Tanoue S, Nakaura T, Iyama Y et  al (2020) Diagnostic
CT at 120 kVp: comparison of radiation exposure, performance of dual-layer computed tomography for
diagnostic performance and image quality. Eur J deep vein thrombosis in indirect computed tomogra-
Radiol Open 7:100234. https://doi.org/10.1016/j. phy venography. Circ J 84(4):636–641. https://doi.
ejro.2020.100234 org/10.1253/circj.CJ-­19-­0722
Schabel C, Bongers M, Sedlmair M et al (2014) Assessment Uhrig M, Simons D, Schlemmer H-P (2018) Incidental
of the hepatic veins in poor contrast conditions using pulmonary emboli in stage IV melanoma patients:
dual energy CT: evaluation of a novel monoenergetic Prevalence in CT staging examinations and improved
extrapolation software algorithm. Röfo 186(6):591– detection with vessel reconstructions based on dual
597. https://doi.org/10.1055/s-­0034-­1366423 energy CT.  PLoS One 13(7):e0199458. https://doi.
Schabel C, Bongers MN, Ketelsen D et  al (2015) org/10.1371/journal.pone.0199458
Diagnostische Genauigkeit der Dual-energy-CT-­ Uotani K, Watanabe Y, Higashi M et  al (2009) Dual-­
Angiographie bei Patienten mit Diabetes mellitus. energy CT head bone and hard plaque removal for
Radiologe 55(4):314–322. https://doi.org/10.1007/ quantification of calcified carotid stenosis: utility and
s00117-­014-­2721-­8 comparison with digital subtraction angiography. Eur
Schmoee J, Dirrichs T, Fehrenbacher K et  al (2020) Radiol 19(8):2060–2065. https://doi.org/10.1007/
Virtual monoenergetic images (VMI+) in dual-source s00330-­009-­1358-­x
dual-energy CT venography (DSDE-CTV) of the van der Molen AJ, Reimer P, Dekkers IA et  al (2018)
lower extremity prior to coronary artery bypass graft Post-contrast acute kidney injury. Part 2: risk strati-
(CABG): a feasibility study. Acad Radiol 27(9):1249– fication, role of hydration and other prophylactic
1254. https://doi.org/10.1016/j.acra.2019.11.005 measures, patients taking metformin and chronic
Schulz B, Kuehling K, Kromen W et al (2012) Automatic dialysis patients: recommendations for updated ESUR
bone removal technique in whole-body dual-energy Contrast Medium Safety Committee guidelines. Eur
CT angiography: performance and image quality. AJR Radiol 28(7):2856–2869. https://doi.org/10.1007/
Am J Roentgenol 199(5):W646–W650. https://doi. s00330-­017-­5247-­4
org/10.2214/AJR.12.9176 Wang Q, Shi G, Liu X, Wu R, Wang S (2013) Optimal
Shaida N, Bowden DJ, Barrett T et al (2012) Acceptability contrast of computed tomography portal venog-
of virtual unenhanced CT of the aorta as a replace- raphy using dual-energy computed tomography. J
ment for the conventional unenhanced phase. Clin Comput Assist Tomogr 37(2):142–148. https://doi.
Radiol 67(5):461–467. https://doi.org/10.1016/j. org/10.1097/RCT.0b013e31827cd656
crad.2011.10.023 Wang Q, Gaofeng S, Xueli F, Lijia W, Runze W (2014)
Shinohara Y, Sakamoto M, Kuya K et al (2015) Assessment Comparative study of hepatic venography using
of carotid plaque composition using fast-kV switching non-linear-blending images, monochromatic images
dual-energy CT with gemstone detector: comparison and low-voltage images of dual-energy CT.  Br J
with extracorporeal and virtual histology-intravascular Radiol 87(1042):20140220. https://doi.org/10.1259/
ultrasound. Neuroradiology 57(9):889–895. https:// bjr.20140220
doi.org/10.1007/s00234-­015-­1541-­5 Watanabe Y, Nakazawa T, Higashi M, Itoh T, Naito H
Shuman WP, O’Malley RB, Busey JM, Ramos MM, (2011) Assessment of calcified carotid plaque volume:
Koprowicz KM (2017) Prospective comparison of comparison of contrast-enhanced dual-energy CT
dual-energy CT aortography using 70% reduced angiography and native single-energy CT. AJR Am J
iodine dose versus single-energy CT aortography Roentgenol 196(6):796–799. https://doi.org/10.2214/
using standard iodine dose in the same patient. Abdom AJR.10.4938
Radiol 42(3):759–765. https://doi.org/10.1007/ Wichmann JL, Gillott MR, Cecco CN et  al (2016)
s00261-­016-­1041-­z Dual-­ energy computed tomography angiography
Sommer WH, Johnson TR, Becker CR et al (2009) The of the lower extremity runoff: impact of noise-
value of dual-energy bone removal in maximum inten- optimized virtual monochromatic imaging on
sity projections of lower extremity computed tomog- image quality and diagnostic accuracy. Investig
raphy angiography. Investig Radiol 44(5):285–292. Radiol 51(2):139–146. https://doi.org/10.1097/
https://doi.org/10.1097/RLI.0b013e31819b70ba RLI.0000000000000216
Sommer WH, Graser A, Becker CR et  al (2010) Image Xin L, Yang X, Huang N et al (2015) The initial experi-
quality of virtual noncontrast images derived from ence of the upper abdominal CT angiography using
dual-energy CT angiography after endovascular aneu- low-concentration contrast medium on dual energy
rysm repair. J Vasc Interv Radiol 21(3):315–321. spectral CT.  Abdom Imaging 40(7):2894–2899.
https://doi.org/10.1016/j.jvir.2009.10.040 https://doi.org/10.1007/s00261-­015-­0462-­4
Sun H, Hou X-Y, Xue H-D et al (2015) Dual-source dual-­ Yamamoto S, McWilliams J, Arellano C et al (2009) Dual-­
energy CT angiography with virtual non-enhanced energy CT angiography of pelvic and lower extremity
images and iodine map for active gastrointestinal arteries: dual-energy bone subtraction versus manual
Dual-Energy CT Angiography 181

bone subtraction. Clin Radiol 64(11):1088–1096. Zhang D, Xie Y, Wang Y et al (2020) Initial clinical expe-
https://doi.org/10.1016/j.crad.2009.07.009 rience of virtual monoenergetic imaging improves
Yoshida M, Nakaura T, Sentaro T et al (2020) Prospective stent visualization in lower extremity run-off CT
comparison of 70-kVp single-energy CT versus dual- angiography by dual-layer spectral detector CT. Acad
energy CT: which is more suitable for CT angiography Radiol 27(6):825–832. https://doi.org/10.1016/j.
with low contrast media dosage? Acad Radiol 27(5):116– acra.2019.07.022
122. https://doi.org/10.1016/j.acra.2019.07.016 Zhou K, He J, Qu F et  al (2017) Comparison of 270
Zeng X, Wang X, Chen H et  al (2020) Evaluating the versus 320 mg I/mL of Iodixanol in 1 image assess-
image quality of monoenergetic images from dual-­ ment of both renal arteries and veins with dual-
energy computed tomography with low-concentration energy spectral CT imaging in late arterial phase and
and low-flow-rate contrast media for the arterials their influence on renal function. J Comput Assist
supply to the nipple-areola complex in breast cancer Tomogr 41(5):798–803. https://doi.org/10.1097/
compared with conventional computed tomography RCT.0000000000000595
angiography. J Comput Assist Tomogr 44(6):921–927. Zopfs D, Lennartz S, Laukamp K et al (2018) Improved
https://doi.org/10.1097/RCT.0000000000001063 depiction of atherosclerotic carotid artery stenosis in
Zhang LJ, Wang ZJ, Lu L, Chen B, Lu GM (2011) virtual monoenergetic reconstructions of venous phase
Feasibility of gadolinium-enhanced dual energy CT dual-layer computed tomography in comparison to
pulmonary angiography: a pilot study in rabbits. Int polyenergetic reconstructions. Eur J Radiol 100:36–
J Cardiovasc Imaging 27(7):1069–1080. https://doi. 42. https://doi.org/10.1016/j.ejrad.2018.01.008
org/10.1007/s10554-­010-­9755-­4
 Thoracic Imaging: Ventilation/
Perfusion

Hye Jeon Hwang, Sang Min Lee,
and Joon Beom Seo

Contents
1  erfusion Dual-Energy CT 
P  184
1.1  Imaging Protocol   184
1.2  Clinical Applications   184
2  entilation Dual-Energy CT 
V  188
2.1  Technical Aspects   188
2.2  Clinical Applications   190
3  omprehensive Assessment of Morphology, Pulmonary Ventilation,
C
Perfusion, and Relationship of Ventilation and Perfusion Using
Dual-Energy CT   194
3.1  Concept   194
3.2  Clinical Application: Pulmonary Embolism, COPD   194
4 Perspective and Conclusion   197
References   197

Abstract to evaluate and visualize the pulmonary perfu-

sion or ventilation status. The above methods
Various lung diseases cause structural or reflex have limitations of spatial resolution and pres-
changes in the lungs, resulting in changes in ent challenges in the generation of the radio-
ventilation or perfusion status, which lead to active tag. Multidetector computed
ventilation–perfusion imbalances. Various tomography is the modality of choice in lung
imaging methods, including magnetic reso- imaging for anatomic evaluation by providing
nance imaging (MRI), gamma camera-based excellent spatial resolution: however, it
planar scintigraphy, single-photon emission requires repeated CT scans for evaluating pul-
computed tomography (SPECT), or positron monary ventilation or perfusion, which may
emission tomography (PET) have been used result in increased radiation exposure and
potential misregistration of images. With the
advance of dual-energy computed tomogra-
H. J. Hwang · S. M. Lee · J. B. Seo (*)
Department of Radiology and Research Institute of
phy (DECT), it has become possible to pro-
Radiology, University of Ulsan College of Medicine, vide pulmonary functional information such
Asan Medical Center, Seoul, South Korea as pulmonary perfusion or ventilation map
e-mail: [email protected]

© The Author(s), under exclusive license to Springer Nature Switzerland AG 2022 183
H. Alkadhi et al. (eds.), Spectral Imaging, Medical Radiology Diagnostic Imaging,
https://doi.org/10.1007/978-3-030-96285-2_12
184 H. J. Hwang et al.

and high-resolution morphologic information on the pulmonary circulation from the same data
within a single CT scan. This advantage of set. Many factors influence the enhancement
DECT may facilitate the acquisition of coreg- within lung microvessels: the volume and flow
istered structural and functional information rate of the contrast media, the contrast media
within a clinical setting’s constraints. This administration site, and the anatomic structures
chapter will present the clinical applications through which the iodinated contrast media
for the assessment of pulmonary perfusion, courses (Thieme et  al. 2009). Another factor is
ventilation, and ventilation–perfusion imbal- the systemic circulation, because of its role in
ance using DECT in various lung diseases. collateral supply and the numerous anastomoses
within the pulmonary circulation.

1 Perfusion Dual-Energy CT 1.2 Clinical Applications

1.1 Imaging Protocol 1.2.1 Pulmonary Vascular Diseases

(Pulmonary Embolism
Lung perfusion or pulmonary blood volume and Pulmonary Hypertension)
(PBV) image is based on quantification of the Pulmonary embolism (PE) is a common disease
enhancement in tissue and blood at certain time cardiovascular disease. The emboli, migrated
points after IV administration of contrast media from deep vein thrombosis of the lower limb,
(Gorgos et al. 2009). This image is not true perfu- cause either total or partial blockage of the pul-
sion but reflects the regional volume of blood to monary arteries. Because outcomes without
which fresh blood is being delivered. If this fresh proper treatment are poor, accurate and fast diag-
blood is equilibrated with iodine, then the nosis of PE is essential. CT pulmonary angiogra-
regional iodine concentration measured by DECT phy (CTPA) can visualize emboli and the ancillary
reflects the regional PBV.  As increased blood findings associated with PE.  CTPA has been
flow to a region of lung may serve to dilate the accepted as the method of choice for imaging sus-
capillary beds and recruit new capillaries, a mea- pected PE.  However, CTPA does not provide
sure of regional PBV may serve as a surrogate for functional assessment of lung, which can be eval-
pulmonary parenchymal perfusion. uated with ventilation/perfusion scintigraphy.
For accurate assessment of PBV, there are sev- With a single scan of DECT, both the high
eral important considerations (Hwang et  al. spatial/temporal resolution axial image and the
2017). First, PBV can be affected by the amount PBV map can be acquired simultaneously with
of breathing volume at the time of breath-hold. similar radiation dose of CTPA (Fig.  1). PBV
And important assumption of PBV is that an map can provide perfusion defects in the lung
amount of iodine is equilibrated in the blood at parenchyma for the diagnosis of acute
the time of imaging and that equilibrated iodine PE. Perfusion defects distal to the occlusion usu-
concentration remains constant during the scan- ally have the typical wedge shape.
ning of the lung. In calculating PBV, it is impor- In acute PE, the PBV map of DECT shows
tant to standardize the regional measurements by strong correlation with pulmonary perfusion
the iodine signal within the feeding pulmonary scintigraphy and SPECT images (Thieme et  al.
artery. 2008, 2012a). The additional information of
Lung perfusion image from DECT provides parenchymal perfusion impairment from DECT
an iodine map of the lung microcirculation which can improve the diagnosis of acute PE (Okada
represents a measurement at one time point only. et al. 2015). In clinical practice, it is difficult to
Though this image shows only blood volume at detect small thrombi in the segmental or subseg-
certain time points, this modality provides high-­ mental pulmonary arteries on conventional CT
quality morphologic and functional information angiography. With the postprocessing of DECT
Thoracic Imaging: Ventilation/Perfusion 185

a b

c d

Fig. 1  Example of CTPA using DECT in a 71-year-old fusion images of PBV maps show wedge shape perfusion
man with acute pulmonary embolism. (a, b) Axial and defects in the right middle lobe and left lingular segment
coronal CTPA images show large clots in both lobar and of the left upper lobe. A large perfusion defect is also
segmental pulmonary arteries. (c, d) Axial and coronal noted in the left lower lobe

image, the direct color-coded differentiation of DECT-based perfusion defects in patients with
thrombosed vessels and contrast-filled non-­ PE correlated with clinical parameters of acute
thrombosed vessels can be assessed. The perfu- PE severity (Thieme et al. 2012b). The quantifi-
sion impairment of PBV and the differentiation cation assessment of perfusion defect volume of
of thrombosed/non-thrombosed vessels map help the PBV map was correlated with established CT
detection of small endovascular thrombi (Lee parameters (Meinel et al. 2013a; Sakamoto et al.
et al. 2011) (Fig. 2). 2014).
Using the perfusion defects on PBV map, the Chronic PE may show a mosaic pattern of
severity of acute PE can be investigated. The lung attenuation due to redistribution of blood
visually assessed perfusion defects score from flow (Remy-Jardin et al. 2010). DECT can dif-
the PBV map showed correlation with estab- ferentiate ground glass attenuation of vascular
lished CT parameters of PE severity such as RV/ origin with high iodine contents from ground
LV ratio and CT angiography obstruction score glass attenuation of other origin. DECT pro-
(Chae et al. 2010a). And visual scoring system of vides virtual non-contrast (VNC) image which
186 H. J. Hwang et al.

a b c

Fig. 2  Example of detection of small peripheral embo- shaped perfusion defect distal to the small indecisive ves-
lism in a 56-year-old man with elevated d-dimer using sel. (c) The color-coded CT image with the dedicated
DECT. (a) Axial CT pulmonary angiography shows low dual-energy software, “Lung Vessels,” shows these small
attenuated small vessels compared with other normally vessels in red, representing thrombosed vessels without
enhanced small vessels in the left upper lobe, suspicious iodine
of peripheral embolic clots. (b) PBV map shows a wedge-­

can detect calcification in chronic clots. In sildenafil in emphysema susceptible normal

chronic PE, the bronchial circulation and sys- smokers (Iyer et al. 2016). These data show that
temic circulation may be increased due to vascular heterogeneity of smokers is not due to a
decreased pulmonary flow and ischemia. Two- fixed remodeling but is the result of an active
phase DECT can differentiate acute and chronic vasoconstriction.
PE (Hong et al. 2013). In delayed second phase In COPD, regional perfusion can be impaired
PBV map, lung area with systemic collateral due to loss of capillary beds from emphysema-
circulation showed increased enhancement tous destruction or chronic inflammation and
compared with arterial phase PBV map. Using hypoxic vasoconstriction (Barbera et  al. 2003).
DECT, enhancement values of pulmonary DECT can be helpful to assess patients with
artery and whole lung parenchyma can be quan- COPD, because the severity of anatomic destruc-
tified, and the derived central-to-peripheral tion and the change of parenchymal perfusion
ratio can help to detect chronic thromboem- make the functional effect of the disease. The
bolic pulmonary hypertension (Ameli-Renani regional lung perfusion change on PBV map is
et al. 2014). matched to the parenchymal destruction and
quantitative information of parenchymal destruc-
1.2.2 Diffuse Parenchymal/Airway tion and perfusion can be provided simultane-
Disease ously using DECT (Lee et al. 2012; Pansini et al.
With the assessment of PBV map, pulmonary 2009). Enhanced CT cannot be used for quantifi-
vascular dysfunction can be evaluated as a possi- cation of emphysema; however, VNC image
ble role in the etiology of smoking associated showed moderate correlation with PFTs (Lee
emphysema. In smokers with normal pulmonary et al. 2012). Using the automated quantified PBV
function tests (PFTs), the smokers with centriaci- value, pulmonary perfusion in emphysema can
nar emphysema on CT showed nearly double het- be assessed (Meinel et al. 2013b). The quantified
erogeneity of parenchymal perfusion than PBV showed correlation with PFTs and carbon
smokers without early emphysema (Alford et al. monoxide diffusing capacity (DLCO). The global
2010). This increased heterogeneity of PBV can PBV showed stronger correlation with DLCO than
be reversible with a sing oral dose of 20 mg of emphysema severity.
Thoracic Imaging: Ventilation/Perfusion 187

Fig. 3  Perfusion DECT in a 66-year-old man with severe sion DECT, the left lower lobe was selected for the target
emphysema shows hyperinflation and decreased paren- lobe of bronchoscopic lung volume reduction.
chymal perfusion in the left lower lobe. After lobe seg- Bronchoscopic lung volume reduction was performed
mentation, perfusion ratio of LLL (12%) is more successfully in this patient, and the percent predicted
decreased than volume ratio of LLL (22%). Considering FEV1 was improved from 18% to 32%
the conventional CT findings and perfusion map of perfu-

In severe COPD patients, assessment of can be used for prediction of postoperative lung
emphysema distribution and regional perfusion function in patients undergoing lung resection
using DECT can help target lobe selection of sur- and can predict more accurately than perfusion
gical or bronchoscopic lung volume reduction scintigraphy, because DECT can do the precise
(Park et al. 2014) (Fig. 3). Most hyperinflated and segmentation and measurement of lobar perfu-
least perfused lobe of the emphysematous lung sion on high-resolution CT imaging and perfectly
on DECT was selected as the target lobe. matched PBV maps (Chae et al. 2013) (Fig. 4).
Additionally, CT image can provide fissure integ- The maximum iodine-related attenuation of
rity of the target lobe. lung cancer on DECT showed a moderate corre-
In interstitial lung disease (ILD), DECT can lation with the maximum standardized uptake
be used for assessing ILD and predicting progno- value (SUVmax) on FDG-PET/CT.  Especially
sis with assessment of texture pattern-based non-small cell lung cancer showed a strong cor-
quantification and the PBV map (Moon et  al. relation between SUVmax and maximum iodine-­
2016). Perfusion and morphologic assessments related attenuation (Schmid-Bindert et al. 2012).
on DECT were correlated with clinical parame- The maximum iodine-related attenuation on
ters such as PFTs, DLco, or 6-min walk test and DECT may be a useful surrogate parameter for
can predict survival of ILD patients. assessment of therapy response in non-small cell
lung cancer patient.
1.2.3 Lung Cancer Dual-phase DECT can assess therapeutic
The lobar perfusion ratio on perfusion scintigra- effect after conservative therapy including anti-­
phy is widely used to predict postoperative forced angiogenesis therapy for not only primary
expiratory volume during 1 s (FEV1) for preop- lesions, but also mediastinal lymph node metas-
erative risk stratification (Brunelli and Fianchini tases in non-small cell lung cancer (Baxa et  al.
1997; Pierce et al. 1994). DECT with PBV map 2014, 2016). Quantification of arterial iodine
188 H. J. Hwang et al.

Fig. 4  Perfusion DECT in a 56-year-old man with a cen- and predicted postoperative FEV1 using perfusion DECT
tral lung mass in the left lower lobe abutting to left main was 1.56L (43%). After 3 months from left pneumonec-
pulmonary artery. After lobe segmentation, perfusion ratio tomy, postoperative FEV1 was 1.65L (45%)
of Lt. lung is 44%. Preoperative FEV1 was 2.79L (76%)

uptake can predict and assess therapeutic effect. tion contrast agents (Gur et  al. 1979; Herbert
In addition, ratio of early and late iodine uptake et al. 1982; Chae et al. 2010b; Chon et al. 1985;
quantification can evaluate function of mediasti- Hong et al. 2016). These are radio-opaque gases
nal lymph node. with high atomic numbers (54 for xenon and 36
for krypton) that resemble the X-ray absorption
characteristics of iodine. Their concentration in
2 Ventilation Dual-Energy CT the airspaces of the lung including alveolar space
can be measured based on the CT attenuation,
2.1 Technical Aspects and it has been shown that local xenon concentra-
tion is linearly related to CT attenuation (Gur
2.1.1 Noble Gas Contrast Agents et al. 1979; Murphy et al. 1989). Xenon has anes-
For ventilation imaging with DECT, two stable thetic properties and respiratory depression effect
gases, xenon and krypton are eligible as inhala- at a concentration of over 30–40% (Bedi et  al.
Thoracic Imaging: Ventilation/Perfusion 189

2002; Latchaw et  al. 1987; Tonner 2006); thus, may vary depending on study protocols; and the
the 30% concentration of xenon gas is clinically multiple breath-in method for less than 2 min has
used in ventilation imaging. Other side effects been used in most of the studies for patients’
included headache, somnolence nausea and vom- safety and sufficient CT attenuation for visual
iting, and physiologic effects, such as lighthead- and quantitative analyses (Chae et  al. 2010b;
edness and labile emotion (Latchaw et al. 1987; Park et  al. 2010; Hwang et  al. 2020). However,
Yonas et  al. 1981). These adverse effects are the multiple breath-in method might obscure the
mostly mild and uncommon and not resulted in mild degree of ventilation abnormalities that can
any persistent neurologic change or other be identified on 3He MRI with the single breath-
sequelae. Stable krypton gas can be an alternative in method. In recent years, xenon ventilation
­
to xenon gas for ventilation imaging, and its DECT using a single vital capacity inhalation of
potential usefulness as an inhalation contrast 35% xenon gas also has been tried (Honda et al.
agent for ventilation imaging has been investi- 2012; Kyoyama et al. 2017). Monitoring of respi-
gated in several studies (Hong et al. 2016; Chung ratory rate, oxygen saturation, and blood pressure
et al. 2014; Winkler et al. 1977). Krypton gas is of patients, and xenon concentrations within
less radio-opaque than xenon gas (Chon et  al. inhaled and exhaled gas and carbon dioxide con-
1985); however, it has no anesthetic properties centrations within exhaled gas are recommended
like xenon gas. With improvements in spectral throughout the study.
separation and detector sensitivity of scanners, Ventilation DECT scanning with either single
krypton is expected to become a viable alterna- static or dynamic acquisition protocols is avail-
tive to xenon gas (Chung et al. 2014). able depending on the purpose of the ventilation
imaging. The single static scan is usually per-
2.1.2 Imaging Protocol formed with multiple breath-in method, and CT
Xenon ventilation DECT can be successfully image acquisition is usually performed with full
performed using the dual-source CT scanners lung coverage at the end of the wash-in period,
(Somatom Definition scanner and Somatom and an additional CT scan with full lung coverage
Definition FLASH, Siemens Healthcare, may be performed at the end of the wash-out
Forchheim, Germany) (Chae et al. 2010b, 2008; period. This method can fully answer whether a
Goo et al. 2008; Park et al. 2010). region of the lung is or is not being ventilated,
Patients inhale xenon gas using the close-­ although the minor regional ventilation heteroge-
fitting face mask designed to deliver positive neity may be obscured. The dynamic scan of ven-
pressure ventilation treatment (Chae et al. 2010b, tilation DECT during wash-in and wash-out of
2008; Hong et al. 2016; Park et al. 2010; Hwang xenon gas may provide more detailed information
et al. 2020). As Chae et al. (2008) demonstrated, of regional ventilation, reflecting real ventilation,
a xenon concentration of 30% is sufficient for an as pathologies of various airway diseases affect
adequate CT attenuation, and many studies used the outflow of the air, as well as the inflow of the
the mixture of 30% xenon and 70% oxygen with air. However, additional radiation exposure is
the use of an inhalation system (Chae et  al. required in this protocol. In a dynamic acquisition
2010b, 2008; Goo et al. 2008; Park et al. 2010; protocol, CT images are usually obtained every
Hwang et  al. 2020). The patient may inhale a 18–30 s during both the wash-in and wash-­out
high concentration (>60%) of oxygen for 2–3 min period with limited or targeted coverage, and sin-
before inhaling xenon gas to reduce the probabil- gle CT scans that covered the whole thorax are
ity of respiratory difficulties. The inhalation performed at the end of the wash-in and washout
times vary depending on imaging protocols, and periods. Chae et al. (Chae et al. 2008) performed
it is recommended not to exceed 2 min, because a dynamic examination using xenon ventilation
most adverse effects have occurred after inhala- DECT in the limited area of the lung and analyzed
tion lasting longer than 4  min (Latchaw et  al. the DECT data with a single-compartment model
1987; Winkler et al. 1977). The inhalation times based on the Kety model, which assumes that the
190 H. J. Hwang et al.

xenon wash-in rate is equal to the xenon wash-out 2.2 Clinical Applications
rate. This study showed the diseased lung areas
showed the different dynamics of wash-out as 2.2.1 C  hronic Obstructive Pulmonary
well as wash-in xenon curves compared with the Disease
normal lung areas, showing an irregular delayed Chronic Obstructive Pulmonary Disease (COPD)
and redistributed pattern. As the different ventila- is characterized by airflow limitation that is
tion dynamics can be assessed depending upon caused by airway obstruction due to persistent
the scanning protocols, it is crucial to determine inflammation and emphysematous alveolar
the physiologic question to be evaluated in imple- destruction (Hogg et al. 2004; Mead et al. 1967).
menting a DECT imaging protocol using xenon or The diagnosis of COPD is based on clinical symp-
krypton gases. The appropriate protocols to assess toms and the results of pulmonary function test
regional ventilation change remain an important (PFT), and the severity of COPD is also graded on
topic for further investigation. the basis of the results of PFT.  Although obvi-
ously useful, PFT provides a global measure of
2.1.3 Postprocessing lung function inferred from primary structural
The image reconstruction system provides low- and functional alterations in the lung and does not
and high-kVp images and weighted average demonstrate the regional distribution of structural
images from the acquired data from both detec- and functional abnormalities. Conventional CT
tors. The weighted average images integrate both scan is also widely used in COPD patients to
low- and high-kVp data in a certain proportion assess the regional distribution of low attenuation
for diagnostic and morphologic assessment, and areas, but when assessed as a single full inspira-
the proportion of data from both acquisitions can tory scan, it is limited because it provides only
be adjusted using dedicated postprocessing soft- structural information of the lung parenchyma
ware. Usually, 30% image information from and airways. With the introduction of a dual-­
80-kVp series and 70% from 140-kVp series are energy CT technique, which can provide high-­
used to get an average weighted series corre- resolution anatomic information and functional
sponding to a 120-kVp image, and 50% for each information such as parenchymal ventilation,
140- and 100-kVp images. concurrently, many investigators have tried to
For the generation of ventilation maps, the evaluate pulmonary ventilation in COPD patients
dedicated dual-energy postprocessing software as well as structural changes using DECT (Fig. 5).
(Syngo Dual-Energy software; Siemens Chae et al. reported the first clinical study with
Healthcare) analyzes the density values in the xenon ventilation imaging using DECT in four
corresponding low- and high-kVp separate series patients with lung disease including COPD and
using a three-material decomposition algorithm eight healthy volunteers (Chae et al. 2008). The
for air, soft tissue, and xenon. By which, xenon study has reported the heterogeneous xenon
can be differentiated from other materials in the enhancement of the lung parenchyma in a patient
lung such as air and soft tissue, and the distribu- with COPD.  Park et  al. performed two-phase
tion of xenon within lung parenchyma is dis- (wash-in and wash-out phase) xenon ventilation
played with color-coded map. The preset material imaging with DECT in 32 COPD patients. This
parameters for xenon extraction are set to −1000 study has shown that regionally quantified value
HU for air at both photon energies, 60/56 HU at of xenon enhancement in low attenuating lung
80/100 kVp and 54/52 HU at 140/Sn140 kVp for areas on wash-in and wash-out phases showed
soft tissue, minimum value, −960 HU; and maxi- inverse correlation with PFT results, and the
mum value, −600 HU, a slope of 2.00 for 140/80 xenon value of low attenuating lung areas on
kVp or 2.18 for 100/Sn140 kVp. These parame- wash-out phase (suggesting abnormal xenon
ters are based on the empirical observation of retention in a diseased lung) showed a better cor-
dual-energy scans. relation with FEV1 than CT density based quanti-
Thoracic Imaging: Ventilation/Perfusion 191

a b

c d

Fig. 5  Xenon ventilation DECTs in a patient with COPD slightly decreased xenon enhancement at the periphery on
in 66-year-old male with mild emphysema (a, b) and xenon ventilation map. (c) Axial VNC image shows severe
58-year-old male with moderate to severe emphysema (c, emphysema with diffuse bronchial wall thickening in both
d). (a) Xenon ventilation DECT shows mild centrilobular lungs. (d) On the xenon ventilation map, multifocal areas
emphysema on axial VNC image and (b) homogeneous with decreased xenon enhancement are identified in both
xenon enhancement throughout both upper lobes but lungs

fication of emphysema severity (Park et al. 2010). CTs and the parenchymal attenuation changes
The authors categorized the lung ventilation between inspiration/expiration CTs.
abnormalities on two-phase xenon imaging into Hachulla et  al. have performed ventilation
four patterns. Visual analysis showed the areas DECT imaging using krypton gas in COPD
with emphysema exhibited iso-attenuation or patients. The maximum level of krypton enhance-
high attenuation in both wash-in and the wash-­ ment in the lung parenchyma was 18.5 HU. This
out phase, while the areas with predominant air- value is lower than that reported with xenon gas,
way disease showed low attenuation during the with an average maximum degree of xenon
wash-in period and various attenuation during the enhancement of 23.78 HU; however, it is suffi-
wash-out period. Thus, this categorization corre- cient to visualize ventilation abnormalities, with
lated well with conventional CT imaging findings significant differences of krypton attenuation
for the components of COPD. Lee et al. also per- between the normal and emphysematous lungs
formed two-phase (wash-in inspiration CT and (Hachulla et al. 2012).
wash-out expiration CT) xenon ventilation DECT
in 52 COPD patients (Lee et  al. 2017a). The 2.2.2 Asthma
authors compared between the xenon dynamic of Asthma is characterized by a reversible airway
wash-in and wash-out phase xenon ventilation obstructive disease with increased airway hyper-
192 H. J. Hwang et al.

responsiveness and chronic airway inflammation findings suggested that ventilation abnormalities
(ATS Board of Directors 1987). The clinical assessed using xenon-enhanced DECT may be
diagnosis of bronchial asthma is mainly based on used as novel parameters that reflect the asthma
clinical symptoms and pulmonary function test status and predict therapeutic responses.
results. Conventional CT imaging can show the
parenchymal changes including bronchial wall 2.2.3 Other Ventilation Related
thickening, varying degrees of air trapping, and Diseases: Asthma-COPD
airway mucus (Lee et al. 2004; Svenningsen et al. Overlap Syndrome,
2019), however, clinically, CT has been mainly Bronchiolitis Obliterans
used to evaluate the complications and associated Xenon ventilation DECT may be applicable to
conditions in patients with asthma. other obstructive airway diseases such as asthma-­
In the study by Chae et al., xenon ventilation COPD overlap syndrome (ACOS) and bronchiol-
imaging with DECT has shown the ventilation itis obliterans (BO). Although asthma and COPD
defects in the mid to lower lung peripheary, usu- have characteristic features, they share similar
ally pleural based, frequently wedge-shaped, and physiologic and clinical features (Gibson and
varied in size from tiny to segment in asthma Simpson 2009; Hardin et  al. 2014; Zeki et  al.
patients (Chae et  al. 2010b). The configuration 2011). Thus, the phenotype that shows persistent
and location of the ventilation defects on xenon airflow limitation with several features of both
ventilation DECT were similar to the description asthma and COPD is referred to as ACOS (Hwang
of ventilation defects on 3He-MR studies (Altes et  al. 2020). In 2017, the American Thoracic
et al. 2001; de Lange et al. 2006). In this study, Society and the National Heart, Lung, and Blood
ventilation defect scores are also correlated with Institute published a joint workshop report on
PFT results, including FEV1/forced vital capac- asthma-COPD overlap (Woodruff et  al. 2017).
ity (FVC), total lung capacity, functional residual Hwang et al. have evaluated the regional ventila-
volume, and residual volume. Xenon ventilation tion status in twenty-one ACOS patients using
DECT can also provide the changes in airflow xenon ventilation DECT, and to compare it to
dynamic in response to inhalation drugs such as that in COPD patients (Hwang et  al. 2020). In
methacholine or salbutamol in asthma patients. this study, most patients with ACOS showed the
In several studies, xenon ventilation DECT has peripheral wedge/diffuse defect on xenon maps,
demonstrated the changes in xenon ventilation which is frequently seen on various ventilation
map, including compensatory hyperventilation images performed in asthmatics, whereas patients
which occurred adjacent to the ventilation defects with COPD commonly showed the diffuse het-
after methacholine inhalation and improved erogeneous defect or lobar/segmental/subseg-
regional ventilation defects after inhalation of mental defect (Fig. 6). The airway wall thickening
bronchodilator in asthma patients (Goo and Yu (Pi10) and severity of emphysema (emphysema
2011; Kim et al. 2012). Also, in the study by Jung index) were also quantified on virtual non-­
et al., the authors evaluated the change in airflow contrast (VNC) images of xenon ventilation
dynamics after inhalation of methacholine and DECT, and the measured Pi10 was significantly
salbutamol in 43 non-smoking asthmatics (Jung higher in ACOS patients than in COPD patients,
et al. 2013). This study showed that xenon trap- while emphysema index was not significantly
pings in the wash-out phase after salbutamol different between two group. Xenon ventilation
inhalation were correlated with various clinical DECT may demonstrate the difference in the
symptoms including the asthma control test physiologic change of ventilation in ACOS
scores, wheezing, or night symptoms; whereas patients compared to COPD patients.
FEV1 showed no significant correlation with Bronchiolitis obliterans (BO) is a chronic
asthma symptoms. The degrees of ventilation obstructive lung disease following an injury to
defects were also positively correlated with FEV1 the small airways due to various etiologies
improvement after 3 months of treatment. These including lower respiratory infection, organ
Thoracic Imaging: Ventilation/Perfusion 193

a b

c d

Fig. 6  Xenon ventilation DECT in a 54-year-old man xenon ventilation maps show multifocal wedge-shaped or
with ACOS. (a, b) Axial and coronal VNC images of patchy areas showing blue-to-purple color in peripheral
DECT show diffuse bronchial wall thickening with mini- lung areas, indicating ventilation defects
mal centrilobular emphysema. (c, d) Axial and coronal

transplantation, and others. BO is usually diag- tions would increase the radiation exposure. Goo
nosed based on the typical clinical presentation, et  al. performed xenon ventilation DECT in 17
the finding of fixed airway obstruction in pulmo- children with BO (Goo et al. 2010). They demon-
nary function tests, and characteristic CT find- strated heterogeneously impaired ventilation in
ings. Conventional CT image is quite sensitive lung regions with BO on xenon ventilation map.
and specific in the evaluation of BO, showing Additionally, measured xenon and CT density
characteristic CT findings including bronchial values showed a positive correlation for the
wall thickening, central bronchiectasis, centri- whole lung and the hyperlucent regions. The
lobular opacities, and mosaic parenchymal atten- indexed volumes and volume percentages of
uation due to air trapping. However, for the hyperlucent areas and xenon ventilation defects
comprehensive and accurate evaluation of BO, have shown correlation with PFTs, including
paired inspiratory and expiratory chest CT scans FEV1, FEV1/FVC, and forced mid expiratory
may be required, and such paired CT examina- flow rate. Thus, xenon-enhanced DECT can help
194 H. J. Hwang et al.

identify regional ventilation defects as well as gible and does not warrant correction (Hoag et al.
morphologic abnormalities of BO without addi- 2007). Then, for evaluating the ventilation–per-
tional radiation exposure. fusion relationship, registration of VNC image of
perfusion CT to VNC image of ventilation CT is
performed, and with this information ventilation,
3 Comprehensive Assessment map, registered perfusion map, and then ventila-
of Morphology, Pulmonary tion/perfusion ratio (V/Qratio) map area are gen-
Ventilation, Perfusion, erated (Fig. 7).
and Relationship Recently, Sauter et al. have tried to simultane-
of Ventilation and Perfusion ously evaluate pulmonary ventilation and perfu-
Using Dual-Energy CT sion during a single CT scan in an animal model
using three-material differentiation in a dual-­
3.1 Concept layer CT scan (IQon Spectral CT, Philips
Healthcare, Netherlands) (Sauter et al. 2019).
With the introduction of DECT, pulmonary
parenchymal perfusion and ventilation impair-
ment have been evaluated independently with 3.2 Clinical Application:
DECT in various lung diseases. However, pulmo- Pulmonary Embolism, COPD
nary ventilation and perfusion often change con-
currently, and both ventilation and perfusion are There are few studies on the clinical application
crucial for efficient gas exchange. Combined of combined ventilation and perfusion imaging
ventilation–perfusion DECT can provide com- with DECT.  Thieme et  al. have reported the
prehensive information of regional ventilation, potential of DECT to provide both pulmonary
perfusion, and ventilation–perfusion relationship ventilation and perfusion imaging by investigat-
as well as high-resolution anatomic information ing ten patients with various diseases (e.g., lung
in various pulmonary diseases. transplantation, acute respiratory distress syn-
Based on our own experience and the results drome, aortic dissection, gastric cancer) from an
previously reported by Thieme et  al., combined anesthesiological intensive care unit (Thieme
xenon ventilation and perfusion DECT can be et al. 2010). Zhang et al. have applied this exami-
successfully performed using the dual-source CT nation in patients with suspected pulmonary
scanners (Somatom Definition scanner and embolism (Zhang et al. 2013). In that study, ven-
Somatom Definition FLASH, Siemens tilation and PBV maps have shown the ventila-
Healthcare, Forchheim, Germany) (Hwang et al. tion–perfusion mismatch in most areas with
2016, 2019; Lee et al. 2017b; Thieme et al. 2010). pulmonary embolism, and the authors hypothe-
For evaluating both ventilation and perfusion sized that the combined ventilation and perfusion
using DECT, the xenon ventilation DECT is per- DECT might provide higher sensitivity for
formed, followed by iodine-contrast perfusion detecting peripheral pulmonary embolism com-
DECT. After xenon ventilation DECT and before pared to conventional CT pulmonary
perfusion DECT, patients should inhale room air angiography.
for more than 10  min to wash out xenon gas. In COPD patients, the imbalances between
Because xenon is moderately soluble, the ques- alveolar ventilation and pulmonary blood flow
tion may arise about whether vascular uptake and are the critical features that result in arterial
redistribution back to the lung of inhaled xenon hypoxemia. Thus, combined ventilation–perfu-
may affect the background level lung density of sion DECT has the potential for the evaluation of
pulmonary PBV map. However, Hoag et al. have ventilation–perfusion imbalances as well as mor-
been demonstrated with the unilateral ventilation phologic changes in COPD patients. Hwang et al.
of xenon in intubated canines that the redistribu- have applied the combined xenon ventilation and
tion of inhaled xenon via the circulation is negli- iodine perfusion DECT in 52 COPD patients
Thoracic Imaging: Ventilation/Perfusion 195

Fig. 7  The graph shows the CT scanning protocol for (b) From ventilation and coregistered perfusion maps, a
combined xenon ventilation and contrast-enhanced perfu- coregistered ventilation/perfusion ratio map is generated
sion DECT. (a) Red and blue curves denote tidal xenon for evaluating the ventilation–perfusion relationship
and CO2 concentrations in the exhaled gas, respectively.
196 H. J. Hwang et al.

(Hwang et  al. 2016). In this study, ventilation, went bronchoscopic lung volume reduction,
perfusion, and ventilation/perfusion maps and DECT showed improved pulmonary ventilation
VNC images were used for the visual analysis of and ventilation–perfusion mismatch after bron-
regional ventilation, perfusion, and ventilation– choscopic lung volume reduction (Lee et  al.
perfusion relationships along with the morpho- 2017b). This combined DECT imaging has also
logic evaluation in COPD patients. On visual demonstrated change in the functional changes
analysis, in the normal appearing lung areas on after pharmacologic treatment in COPD patients
VNC images, parenchymal ventilation and perfu- (Hwang et  al. 2019). In this study, 52 COPD
sion were preserved, resulting in a matched ven- patients underwent combined ventilation and
tilation–perfusion relationship. However, in the perfusion DECT in the baseline and after 3
areas of bronchial wall thickening, ventilation months of pharmacologic treatment. The com-
was often decreased while perfusion was pre- bined DECT analysis showed improved ventila-
served, resulting in a ventilation–perfusion mis- tion–perfusion imbalance after the
match (reversed mismatch). In areas with pharmacological treatment in COPD patients,
emphysema, there were no dominant ventilation although the parenchymal disease patterns
or perfusion patterns. Furthermore, quantified remained unchanged (Fig.  8). This comprehen-
DECT parameters for ventilation, perfusion, and sive evaluation of parenchymal destruction and
ventilation–perfusion relationships are signifi- regional ventilation–perfusion relationships
cantly correlated with PFT results. Using this using DECT in COPD patients may provide use-
combined DECT imaging, the functional and ful information to the clinician, such as for the
physiologic changes after bronchoscopic lung assessment of response to medical treatment or
volume reduction in COPD patients have been the target lobe selection for lung volume reduc-
demonstrated. In 30 COPD patients who under- tion treatment.

a b

Fig. 8  A 63-year-old male COPD patient. On a VNC u­ p studies, resulting in the matched V/Q pattern on the V/
image of xenon ventilation CT, minimal centrilobular Qratio map in most areas of the right upper lobe. In the
emphysema and bronchial wall thickening are notable in posterior segment of the right upper lobe, ventilation
both upper lobes. (a) On baseline study, multifocal patchy remained decreased on ventilation map, while perfusion is
decreased ventilation is noted in the right upper lobe, preserved, resulting in the persistent reversed mismatched
while perfusion is preserved, resulting in the reversed V/Q on the V/Qratio map. In this patient, the percent pre-
mismatched V/Q on the V/Qratio map. (b) After 3 months dicted FEV1 increased by 20%
of medical treatment, ventilation is improved on follow-
Thoracic Imaging: Ventilation/Perfusion 197

4 Perspective and Conclusion sema susceptible smokers. Proc Natl Acad Sci U

S A 107(16):7485–7490. https://doi.org/10.1073/
pnas.0913880107
Ventilation or perfusion imaging using DECT is Altes TA, Powers PL, Knight-Scott J, Rakes G, Platts-­
a novel technique for perfusion and ventilation Mills TA, de Lange EE, Alford BA, Mugler JP,
assessment of various pulmonary diseases along Brookeman JR (2001) Hyperpolarized 3He MR lung
ventilation imaging in asthmatics: preliminary find-
with morphologic assessment with high-resolu- ings. J Magn Reson Imaging 13(3):378–384. https://
tion CT information without additional radiation doi.org/10.1002/jmri.1054
exposure. DECT analysis can help us diagnose Ameli-Renani S, Ramsay L, Bacon JL, Rahman F, Nair A,
disease, assess disease severity more accurately, Smith V, Baskerville K, Devaraj A, Madden B, Vlahos
I (2014) Dual-energy computed tomography in the
and understand unrevealed pathophysiologies assessment of vascular and parenchymal enhance-
related to lung function in various lung diseases. ment in suspected pulmonary hypertension. J Thorac
However, there is still insufficient evidence that Imaging 29(2):98–106. https://doi.org/10.1097/
DECT analysis can add clinically valuable infor- RTI.0000000000000061
ATS Board of Directors (1987) Standards for the diag-
mation in addition to anatomic information, nosis and care of patients with chronic obstructive
change the treatment plans, and predict therapeu- pulmonary disease (COPD) and asthma. Am Rev
tic response or prognosis. Therefore, further Respir Dis 136(1):225–244. https://doi.org/10.1164/
research will be needed to solidify the evidence ajrccm/136.1.225
Barbera JA, Peinado VI, Santos S (2003) Pulmonary
of the values and roles of DECT imaging meth- hypertension in chronic obstructive pulmonary
ods for improved diagnosis and outcomes assess- ­disease. Eur Respir J 21(5):892–905. https://doi.org/
ment in diseases associated with the lung function 10.1183/09031936.03.00115402
alteration. More investigations will also be Baxa J, Vondrakova A, Matouskova T, Ruzickova O,
Schmidt B, Flohr T, Sedlmair M, Ferda J (2014)
needed to establish appropriate imaging proto- Dual-phase dual-energy CT in patients with lung
cols that can be applied to a clinical diagnostic cancer: assessment of the additional value of iodine
work-up. With the continued development of quantification in lymph node therapy response. Eur
image acquisition protocols, post-processing Radiol 24(8):1981–1988. https://doi.org/10.1007/
s00330-­014-­3223-­9
techniques, and scanner designs, DECT will be Baxa J, Matouskova T, Krakorova G, Schmidt B, Flohr
applied widely to assessing various lung diseases T, Sedlmair M, Bejcek J, Ferda J (2016) Dual-phase
in clinical practices and research. dual-energy CT in patients treated with erlotinib for
advanced non-small cell lung cancer: possible ben-
efits of iodine quantification in response assessment.
Compliance with Ethical Standards
Eur Radiol 26(8):2828–2836. https://doi.org/10.1007/
s00330-­015-­4092-­6
Funding None. Bedi A, McCarroll C, Murray JM, Stevenson MA, Fee JP
(2002) The effects of subanaesthetic concentrations
Disclosure of Interests  All authors declare they have no of xenon in volunteers. Anaesthesia 57(3):233–241.
conflict of interest. https://doi.org/10.1046/j.0003-­2409.2001.02455.x
Brunelli A, Fianchini A (1997) Predicted postoperative
Studies involving human FEV1 and complications in lung resection candidates.
Chest 111(4):1145–1146. https://doi.org/10.1378/
Ethical Approval  This article does not contain any spe- chest.111.4.1145
cific studies with human participants performed by any of Chae EJ, Seo JB, Goo HW, Kim N, Song KS, Lee SD,
the authors. Hong SJ, Krauss B (2008) Xenon ventilation CT with
a dual-energy technique of dual-source CT: initial
experience. Radiology 248(2):615–624. https://doi.
Informed Consent  Informed consent was not needed for
org/10.1148/radiol.2482071482
this chapter.
Chae EJ, Seo JB, Jang YM, Krauss B, Lee CW, Lee HJ,
Song KS (2010a) Dual-energy CT for assessment
of the severity of acute pulmonary embolism: pul-
References monary perfusion defect score compared with CT
angiographic obstruction score and right ventricular/
left ventricular diameter ratio. AJR Am J Roentgenol
Alford SK, van Beek EJ, McLennan G, Hoffman EA
194(3):604–610. https://doi.org/10.2214/AJR.09.2681
(2010) Heterogeneity of pulmonary perfusion as
Chae EJ, Seo JB, Lee J, Kim N, Goo HW, Lee HJ, Lee
a mechanistic image-based phenotype in emphy-
CW, Ra SW, Oh YM, Cho YS (2010b) Xenon venti-
198 H. J. Hwang et al.

lation imaging using dual-energy computed tomog- Krypton ventilation imaging using dual-energy CT in
raphy in asthmatics: initial experience. Investig chronic obstructive pulmonary disease patients: initial
Radiol 45(6):354–361. https://doi.org/10.1097/ experience. Radiology 263(1):253–259. https://doi.
RLI.0b013e3181dfdae0 org/10.1148/radiol.12111211
Chae EJ, Kim N, Seo JB, Park JY, Song JW, Lee HJ, Hardin M, Cho M, McDonald ML, Beaty T, Ramsdell
Hwang HJ, Lim C, Chang YJ, Kim YH (2013) J, Bhatt S, van Beek EJ, Make BJ, Crapo JD,
Prediction of postoperative lung function in patients Silverman EK, Hersh CP (2014) The clinical and
undergoing lung resection: dual-energy perfusion genetic features of COPD-asthma overlap syn-
computed tomography versus perfusion scintigraphy. drome. Eur Respir J 44(2):341–350. https://doi.
Invest Radiol. 48(8):622–627. org/10.1183/09031936.00216013
Chon D, Beck KC, Simon BA, Shikata H, Saba OI, Herbert DL, Gur D, Shabason L, Good WF, Rinaldo
Hoffman EA (1985) Effect of low-xenon and kryp- JE, Snyder JV, Borovetz HS, Mancici MC (1982)
ton supplementation on signal/noise of regional Mapping of human local pulmonary ventila-
CT-based ventilation measurements. J Appl Physiol tion by xenon enhanced computed tomography. J
102(4):1535–1544. https://doi.org/10.1152/ Comput Assist Tomogr 6(6):1088–1093. https://doi.
japplphysiol.01235.2005 org/10.1097/00004728-­198212000-­00006
Chung YE, Hong SR, Lee MJ, Lee M, Lee HJ (2014) Hoag JB, Fuld M, Brown RH, Simon BA (2007)
Krypton-enhanced ventilation CT with dual energy Recirculation of inhaled xenon does not alter lung
technique: experimental study for optimal krypton CT density. Acad Radiol 14(1):81–84. https://doi.
concentration. Exp Lung Res 40(9):439–446. https:// org/10.1016/j.acra.2006.10.012
doi.org/10.3109/01902148.2014.946630 Hogg JC, Chu F, Utokaparch S, Woods R, Elliott WM,
de Lange EE, Altes TA, Patrie JT, Gaare JD, Knake JJ, Buzatu L, Cherniack RM, Rogers RM, Sciurba FC,
Mugler JP, Platts-Mills TA (2006) Evaluation of asthma Coxson HO, Pare PD (2004) The nature of small-­
with hyperpolarized helium-3 MRI: correlation with airway obstruction in chronic obstructive pulmonary
clinical severity and spirometry. Chest 130(4):1055– disease. N Engl J Med 350(26):2645–2653. https://
1062. https://doi.org/10.1378/chest.130.4.1055 doi.org/10.1056/NEJMoa032158
Gibson PG, Simpson JL (2009) The overlap syndrome Honda N, Osada H, Watanabe W, Nakayama M,
of asthma and COPD: what are its features and how Nishimura K, Krauss B, Otani K (2012) Imaging of
important is it? Thorax 64(8):728–735. https://doi. ventilation with dual-energy CT during breath hold
org/10.1136/thx.2008.108027 after single vital-capacity inspiration of stable xenon.
Goo HW, Yu J (2011) Redistributed regional ventilation Radiology 262(1):262–268. https://doi.org/10.1148/
after the administration of a bronchodilator demon- radiol.11110569
strated on xenon-inhaled dual-energy CT in a patient Hong YJ, Kim JY, Choe KO, Hur J, Lee HJ, Choi BW,
with asthma. Korean J Radiol 12(3):386–389. https:// Kim YJ (2013) Different perfusion pattern between
doi.org/10.3348/kjr.2011.12.3.386 acute and chronic pulmonary thromboembolism:
Goo HW, Chae EJ, Seo JB, Hong SJ (2008) Xenon ven- evaluation with two-phase dual-energy perfusion
tilation CT using a dual-source dual-energy tech- CT.  AJR Am J Roentgenol 200(4):812–817. https://
nique: dynamic ventilation abnormality in a child with doi.org/10.2214/AJR.12.8697
bronchial atresia. Pediatr Radiol 38(10):1113–1116. Hong SR, Chang S, Im DJ, Suh YJ, Hong YJ, Hur J, Kim
https://doi.org/10.1007/s00247-­008-­0914-­x YJ, Choi BW, Lee HJ (2016) Feasibility of single scan
Goo HW, Yang DH, Hong SJ, Yu J, Kim BJ, Seo JB, for simultaneous evaluation of regional Krypton and
Chae EJ, Lee J, Krauss B (2010) Xenon ventilation iodine concentrations with dual-energy CT: an experi-
CT using dual-source and dual-energy technique in mental study. Radiology 281(2):597–605. https://doi.
children with bronchiolitis obliterans: correlation of org/10.1148/radiol.16152429
xenon and CT density values with pulmonary function Hwang HJ, Seo JB, Lee SM, Kim N, Oh SY, Lee JS, Lee
test results. Pediatr Radiol 40(9):1490–1497. https:// SW, Oh YM (2016) Assessment of regional xenon
doi.org/10.1007/s00247-­010-­1645-­3 ventilation, perfusion, and ventilation-­ perfusion
Gorgos A, Remy-Jardin M, Duhamel A, Faivre JB, mismatch using dual-energy computed tomog-
Tacelli N, Delannoy V, Remy J (2009) Evaluation of raphy in chronic obstructive pulmonary disease
peripheral pulmonary arteries at 80 kV and at 140 kV: patients. Investig Radiol 51(5):306–315. https://doi.
dual-energy computed tomography assessment in 51 org/10.1097/RLI.0000000000000239
patients. J Comput Assist Tomogr 33(6):981–986. Hwang HJ, Hoffman EA, Lee CH, Goo JM, Levin DL,
https://doi.org/10.1097/RCT.0b013e3181a5cd0f Kauczor HU, Seo JB (2017) The role of dual-energy
Gur D, Drayer BP, Borovetz HS, Griffith BP, Hardesty computed tomography in the assessment of pulmo-
RL, Wolfson SK (1979) Dynamic computed tomogra- nary function. Eur J Radiol 86:320–334. https://doi.
phy of the lung: regional ventilation measurements. J org/10.1016/j.ejrad.2016.11.010
Comput Assist Tomogr 3(6):749–753 Hwang HJ, Lee SM, Seo JB, Lee JS, Kim N, Kim C, Oh
Hachulla AL, Pontana F, Wemeau-Stervinou L, Khung SY, Lee SW (2019) Assessment of changes in regional
S, Faivre JB, Wallaert B, Cazaubon JF, Duhamel A, xenon-ventilation, perfusion, and ventilation-­perfusion
Perez T, Devos P, Remy J, Remy-Jardin M (2012) mismatch using dual-energy computed tomography
Thoracic Imaging: Ventilation/Perfusion 199

after pharmacological treatment in patients with chronic Lee SM, Seo JB, Hwang HJ, Kim N, Oh SY, Lee JS, Lee
obstructive pulmonary disease: visual and quantitative SW, Oh YM, Kim TH (2017a) Assessment of regional
analysis. Int J Chron Obstruct Pulmon Dis 14:2195– emphysema, air-trapping and Xenon-­ventilation using
2203. https://doi.org/10.2147/COPD.S210555 dual-energy computed tomography in chronic obstruc-
Hwang HJ, Lee SM, Seo JB, Lee JS, Kim N, Lee SW, tive pulmonary disease patients. Eur Radiol 27(7):2818–
Oh YM (2020) Visual and quantitative assessments of 2827. https://doi.org/10.1007/s00330-­016-­4657-­z
regional xenon-ventilation using dual-energy CT in Lee SW, Lee SM, Shin SY, Park TS, Oh SY, Kim N,
asthma-chronic obstructive pulmonary disease overlap Hong Y, Lee JS, Oh YM, Lee SD, Seo JB (2017b)
syndrome: a comparison with chronic obstructive pul- Improvement in ventilation-perfusion mismatch after
monary disease. Korean J Radiol 21(9):1104–1113. bronchoscopic lung volume reduction: quantitative
https://doi.org/10.3348/kjr.2019.0936 image analysis. Radiology 285(1):250–260. https://
Iyer KS, Newell JD Jr, Jin D, Fuld MK, Saha PK, doi.org/10.1148/radiol.2017162148
Hansdottir S, Hoffman EA (2016) Quantitative dual-­ Mead J, Turner JM, Macklem PT, Little JB (1967)
energy computed tomography supports a vascular eti- Significance of the relationship between lung recoil and
ology of smoking-induced inflammatory lung disease. maximum expiratory flow. J Appl Physiol 22(1):95–
Am J Respir Crit Care Med 193(6):652–661. https:// 108. https://doi.org/10.1152/jappl.1967.22.1.95
doi.org/10.1164/rccm.201506-­1196OC Meinel FG, Graef A, Bamberg F, Thieme SF, Schwarz
Jung JW, Kwon JW, Kim TW, Lee SH, Kim KM, Kang F, Sommer WH, Neurohr C, Kupatt C, Reiser MF,
HR, Park HW, Lee CH, Goo JM, Min KU, Cho SH Johnson TR (2013a) Effectiveness of automated quan-
(2013) New insight into the assessment of asthma tification of pulmonary perfused blood volume using
using xenon ventilation computed tomography. Ann dual-energy CTPA for the severity assessment of acute
Allergy Asthma Immunol 111(2):90–95. https://doi. pulmonary embolism. Investig Radiol 48(8):563–569.
org/10.1016/j.anai.2013.04.019 https://doi.org/10.1097/RLI.0b013e3182879482
Kim WW, Lee CH, Goo JM, Park SJ, Kim JH, Park EA, Meinel FG, Graef A, Thieme SF, Bamberg F, Schwarz
Cho SH (2012) Xenon-enhanced dual-energy CT of F, Sommer WH, Helck AD, Neurohr C, Reiser MF,
patients with asthma: dynamic ventilation changes Johnson TR (2013b) Assessing pulmonary p­ erfusion
after methacholine and salbutamol inhalation. AJR in emphysema: automated quantification of per-
Am J Roentgenol 199(5):975–981. https://doi. fused blood volume in dual-energy CTPA.  Investig
org/10.2214/AJR.11.7624 Radiol 48(2):79–85. https://doi.org/10.1097/
Kyoyama H, Hirata Y, Kikuchi S, Sakai K, Saito Y, RLI.0b013e3182778f07
Mikami S, Moriyama G, Yanagita H, Watanabe W, Moon JW, Bae JP, Lee HY, Kim N, Chung MP, Park
Otani K, Honda N, Uematsu K (2017) Evaluation of HY, Chang Y, Seo JB, Lee KS (2016) Perfusion- and
pulmonary function using single-breath-hold dual-­ pattern-based quantitative CT indexes using contrast-­
energy computed tomography with xenon: Results of a enhanced dual-energy computed tomography in
preliminary study. Medicine 96(3):e5937. https://doi. diffuse interstitial lung disease: relationships with
org/10.1097/MD.0000000000005937 physiologic impairment and prediction of prognosis.
Latchaw RE, Yonas H, Pentheny SL, Gur D (1987) Eur Radiol 26(5):1368–1377. https://doi.org/10.1007/
Adverse reactions to xenon-enhanced CT cerebral s00330-­015-­3946-­2
blood flow determination. Radiology 163(1):251–254. Murphy DM, Nicewicz JT, Zabbatino SM, Moore RA
https://doi.org/10.1148/radiology.163.1.3823444 (1989) Local pulmonary ventilation using nonradio-
Lee YM, Park JS, Hwang JH, Park SW, Uh ST, Kim active xenon-enhanced ultrafast computed tomogra-
YH, Park CS (2004) High-resolution CT findings in phy. Chest 96(4):799–804. https://doi.org/10.1378/
patients with near-fatal asthma: comparison of patients chest.96.4.799
with mild-to-severe asthma and normal control sub- Okada M, Kunihiro Y, Nakashima Y, Nomura T, Kudomi
jects and changes in airway abnormalities following S, Yonezawa T, Suga K, Matsunaga N (2015) Added
steroid treatment. Chest 126(6):1840–1848. https:// value of lung perfused blood volume images using
doi.org/10.1378/chest.126.6.1840 dual-energy CT for assessment of acute pulmonary
Lee CW, Seo JB, Song JW, Kim MY, Lee HY, Park YS, embolism. Eur J Radiol 84(1):172–177. https://doi.
Chae EJ, Jang YM, Kim N, Krauss B (2011) Evaluation org/10.1016/j.ejrad.2014.09.009
of computer-aided detection and dual energy software Pansini V, Remy-Jardin M, Faivre JB, Schmidt B,
in detection of peripheral pulmonary embolism on dual- Dejardin-Bothelo A, Perez T, Delannoy V, Duhamel
energy pulmonary CT angiography. Eur Radiol 21(1):54– A, Remy J (2009) Assessment of lobar perfusion in
62. https://doi.org/10.1007/s00330-­010-­1903-­7 smokers according to the presence and severity of
Lee CW, Seo JB, Lee Y, Chae EJ, Kim N, Lee HJ, Hwang emphysema: preliminary experience with dual-energy
HJ, Lim CH (2012) A pilot trial on pulmonary emphy- CT angiography. Eur Radiol 19(12):2834–2843.
sema quantification and perfusion mapping in a single-­ https://doi.org/10.1007/s00330-­009-­1475-­6
step using contrast-enhanced dual-energy computed Park EA, Goo JM, Park SJ, Lee HJ, Lee CH, Park CM,
tomography. Investig Radiol 47(1):92–97. https://doi. Yoo CG, Kim JH (2010) Chronic obstructive pulmo-
org/10.1097/RLI.0b013e318228359a nary disease: quantitative and visual ventilation pattern
200 H. J. Hwang et al.

analysis at xenon ventilation CT performed by using the pulmonary parenchyma. AJR Am J Roentgenol
a dual-energy technique. Radiology 256(3):985–997. 193(1):144–149. https://doi.org/10.2214/AJR.08.1653
https://doi.org/10.1148/radiol.10091502 Thieme SF, Hoegl S, Nikolaou K, Fisahn J, Irlbeck M,
Park TS, Hong Y, Lee JS, Lee SM, Seo JB, Oh YM, Lee Maxien D, Reiser MF, Becker CR, Johnson TR
SD, Lee SW (2014) Efficacy of bronchoscopic lung (2010) Pulmonary ventilation and perfusion imaging
volume reduction by endobronchial valves in patients with dual-energy CT.  Eur Radiol 20(12):2882–2889.
with heterogeneous emphysema: report on the first https://doi.org/10.1007/s00330-­010-­1866-­8
Asian cases. J Korean Med Sci 29(10):1404–1410. Thieme SF, Graute V, Nikolaou K, Maxien D, Reiser MF,
https://doi.org/10.3346/jkms.2014.29.10.1404 Hacker M, Johnson TR (2012a) Dual energy CT lung
Pierce RJ, Copland JM, Sharpe K, Barter CE (1994) perfusion imaging–correlation with SPECT/CT.  Eur
Preoperative risk evaluation for lung cancer resection: J Radiol 81(2):360–365. https://doi.org/10.1016/j.
predicted postoperative product as a predictor of surgi- ejrad.2010.11.037
cal mortality. Am J Respir Crit Care Med 150(4):947– Thieme SF, Ashoori N, Bamberg F, Sommer WH,
955. https://doi.org/10.1164/ajrccm.150.4.7921468 Johnson TR, Leuchte H, Becker A, Maxien D, Helck
Remy-Jardin M, Faivre JB, Pontana F, Hachulla AL, AD, Behr J, Reiser MF, Nikolaou K (2012b) Severity
Tacelli N, Santangelo T, Remy J (2010) Thoracic appli- assessment of pulmonary embolism using dual energy
cations of dual energy. Radiol Clin N Am 48(1):193– CT-correlation of a pulmonary perfusion defect
205. https://doi.org/10.1016/j.rcl.2009.08.013 score with clinical and morphological parameters
Sakamoto A, Sakamoto I, Nagayama H, Koike H, of blood oxygenation and right ventricular failure.
Sueyoshi E, Uetani M (2014) Quantification of lung Eur Radiol 22(2):269–278. https://doi.org/10.1007/
perfusion blood volume with dual-energy CT: assess- s00330-­011-­2267-­3
ment of the severity of acute pulmonary thrombo- Tonner PH (2006) Xenon: one small step for anaesthe-
embolism. AJR Am J Roentgenol 203(2):287–291. sia...? Curr Opin Anaesthesiol 19(4):382–384. https://
https://doi.org/10.2214/AJR.13.11586 doi.org/10.1097/01.aco.0000236136.85356.13
Sauter AP, Hammel J, Ehn S, Achterhold K, Kopp FK, Winkler SS, Holden JE, Sackett JF, Flemming DC,
Kimm MA, Mei K, Laugerette A, Pfeiffer F, Rummeny Alexander SC (1977) Xenon and krypton as radiographic
EJ, Pfeiffer D, Noel PB (2019) Perfusion-ventilation inhalation contrast media with computerized tomog-
CT via three-material differentiation in dual-layer raphy: preliminary note. Investig Radiol 12(1):19–20.
CT: a feasibility study. Sci Rep 9(1):5837. https://doi. https://doi.org/10.1097/00004424-­197701000-­00007
org/10.1038/s41598-­019-­42330-­7 Woodruff PG, van den Berge M, Boucher RC, Brightling
Schmid-Bindert G, Henzler T, Chu TQ, Meyer M, Nance C, Burchard EG, Christenson SA, Han MK, Holtzman
JW Jr, Schoepf UJ, Dinter DJ, Apfaltrer P, Krissak MJ, Kraft M, Lynch DA, Martinez FD, Reddel HK,
R, Manegold C, Schoenberg SO, Fink C (2012) Sin DD, Washko GR, Wenzel SE, Punturieri A,
Functional imaging of lung cancer using dual energy Freemer MM, Wise RA (2017) Am J Respir Crit
CT: how does iodine related attenuation correlate Care Med 196(3):375–381. https://doi.org/10.1164/
with standardized uptake value of 18FDG-PET-CT? rccm.201705-­0973WS
Eur Radiol 22(1):93–103. https://doi.org/10.1007/ Yonas H, Grundy B, Gur D, Shabason L, Wolfson SK
s00330-­011-­2230-­3 Jr, Cook EE (1981) Side effects of xenon inhalation.
Svenningsen S, Haider E, Boylan C, Mukherjee M, J Comput Assist Tomogr 5(4):591–592. https://doi.
Eddy RL, Capaldi DPI, Parraga G, Nair P (2019) org/10.1097/00004728-­198108000-­00029
CT and functional MRI to evaluate airway mucus in Zeki AA, Schivo M, Chan A, Albertson TE, Louie S
severe asthma. Chest 155(6):1178–1189. https://doi. (2011) The asthma-COPD overlap syndrome: a
org/10.1016/j.chest.2019.02.403 common clinical problem in the elderly. J Allergy
Thieme SF, Becker CR, Hacker M, Nikolaou K, Reiser 2011:861926. https://doi.org/10.1155/2011/861926
MF, Johnson TR (2008) Dual energy CT for the assess- Zhang LJ, Zhou CS, Schoepf UJ, Sheng HX, Wu SY,
ment of lung perfusion–correlation to scintigraphy. Krazinski AW, Silverman JR, Meinel FG, Zhao YE,
Eur J Radiol 68(3):369–374. https://doi.org/10.1016/j. Zhang ZJ, Lu GM (2013) Dual-energy CT lung ven-
ejrad.2008.07.031 tilation/perfusion imaging for diagnosing pulmonary
Thieme SF, Johnson TR, Lee C, McWilliams J, Becker embolism. Eur Radiol 23(10):2666–2675. https://doi.
CR, Reiser MF, Nikolaou K (2009) Dual-energy CT org/10.1007/s00330-­013-­2907-­x
for the assessment of contrast material distribution in
 Thoracic Oncology

Philip Konietzke

Contents
1  rtifact Reduction and Improved Image Quality 
A  202
1.1  Metal Artifacts   202
1.2  Beam-Hardening Artifacts Due to Contrast Media   202
2 Imaging Pulmonary Nodules   203
2.1  I ntroduction   203
2.2  Detection of Pulmonary Nodules   204
2.3  Classification of Pulmonary Nodules   204
3 Imaging Lung Cancer   206
3.1  I ntroduction   206
3.2  Detecting Lung Cancer   206
3.3  Differentiate Lung Cancer and Inflammation   207
3.4  Characterization of Lung Cancer   208
3.5  Staging Lung Cancer   210
4  herapy Evaluation of Lung Cancer 
T  212
4.1  Introduction   212
4.2  Therapy Response with DECT   212
References   215

Abstract chromatic imaging (VMI), virtual non-contrast

or unenhanced imaging (VNC), iodine concen-
In recent years, dual-energy computed tomog- tration (IC) measurements, and the effective
raphy (DECT) is increasingly used in routine atomic number (Zeff). In thoracic oncology,
clinical practice due to technical developments. these applications might have additional bene-
DECT can provide both material-­nonspecific fits, in the diagnostic and therapy response
and material-specific energy-dependent infor- evaluation of lung cancer, especially in the con-
mation like virtual monoenergetic or mono- text with new treatments based on antiangioge-
netic agents and tyrosine kinase inhibitors.
P. Konietzke (*)
However, besides the problem that DECT
Department of Diagnostic and Interventional parameter values are not uniform across stud-
Radiology, University Hospital of Heidelberg, ies, the DECT technique also has some general
Heidelberg, Germany limitations. First, the use of DECT is limited in
e-mail: [email protected]

© The Author(s), under exclusive license to Springer Nature Switzerland AG 2022 201
H. Alkadhi et al. (eds.), Spectral Imaging, Medical Radiology Diagnostic Imaging,
https://doi.org/10.1007/978-3-030-96285-2_13
202 P. Konietzke

obese patients because of high image noise detect pulmonary nodules. In contrast, low VMIs
often interfering with structural and functional of 40  keV increased metal artifacts and worsen
image analysis. Second, larger image datasets the nodule detection, at least in thick slices, con-
require increased data storage capabilities and cluding that high monochromatic energy images
appropriate adjustment of clinical workflows. can reduce metal artifacts without nodule detec-
Photon-counting detector computed tion changes (Gyobu et al. 2013). Another group
tomography (PCDCT) is an emerging technol- used specially designed phantoms containing
ogy, using energy-resolving X-ray detectors dental, spine, and hip implants. They tested a
that differ substantially from conventional MAR algorithm and two types of VMI recon-
energy-integrating detectors. The results of structions separately and in combination to iden-
initial studies indicate its potentially high tify the optimal technique for each implant site.
impact on lung imaging by correcting for arti- Both methods significantly improved delineation
facts, having an excellent spatial resolution, accuracy, but the combination of 130 keV VMIs
reducing radiation exposure, and optimizing with MAR showed the best overall results.
the use of contrast agents. However, till a However, the optimal reconstruction technique
broad use in clinical routines, further research depends on the type of metal implant (Kovacs
is necessary. et al. 2018). Finally, Liu et al. evaluated the clini-
cal value of VMI combined with a MAR algo-
rithm to assess small pulmonary nodules after the
placement of microcoils before video-assisted
1 Artifact Reduction thoracoscopic surgery (VATS). They reported
and Improved Image Quality that 74 keV was the optimal level for VMIs since
it improved the image quality for microcoil local-
1.1 Metal Artifacts ization in pulmonary nodules (Liu et al. 2018a).
However, the presented results indicate that each
The photon absorption of metal objects leads to situation’s most suitable energy level is different
prominent metal- and beam-hardening artifacts. depending on whether artifact reduction or
In thoracic oncology, these artifacts may hamper improved signal-to-noise ratio is preferred.
tumor delineation when patients have metal Therefore, the energy level needs to be adjusted
objects such as spinal stabilization implants close depending on the situation.
to the treatment area, causing problems in follow- PCDCT can also reduce metal- and beam-­
­up situations or planning radiation therapy. hardening artifacts by relying on multiple energy
With its virtual monochromatic imaging bins (Shikhaliev 2005). However, there are cur-
(VMI), DECT can improve image quality by rently only limited data available addressing arti-
decreasing metal or beam-hardening artifacts. fact reduction in the specific setting of thoracic
However, if there is no desire to obtain material-­ oncology imaging.
specific information or correct metal or beam-­
hardening artifacts, it is better to perform a
conventional single-energy scan at the optimal 1.2 Beam-Hardening Artifacts
tube potential (Yu et al. 2012). VMI can be com- Due to Contrast Media
bined with a metal artifact reduction (MAR)
algorithm to improve image quality further. Three DECT VMI can also reduce beam-hardening arti-
studies investigated VMI and MAR for artifact facts caused by contrast media in the thorax, par-
reduction in the thorax. Gyobu et  al. placed 12 ticularly when assessing lesions in the upper
simulated nodules randomly inside a chest phan- mediastinum adjacent to the brachiocephalic vein
tom with a pacemaker. They reported that high (Fig. 1). In a retrospective study with 101 patients,
energy VMIs of 100 or 140  keV reduced metal Kim et al. determined the optimal energy level of
artifacts without influencing the capability to VMIs for reducing beam-hardening artifacts
Thoracic Oncology 203

a b c

d e f

Fig. 1  Image (a) shows a conventional polychormatic keV levels (70, 90, 110, 130, 150). The best contrast with
reconstruction of the upper thoracic apparatre with a con- significantly reduced beam-hardening artifacts is achieved
trast bolus in the left subclavian vein. Images (b–f) are at KeV levels of 100 to 130 KeV
showing monochromatic reconstructions with ascending

caused by contrast media. They compared VMIs gins. Malignancy is often indicated by deep
at different energy levels with conventional poly- lobulated or coarse spiculated margins or a max-
chromatic images and reported the best subjec- imum attenuation of 20–60 Hounsfield units
tive image quality and image noise at an energy (HU), while smooth borders and a maximum
level of 100 and 130 keV (Kim et al. 2018). attenuation of ≤15 HU make benignity more
likely (Ohno et al. 2014). However, there is con-
siderable overlap between benign and malignant
2 Imaging Pulmonary Nodules PNs in imaging features, and oversimplifying
can lead to misdiagnosis. Furthermore, the eval-
2.1 Introduction uation of contrast enhancement in pulmonary
lesions with ground-glass attenuation (GGA)
Conventional contrast-enhanced or non-enhanced remains challenging with conventional imaging
chest CT is the best imaging modality for diag- techniques.
nosing and characterizing pulmonary nodules However, there are some pitfalls when using
(PN). Medium to large lung pulmonary nodules DECT for PN evaluation. Harder et al. measured
are consistently detected on clinical CT systems, a total of 63 PN in 24 patients to assess the effect
but the interreader agreement and reader sensitiv- of different VMI reconstructions on iodine atten-
ity diminish substantially if the nodule size is uation and pulmonary nodule volumetry, show-
below 8–10  mm (Rubin 2015). Over the last ing a good correlation between nodule attenuation
decades, the case detection rates of pulmonary and nodule volume (R2  =  0.77). Furthermore,
nodules have continuously increased and might high energy VMI resulted in lower attenuation
rise even faster due to recent lung cancer screen- and nodule volumes, while low energy VMI
ing recommendations (Kauczor et  al. 2015). resulted in higher attenuation and nodule vol-
However, the correct management of detected umes, which is probably caused by differences in
pulmonary nodules is sometimes demanding. nodules’ peripheral enhancement at different
On CT, the diagnostic evaluation of PN usu- energy levels. Therefore, readers should consider
ally consists of two major parts: the degree and the possibility of over-or under-estimating pul-
pattern of contrast enhancement and the mor- monary nodules volume while using VMI (den
phologic features such as growth, size, and mar- Harder et al. 2017).
204 P. Konietzke

2.2 Detection of Pulmonary improved spatial resolution of the evaluated

Nodules PCDCT system could improve nodule growth
measurements and nodule shape characterization
Leng et  al. established an ultrahigh-resolution (Zhou et  al. 2017). Furthermore, SPCCT offers
(UHR) data collection mode on a whole-body, high-quality images at low radiation doses, which
research photon-counting detector computed is advantageous in low-dose lung cancer screen-
tomography (PCDCT) system with 64 rows of ing (Symons et al. 2017a).
0.45 mm × 0.45 mm detector pixels, which cor-
responded to a pixel size of 0.25  mm × 0.25  mm
at the isocenter (Leng et al. 2016). In a phantom 2.3 Classification of Pulmonary
study, the same group evaluated this UHR mode Nodules
to assess its performance in volume quantifica-
tion and shape differentiation for lung nodules. 2.3.1 Solid Pulmonary Nodules
Therefore, they scanned 20 synthetic lung nod- DECT offers several parameters derived from
ules with different sizes, shapes, and radio-­ iodine concentration (IC) measurements, provid-
densities to assess the influence of nodule ing additional information when evaluating pul-
properties and reconstruction kernels. The vol- monary nodules (PN). Several studies investigated
ume measurements of small or star-shape nod- values of DECT to differentiate benign from
ules were promising, demonstrating the malignant PN.
advantages of PCDCT as a robust quantitative In a pilot study, Xiao et al. assessed the combi-
tool for lung nodule characterization. Moreover, nation of low-dose DECT and ASIR (Adaptive
the results indicated that the UHR mode has a Statistical Iterative Reconstruction) algorithm to
high capability to differentiate sphere- from star-­ classify solitary pulmonary nodules (SPN). Sixty-
shaped nodules (Fig. 2). They concluded that the two patients with pathology-proved 42 benign

Fig. 2  Lung cancer in the left lower lung lobe showing CT scanner (Somatom Definition AS) using a I70 kernel
spiculae and cental cavitations. Left: Transverse photon-­ and achieving a slice thicksness of 1  mm. (Kindly pro-
counting detector CT images using a U70 kernel and vided by the DKFZ Heidelberg and the Thoraxklinik at
achieving a slice thickness of 0.25 mm. Right: Transverse University of Heidelberg)
images of the same lesion obtained with a conventional
Thoracic Oncology 205

and 20 malignant SPNs were scanned with arte- 2.3.2 Ground-Glass Pulmonary
rial and venous phase DECT.  The iodine and Nodules
water concentration (IC and WC), the normalized Pulmonary ground-glass attenuation (GGA) or
iodine and water concentration (NIC and NWC) lesions with GGA are still challenging with con-
of the lesions were measured, and the normalized ventional imaging techniques since the evalua-
iodine and water concentration difference (ICD tion of contrast enhancement is difficult.
and WCD) between the arterial and venous phases In a phantom study, Kawai et  al. used phan-
(AP and VP) were calculated. The spectral HU toms containing various iodine or calcium con-
curve was divided into three sections based on the centrations to simulate soft tissue and GGA to
energy (40–70, 70–100, and 100–140 keV), and examine the relationship between iodine concen-
the slopes (λHU) in both phases were calculated. tration and calculated iodine value contrast-­
The results showed that iodine-­related parameters mapping images (CMIs). They reported a good
(ICAP, ICVP, NICAP, NICVP, and the ICD) of malig- correlation between iodine value and iodine con-
nant SPN were significantly higher than that of centration in the soft tissue models (r2 = 0.996)
benign SPN and that the three venous phase λHU and the GGA models (R2  =  0.998). In the next
values in malignant SPN were higher than in step, they applied the technique to clinical cases
benign SPN (P < 0.05). The water-related param- with lung lesions, showing that contrast enhance-
eters showed no difference. Therefore, they con- ment on CMIs was visible in 22 a­ denocarcinomas
cluded that iodine parameters are useful markers but not in pulmonary hemorrhage and inflamma-
to distinguish benign from malignant lung dis- tory changes (Kawai et al. 2011).
eases (Xiao et al. 2015). Wu et al. also used mul- Liu et al. included 48 patients with lung ade-
tiple DECT parameters to assess the spatial nocarcinoma in a retrospective study who under-
distribution of NIC in 39 malignant and 21 benign went arterial phase DECT before treatment. The
pulmonary SPNs. For this purpose, they calcu- iodine concentration (IC) and water content
lated the difference (dNIC) between the proximal (WC) of the GGO were measured and compared
(NICpro) and the distal (NICdis) regions of the nod- to the contralateral and ipsilateral normal lung
ules, showing significant differences between tissue, finding significantly higher IC values in
malignant and benign nodules in the arterial and pGGO and mGGO (P < 0.001). Furthermore, IC,
venous phase (Wu et al. 2018a). NIC, and WC values were compared between
Lin et  al. analyzed 139 patients with groups of pure ground-glass opacity (pGGO),
pathology-­ proved SPNs who also underwent mixed ground-glass opacity (mGGO), preinva-
double-phase enhanced DECT scans. They sive lesions, minimally invasive adenocarcinoma
divided the patients into an active inflammatory, (MIA), and invasive adenocarcinoma (IA). The
a malignant, and a tuberculosis group. Normalized NIC and WC values for pGGO and mGGO, and
(NICs) and non-normalized iodine concentra- the WC values of the groups with preinvasive
tions (ICs) were derived from iodine-based mate- lesions and MIA and IA were statistically differ-
rial decomposition CT images, and the slope rate ent (P = 0.049, P < 0.001, P < 0.001) (Liu et al.
was calculated from the spectral curve. The 2018b).
results showed that the mean slope rate, IC, and Chen et al. quantitatively assessed the DECT
NIC for the active inflammatory group were sig- imaging’s efficacy for differentiation of benign
nificantly higher than for the malignant group, and malignant ground-glass nodules (GGN) and
and the parameters of the malignant group were solid nodules (SN). The study included 114
considerably higher than the tuberculosis group patients with SPNs (61 GGNs and 53 SNs) who
(P < 0.05) (Lin et al. 2016). underwent DECT plain and enhanced scans in
In summary, all three studies imply that DECT the arterial (a) and venous (v) phases. The spec-
imaging provides a novel method for a better tral CT imaging parameters included: iodine con-
characterization of pulmonary nodules in double-­ centrations (IC) of lesions in the arterial (ICLa)
phase contrast-enhanced scanning. and venous (ICLv) phases; normalized IC (NICa/
206 P. Konietzke

NICv), the slope of the spectral curve (λHUa/ oped world (Sung et al. 2021). Early-stage lung
λHUv), and VMI images on 40 and 70  keV cancer can be treated with potentially curative
(CT40keVa/v, CT70keVa/v). Pathology revealed intent, but most patients present at an advanced
75 lung cancer cases, three metastatic nodules, stage, which results in an overall low 5-year sur-
14 benign nodules, and 22 inflammatory nodules. vival for all stages. Furthermore, the staging of
Among the 53 SNs were 37 malignant and 16 lung cancer is essential because treatment options
benign nodules, while among the 61 GGNs were and prognosis differ significantly by stage.
41 malignant and 20 benign nodules. Overall, the Computed tomography (CT) plays an essen-
CT40keVa, λHUa, CT40keVv, λHUv, and ICLv tial role in noninvasively characterizing pulmo-
of benign SPNs were greater than those of malig- nary masses according to the morphology,
nant SPNs (all P < 0.05). For GGNs, CT40keVa/v, interfaces, inner densities, and enhancement of
CT70keVa/v, λHUa/λHUv, and ICLv of malig- masses (MacMahon et al. 2005). DECT has the
nant GGNs were all lower than those of benign potential to describe the contrast enhancement
GGNs (Chen et al. 2019). even further by depicting microvessel density
All three studies mentioned above showed and blood supply with IC measurements. In this
that DECT is a promising method for distinguish- context, significant correlations between the
ing malignant from benign GGN by indicating iodine uptake derived from DECT and perfusion
their blood supply status. parameters derived from first-pass dual-input
perfusion computed tomography (DIPCT) have
2.3.3 Calcified Pulmonary Nodules been reported (Chen et al. 2017). Several DECT
The DESCT technique allows data acquisition at parameters are available describing contrast
different kilovoltage settings, enabling recon- enhancement, which can help detect lung cancer,
structing virtual non-contrast (VNC) images by differentiate lung cancer from inflammation,
substracting the iodine content in contrast-­ classify lung cancer subtypes, and identify lymph
enhanced images with CT numbers similar to node metastasis:
true non-contrast imaging (TNC).
VNC has the potential to differentiate calcifi- 1. Iodine concentration (IC) in enhanced arterial
cations and strongly attenuating nodules without (AP) or venous phase (VP) images.
additional TNC, which can help characterize pul- 2. Normalized iodine concentration (NIC) as the
monary nodes (Fig. 3). Chae et al. detected 17 out ratio of the lesions IC and the IC of the artery
of 20 calcifications in SPN on VNC images with- at the same level.
out using TNC (Chae et al. 2008). Nevertheless, 3. The spectral attenuation curve slope (λHU)
VNC can underestimate calcification’s extent due calculated with spectral images from 40 to
to post-processing subtraction errors and influ- 140 keV (spaced at 10 keV intervals).
ence the measured volume of pulmonary nod-
ules. In this context, Harder et al. compared the
volume of pulmonary nodules on VNC and con- 3.2 Detecting Lung Cancer
ventional reconstructions, finding a significant
reduction of 5.5% (2.6–11.2%, P  <  0.001) on DECT techniques may improve lung cancer
VNC (den Harder et al. 2017). detection by decreasing image noise and increas-
ing the signal-to-noise and contrast-to-noise
ratio. For this purpose, two studies explored the
3 Imaging Lung Cancer optimal energy level of virtual monochromatic
imaging (VMI) to improve lung cancer imaging
3.1 Introduction quality. Hou et al. scanned 50 lung cancer patients
with DECT, generating monochromatic images
Lung cancer is the leading cause of cancer-related at 50, 60, 70, and 80 keV energy levels. Subjective
deaths for both men and women across the devel- assessment about the overall image quality and
Thoracic Oncology 207

Fig. 3  Partially calcified hamartoma in the right lower contrast reconstuction (VNC). Using VNC images, calci-
lobe. The left image shows a conventional polychromatic fications can be well separated from a possible nodular
reconstruction in the venous phase. The right image shows contrast enhancement within the lesion
the same slice with same windowing as virtual non-­

inhomogeneity enhancement was performed by CT numbers in the distal lung of 32 patients with
analyzing image noise, lesion-to-lung contrast-­ central lung cancer with the corresponding areas
to-­noise ratio, and CT number difference between in the normal contralateral lung. The results
central and peripheral regions of tumor (dCT showed that IC of the side with lung cancer
value). The highest contrast-to-noise ratio value (0.70 ± 0.42 mg/ml) was significantly lower than
and the best subjective score of image quality the corresponding area in the normal contralateral
were obtained at 70 keV (P < 0.05), whereas the lung (1.19 ± 0.62 mg/ml) (P < 0.001), indicating
highest subjective score of inhomogeneity evalu- that DECT is feasible and to identify perfusion
ation was at 60  keV (P  <  0.05). Therefore, the defects that are induced by central lung cancer
combination of 60 and 70  keV monochromatic (Sun et al. 2013).
images might be useful in lung cancer imaging
(Hou et al. 2016). Kaup et al. reported compara-
ble results. In their study, the scans of 59 lung 3.3  ifferentiate Lung Cancer
D
cancer patients who underwent chest DECT were and Inflammation
reconstructed at different energy levels (40, 60,
80, 100  keV) with VMI.  After assessing each Inflammatory masses with benign nature such as
reconstruction’s objective and subjective image granulomatous inflammation, focal organizing
qualities, the best performance for lung cancer pneumonia, and lung abscess are a common find-
imaging was reported at 60  keV (Kaup et  al. ing in chest CT. The differentiation against lung
2016). cancer is essential because inflammatory masses
Central lung cancer often involves lobar or seg- can be treated with high-dose steroids, irradia-
mental bronchi, causing ventilation impairment in tion, or antibiotics, and unnecessary pulmonary
the distal lung parenchyma. The close relationship resection should be avoided. DECT parameters
between pulmonary ventilation and perfusion may reflect the blood supply of lung masses.
also cause perfusion deficits in the corresponding Granulomatous inflammation and organizing
lung areas. Sun et al. investigated the capability of pneumonia are formed by the proliferation of
DECT to quantitatively evaluate lung perfusion inflammatory granulation tissue or acute inflam-
defects that are induced by central lung cancer. mation residuals. The inflammation may stimu-
They compared the iodine ­concentrations (IC) and late rich and dilated capillaries leading to a high
208 P. Konietzke

and homogenous enhancement (Maldonado et al. while the combination of these parameters
2007; Diederich et al. 2006). However, inappro- achieved a sensitivity and specificity of 100%
priate angiogenesis and establishment of vascular and 81.3% (Yu et al. 2019). In conclusion, these
networks may occur in lung cancer, leading to a studies indicate that DESCT imaging can help to
subsequent reduction in oxygen delivery and an differentiate lung cancers from inflammatory
insufficient supply of fast-growing masses, thus masses.
explaining the inhomogeneous enhancement in
lung cancer (Yi et al. 2004; Zhao et al. 2014).
Several studies investigate the value of DECT 3.4 Characterization of Lung
in differentiating malignant from inflammatory Cancer
pulmonary masses. Wang et al. evaluated the fea-
sibility of qualitative and quantitative informa- 3.4.1 Virtual Biopsy with DECT
tion from Gemstone Spectral Imaging (GSI) to The heterogeneity and complexity of lung cancer
differentiate lung cancer and benign lung lesions. are determined by genes that play a crucial role in
Sixty-eight patients, divided into pneumonia its occurrence, type, development, and prognosis
(n  =  24) and malignant tumor groups (n  =  44), (Vogelstein et al. 2013). The lung tumor classifi-
were compared by iodine concentration (IC), cation system of the World Health Organization
water concentration (WP), spectral curve slope, divides lung carcinomas into non-small cell lung
and CT numbers at 40 keV, and significant differ- carcinoma (NSCLC) and small cell lung
ences were found for all parameters (P  <  0.05) ­carcinoma (SCLC). NSCLCs are further classi-
(Wang et  al. 2014). Hou et  al. analyzed 60 fied into squamous cell carcinoma (SC), adeno-
patients with 35 lung cancers and 25 inflamma- carcinoma (AC), and large cell carcinoma (Travis
tory masses who underwent arterial phase (AP) et  al. 2015). The definitive diagnosis for lung
and venous phase (VP) DECT.  They measured mass is achieved via invasive histological exami-
the normalized iodine concentration (NIC), the nation, such as thoracoscopic surgery, biopsy via
slopes of spectral attenuation curves (λHU), and bronchoscopy, or transthoracic puncture guided
the CT numbers in 70 keV VMI in the central and by ultrasonography. However, a non-invasive and
the peripheral regions of lung masses and calcu- accurate method for evaluating the histological
lated the difference in CT numbers (dCT) type of lung cancer would be desirable since his-
between both regions. The results showed that tological examination may be associated with
NICs and CT numbers in the central and λHU complications in some cases.
values in the central and peripheral regions of Fehrenbach et al. showed that in 52 untreated
lung cancers were significantly lower than those primary NSCLC lesions, lung adenocarcinoma
of inflammatory masses. In contrast, dCT values had significantly higher normalized iodine con-
of lung cancers were higher than those of inflam- centrations than (NIC: 19.37) squamous cell car-
matory masses. (Hou et  al. 2015). Yu et  al. cinoma (NIC: 12.03; P  =  0.035) (Fehrenbach
explored the value of arterial phase (AP) and et al. 2019a). Jia et al. combined the quantitative
venous phase (VP) DECT in differentiating lung DECT parameters CT numbers, the slope of the
cancer from an inflammatory myofibroblastic spectral attenuation curve (λHU), iodine concen-
tumor (IMT) in 96 patients with lung cancer and tration, water concentration, and the effective
16 with IMT.  For this purpose, the normalized atomic number (Zeff) with serum tumor markers
iodine concentration in AP (NICAP) and VP to evaluate the lung cancer histology type. Tumor
(NICVP), the slope of the spectral curve in AP markers were the serum levels of carcinoembry-
(λAP) and VP (λVP), and the normalized iodine onic antigen (CEA), neuron-specific enolase
concentration difference between AP and VP (NSE), squamous cell carcinoma antigen
(ICD) were calculated and compared. A signifi- (SCC-Ag), and cytokeratin fragment
cantly higher NICAP, NICVP, λAP, λVP, and ICD was CYFRA21-­1. CEA and NSE levels were higher
found in IMT than in lung cancer (P  <  0.05), in adenocarcinoma and neuroendocrine tumors,
Thoracic Oncology 209

while SCC-Ag and CYFRA21-1 levels were were divided into an EGFR mutation and an
higher in squamous cell cancer. There was no sig- EGFR wild-type group. The evaluated DECT
nificant difference in CT number attenuation parameters were the CT numbers at 70 keV, nor-
among the groups (P > 0.05), whereas HUλ in the malized iodine concentration (NIC), normalized
arterial phase and Zeff and IC in the arterial and water concentration, and slopes of the spectral
venous phases was significantly different among attenuation curves (λHU). The univariate analy-
the groups (P < 0.05). The diagnostic efficiency sis revealed that sex, smoking history, NIC, and
of serum tumor markers was higher than that of slope λHU were significantly associated with
CT spectral parameters, but combining serum EGFR mutation status (P > 0.05). Furthermore,
markers and CT parameters showed a larger smoking history and NIC were the two signifi-
diagnostic efficiency than combined serum mark- cant predictive factors associated with EGFR
ers and combined CT parameters alone. mutations (OR = 3.23, P = 0.005; OR = 58.026,
Therefore, DECT parameters and serum tumor P = 0.049). Based on this analysis, the smoking
markers are valuable in evaluating histological history and NIC were combined to determine the
types of lung cancer, and in combination, they predictive value for EGFR mutations with the
can significantly improve diagnostic efficiency. area under the curve of 0.702 (Li et al. 2019). Li
Nevertheless, the accuracy of these two methods et  al. explored 48 patients with NSCLC who
alone is still insufficient (Jia et al. 2018). DECT underwent DECT before surgical tumor resec-
might also differentiate pulmonary metastases tion. They correlated the expression level of
from different primary origins. Deniffel et  al. vascular endothelial growth factor (VEGF) in
­
demonstrated significant differences in the IC non-small-cell lung cancer (NSCLC) with the
between pulmonary metastases of renal cell car- DECT quantitative imaging parameters IC and
cinoma, breast, colorectal, and head/neck carci- CT values at 40 keV and λHU. They found sig-
noma, as well as metastases of colorectal nificant differences in IC, λHU, and CT values at
carcinoma, osteosarcoma, pancreato-biliary, and 40  KeV between NSCLCs with negative and
urinary tract carcinoma (Deniffel et al. 2019). moderately positive VEGF expression (P = 0.001)
Angiogenesis is essential in the process of pri- and between NSCLCs with mildly and moder-
mary tumor growth, proliferation, and metastasis. ately positive expression of VEGF (P  =  0.047–
In many cancers, including non-small cell lung 0.002). Besides, all parameters displayed a
cancer (NSCLC), tumor angiogenesis pathways significant and positive correlation with the level
have been identified as important therapeutic tar- of VEGF expression (R2 = 0.458–0.393, P < 0.05)
gets. Wu et al. used DECT parameters to charac- (Li et al. 2016). All three studies imply that quan-
terize 60 histology confirmed lung squamous cell titative DECT parameters have the potential to
carcinoma (SC) and adenocarcinoma (AC) and predict EGFR and VEGF mutations in lung can-
correlated the findings with the expression of thy- cer, helping to evaluate the status of
roid transcription factor-1 (TTF-1) and epidermal angiogenesis.
growth factor receptor (EGFR). They noted no The grading of cancer is a histological method
significant differences in normalized iodine con- intended to help predict prognosis based on spe-
centration (NIC) and the spectral attenuation cific morphological features. It typically is based
curve slope (λHU) between the TTF-1-positive on architectural or cytological features (nuclear
and TTF-1-negative groups. However, significant grade or the number of mitoses), or in some
differences in NIC and λHU were noted between cases, a combination of both (Travis et al. 2016).
EGFR-positive and EGFR-negative SC and AC Lin et  al. investigated the correlation between
(Wu et al. 2018b). Li et al. explored the role of pathological grades of non-small cell lung can-
DECT in identifying epidermal growth factor cers (NSCLCs) and quantitative DECT parame-
receptor (EGFR) mutation status in a cohort of ters. They evaluated 53 patients with NSCLCs
120 patients with pulmonary adenocarcinoma, 66 who underwent preoperative DECT, dividing
with confirmed EGFR mutations. The patients them into a low-grade and a high-grade group
210 P. Konietzke

based on their histopathological differentiation. pericyte and smooth muscle tissue, increasing
They measured arterial phase (AP) and venous permeability of the blood vessels (Ruoslahti
phase (VP), iodine concentration (IC), the nor- 2002). Therefore, the use of different contrast
malized iodine concentration (NIC), and slope of agents simultaneously (e.g., purely intravascular
the spectral curve (λHU), showing that NIC and for the vascular supply and intra−/extracellular
λHU in the AP and VP were significantly higher for the capillary leak) might offer new possibili-
in the low-grade NSCLC than in the high-grade ties for lung tumors’ characterization. PCDCTs
NSCLC (P  <  0.001). Furthermore, there was a K-edge imaging enables the differentiation
significant negative correlation between DECT between multiple contrast agents, making it pos-
parameters and pathological grades (P < 0.001). sible to administer different contrast agents and
ROC analysis indicated that λHU in VP provided simultaneously show their specific distribution
the best diagnostic performance in distinguishing (Cormode et  al. 2017). Furthermore, multiple
high-grade from low-grade cancers with an AUC contrast agents could be administered at different
of 0.914, a sensitivity of 85.7%, and a specificity time points while imaging is performed at a sin-
of 84.4% (Lin et al. 2018). gle time point (Fornaro et al. 2011). The arterial
phase could be evaluated by removing the con-
3.4.2 PCDCT and Molecular Imaging trast agent administered first, and the portal phase
Molecular imaging using new types of targeted could be assessed by removing the second con-
contrast agents is among the exciting possibilities trast agent. Non-contrast images could be recon-
that may become a reality using PCDCT (Taguchi structed by removing the contrast agents from the
and Iwanczyk 2013; Jaffer and Weissleder 2004). images. Symons et  al. evaluated simultaneous
Molecular CT uses combined particles composed imaging in a canine model with orally adminis-
of a contrast agent labeled with nanoparticles, tered bismuth and intravenously injected gado-
larger than conventional contrast agents, remain- linium and iodine (Symons et al. 2017b, c). All
ing in the cardiovascular system longer than 24 h these techniques may significantly impact lung
(Fornaro et al. 2011). Moreover, these nanoparti- tumor characterization and may offer new meth-
cles can carry targeting molecules for specific ods in evaluating tumor behavior, possibly lead-
cells or enzymes, allowing target-specific imag- ing to a more accurate tumor prognosis. However,
ing. Research has focused on nanoparticles extensive studies are necessary till these tech-
labeled with gold, which can be detected with niques may be part of the clinical routine.
K-edge imaging even in the presence of other
contrast agents, such as iodine (Si-Mohamed
et  al. 2017; Cormode et  al. 2009; Roessl and 3.5 Staging Lung Cancer
Proksa 2007). In oncology, molecular CT may
aid in early cancer diagnosis by quantifying small 3.5.1 Lymph Node Metastasis
tumors’ mass and size and determining the distri- The correct staging of lung cancer is essential
bution of contrast agents and/or particles (Barber because treatment options and prognosis differ
et al. 2015). significantly by stage. However, conventional CT
In 1967, Milne showed that lung tumors might diagnostic criteria based on size (lymph node
have a dual blood supply by the pulmonary and maximal short-axis diameter ≥10  mm) are still
the aortic system, with a trend toward histologi- standard for judging metastatic lymph nodes,
cal type-specific circulatory patterns (Milne reaching accuracy of only about 60% (Silvestri
1967). Furthermore, tumor cells can produce et al. 2013).
angiogenic factors that stimulate and generate Yang et  al. investigated the value of DECT
many new blood vessels. These new blood ves- quantitative parameters for preoperative diagno-
sels’ wall is immature due to a lack of hemangio- sis of metastatic lymph nodes in patients with
Thoracic Oncology 211

non-small cell lung cancer (NSLC). They ment of the pulmonary vessels and the lymph
included 84 patients with suspected lung cancer, nodes is almost the same. Sekiguchi et al. evalu-
evaluating a total of 144 lymph nodes, of which ated the visibility of the hilar lymph nodes (LNs)
48 were metastatic, and 96 were non-metastatic. by comparing virtual monoenergetic low-keV
They measured the normalized iodine concentra- images with early-phase contrast-enhanced CT
tions (NIC), water concentration, and slope of the (Fig. 4). DECT was performed in 50 patients for
spectral curve (λHU) in the arterial and venous evaluation of lung cancer at 20 and 60  s after
phases. The λHU measured during both phases administration of contrast media, and five recon-
was significantly higher in metastatic than in structions were made each (A: 20  s/120  kV; B:
benign lymph nodes (P < 0.05). The λHU of the 60 s/40 keV; C: 60 s/50 keV; D: 60 s/120 kV; E:
arterial phase (AP) with an optimal threshold 60  s/100  kV). The authors measured the differ-
value of λHU of 2.75 reached a sensitivity of ences in CT numbers of the bilateral main pulmo-
88.2% and a specificity of 88.4% for metastatic nary arteries (PAs), pulmonary veins (PVs), and
lymph nodes, which is higher than for conven- hilar LNs and calculated the differences in CT
tional CT-based qualitative size assessment number between the PA/PV and LNs (PA-LN and
(Yang et al. 2017). PV-LN contrast). They also evaluated the arti-
Imaging the hilar lymph nodes using non-­ facts from the superior vena cava (SVC). The
contrast-­enhanced CT is difficult because X-ray results showed that virtual monoenergetic 40-keV
absorption by the hilar nodes is similar to that of imaging at the delayed 60-s phase was beneficial
the neighboring pulmonary arteries and veins. for evaluating hilar lymph nodes since it had
Consequently, the hilar LNs are commonly eval- good PA-LN and PV-LN contrasts with low arti-
uated using contrast-enhanced CT.  Still, even facts (Sekiguchi et al. 2019).
with contrast agents, the detection of hilar lymph- Virtual non-contrast imaging (VNC) can offer
adenopathy can be difficult when the enhance- additional information when differentiating cal-

Fig. 4  Small lymph node in the lower mediastinum with right image shows the same slice with same windowing as
contrast uptake. The left image shows a conventional 40 keV VMI. The lymph node is better delineated on the
polychromatic reconstruction in the venous phase. The VMI image
212 P. Konietzke

cifications and strongly enhancing lymph nodes 4  herapy Evaluation of Lung

T
without additional true non-contrast imaging Cancer
(TNC). Yoo et  al. evaluated VNC images’ reli-
ability compared with TNC images in determin- 4.1 Introduction
ing high CT attenuation or calcification of
mediastinal lymph nodes. Node attenuation in Computed tomography (CT) or magnetic reso-
TNC and VNC images was compared in a total of nance imaging (MRI) is usually used to evaluate
112 mediastinal nodes from 45 patients, objec- tumor response to therapy. Combined chemo-
tively and subjectively, via computed tomogra- therapy and thoracic radiation therapy (i.e.,
phy (CT) attenuation and visual scoring. CT chemoradiotherapy (CRT)) have proven their
attenuation in TNC and VNC had a mean abso- value in nonresectable NSCLC (Ramnath et  al.
lute difference of 7.8 ± 7.6 HU and an absolute 2013), but definitive external beam radiation
difference of equal to or less than 10 HU in 65.2% therapy is associated with local recurrence rates
of cases, which corresponds to a moderate intra- of 55% to 70% (Jones et al. 2015). The success of
class correlation coefficient of 0.612. The visual radiation therapy highly depends on tumor vascu-
scores in TNC and VNC images showed fair larization and is limited in hypoxic tumor tissue
agreement with a κ value of 0.335 (Yoo et  al. (Salem et al. 2018).
2013). The morphological criteria used by Response
Evaluation Criteria in Solid Tumors (RECIST)
3.5.2 Bone Metastasis 1.1 are limited to dimensional changes in tumor
Bone metastasis is a frequent finding in advanced size, whereas other criteria such as density evalu-
lung cancer patients, and the spine is the most ation, functional or metabolic changes are not
common location. The detection of a metastatic considered. However, radiation therapy can be
vertebral lesion in its early development stage is successful without immediately changing the
important because appropriate treatment can hin- morphology of treated masses. Furthermore, new
der complications due to advanced metastasis. response criteria like immune-related response
However, the diagnosis of inconspicuous osteo- criteria (irRC) and modifications of RECIST will
blastic metastases (OBM) from lung cancer is a be necessary due to new treatments based on
challenge in conventional CT images since they antiangiogenetic agents and tyrosine kinase
might have a similar density and are therefore not inhibitors. In this context, imaging techniques
distinguishable from normal vertebrae. Yue et al. such as positron emission tomography (PET-CT),
evaluated the optimal energy level of VMI in 35 diffusion-weighted MRI, and DECT may offer
patients to detect and diagnose OBMs of the ver- some of the promising applications.
tebra. The CT number and standard deviation
(SD) of lesions and adjacent normal bone and the
SD value of subcutaneous fat were measured on 4.2 Therapy Response with DECT
the conventional polychromatic image (140 kVp)
and a set of 11 VMI at different energy levels in 4.2.1 Evaluation of RFA Therapy
the range of 40–140  keV (intervals of 10  keV). with DECT
The contrast-to-noise ratio (CNR) was compared Percutaneous radiofrequency ablation (RFA) is a
between the conventional and all VMI images. minimally invasive therapy for treating advanced
The lowest image noise was in 70 and 140 keV primary or metastatic lung cancers that are not
images, and the highest CNR was noted in suitable for surgical resection treatment and can
70 keV images, suggesting that VMI at 70 keV also be combined with radiotherapy and chemo-
could be the best for diagnosing inconspicuous therapy (Ambrogi et  al. 2006). Liu et  al. evalu-
vertebral metastases (Yue et al. 2017). ated the therapeutic efficacy of lung RFA in 30
Thoracic Oncology 213

patients with contrast-enhanced DECT images. methods described above. Correspondingly,

Conventional CT images were used to obtain before treatment significant differences of AEF
tumor size values, and on water-based and iodine-­ were observed between enlarged (90.4%, 32.3–
based material decomposition images, the densi- 238.5%) and non-enlarged (72.7%, −37.5–
ties of iodine and water in lung tumors were 237.5%) lymph nodes (P = 0.044). A significantly
quantitatively analyzed. In 22 cases, tumor size different change of AEF in responding (decrease
increased after RFA while there was no detect- of 26.3%; P  =  0.022) and non-responding
able change in the remaining eight patients. The (increase of 43.0%; P = 0.031) lymph nodes was
water content in the tumors increased signifi- also demonstrated. Finally, a higher value of AEF
cantly after RFA (P < 0.05), while the iodine con- before treatment was observed in lymph nodes
tent in the tumors was reduced from with a subsequent favorable response (88.6% vs.
2.49 ± 0.74 mg/ml before RFA to 0.45 ± 0.29 mg/ 77.7%; P  =  0.122), but this difference had no
ml in areas of necrosis after RFA (P  =  0.001). reach statistical significance (Baxa et al. 2014).
The authors stated that by comparing the tumor Kim et  al. evaluated tumor response in 10
size, water content, and iodine content before and NSCLC patients treated with anti-angiogenic
after RFA, the metabolic states and therapeutic agents (bevacizumab) by assessing intratumoral
efficiency could be evaluated (Liu et al. 2016). changes with DECT. Tumor responses were eval-
uated and compared with the baseline CT results
4.2.2 Evaluation of Anti-angiogenic using both RECIST (size changes only) and
Therapy with DECT Choi’s criteria (reflecting net tumor enhance-
Baxa et  al. investigate the changes of DECT ment). The weighted κ value for comparing the
parameters and tumor size in advanced non-small RECIST and Choi’s criteria was 0.72, with dis-
cell lung cancer (NSCLC) under anti-EGFR ther- cordant responses found in 5 of 31 lesions.
apy (erlotinib). In 31 patients before and after Therefore, DECT may be a useful tool for
treatment (mean 8 weeks as follow-up) the iodine response evaluation after anti-angiogenic treat-
uptake (IU) was quantified in the arterial and ment in NSCLC patients by providing informa-
venous phases, and the arterial enhancement tion on the net enhancement. In the same study,
fraction (AEF) was calculated. The change of IU iodine-enhanced images also allowed a d­ istinction
in responders and non-responders was compared between tumor enhancement and hemorrhagic
with morphological changes in diameter and vol- response (detected 4 of 29), which is important
ume. A significant decrease of IU was proven in since intratumoral hemorrhage or necrosis during
the venous phase in responders, whereas non-­ tumor treatment can make tumor response evalu-
responders showed variable trends of develop- ation more challenging. The measurement of the
ments and no substantial IU changes. AEF net enhancement allows the discrimination
percentage also showed significant differences between intratumoral hemorrhage, which can be
between both groups (P  <  0.05). Based on the otherwise regarded as a solid enhancing compo-
results, the authors assumed a decrease in vascu- nent, leading to an overestimation of tumor diam-
larization in responding and non-significant vari- eters (Kim et al. 2012).
able development of vascularization in
non-responding tumors. However, the most sig- 4.2.3 Correlation of DECT and PET-CT
nificant change was still observed using the ana- in Therapy Evaluation
tomical parameter diameter and volume (Baxa The information derived from PET-CT imaging
et  al. 2016). The same research group investi- enables treatment evaluation at the molecular/
gated the potential of DECT in staging and ther- metabolic level, far ahead of the macroscopic
apy response monitoring of metastatic lymph visual change in RECIST 1.1. Therefore, the
nodes. They analyzed 110 mediastinal lymph PET-CT response evaluation criterion in solid
nodes in 27 patients with NSCLC with the same tumors (PERCIST) has been suggested as the
214 P. Konietzke

alternative to RECIST 1.1 (Wahl et  al. 2009). by DECT. The sensitivity, which was calculated
Ren et  al. showed that semiautomatic iodine-­ at 100%, was excellent; the NPV was at 100%
related quantitation in DECT correlated well (CI: 91.62, 100) and the specificity was at 85.71%
with metabolism-based measurements in fluo- (CI: 73.33, 92.9). The diagnostic accuracy index
rine-­18 fluorodeoxyglucose (F-FDG) PET/CT, was 86.79% (CI: 75.16, 93.45). Therefore, DECT
suggesting that DECT-based iodine quantitation could be a conceivable alternative for detecting
might be a feasible substitute for assessment of early recurrence after lung RFA (Izaaryene et al.
lung cancer response to chemoradiotherapy/ 2017).
radiotherapy. They acquired a total of 32 pairs of Fehrenbach et al. analyzed the DECT param-
DECT and F-FDG PET/CT imaging from 13 eters in 83 patients with advanced NSCLC treated
patients with primary or metastatic lung cancers by CRT who underwent single-phase, contrast-­
receiving either radiotherapy alone or chemora- enhanced DECT. The evaluation included quanti-
diotherapy. Imaging examinations were per- tative treatment response measurements (RECIST
formed before, immediately, and no later than 1.1), iodine content (IC) measurements, and
6 months after treatment for response evaluation. spectral slope analysis. Secondary outcome
Iodine-related DECT parameters included the parameters were IC and spectral slopes in medi-
total iodine uptake (TIU) and vital volume (VIV), astinal lymph nodes (n = 61). The tumor response
and metabolic metrics of F-FDG-PET/CT were was evaluated by applying RECIST 1.1. 24
the standardized uptake value normalized to lean patients (29%) showed complete remission, 34
body mass (SULpeak), metabolic tumor volume patients (41%) had stable disease (SD) or partial
(MTV), and the total lesion glycolysis (TLG). regression (PR), and 25 (30%) had progressive
Pretreatment imaging data revealed a strong cor- disease (PD). The corresponding hotspot analysis
relation between DECT (RECIST, TIU, and VIV) showed significantly higher iodine values in PD
and F-FDG PET/CT parameters (MTV, TLG) than in SD/PR. Ten patients (12%) with initially
(R2 = 0.86 to 0.90, P < 0.01). After treatment, all SD showed progressive disease during follow-up
DECT and PET/CT parameters significantly for up to 18 months (PDFU). These patients also
decreased, whereas the descending amplitude in had significantly higher hotspot iodine values in
RECIST was substantially smaller than that of the initial scan compared to patients with SD
the other parameters (P < 0.05). During follow- throughout the follow-up period (SDFU) (29%)
­up examinations, all parameters followed a simi- (P < 0.001). Enlarged lymph nodes showed sig-
lar changing pattern, with a strong consistency nificantly lower iodine content and a lower spec-
between RECIST, TIU, VIV and SULpeak, MTV, tral slope pitch than normal-sized nodes
TLG (R2 = 0.78–0.96, P < 0.05) (Ren et al. 2018). (P  =  0.003–0.029). Therefore, the authors con-
cluded that SCT might improve tumor response
4.2.4 Predicting Recurrence evaluation and lymph node assessment in NSCLC
with DECT patients treated with CRT.  Iodine quantification
CT, PET-CT, and clinical patient follow-up are can add information on tumor vascularization
used to detect recurrence. However, no examina- and detection of iodine hotspots indicating resid-
tion is totally reliable, and recurrences might be ual tumor vascularization has the potential to
diagnosed late, when the cancers are locally serve as an imaging marker to predict tumor pro-
extended, or when the patients are metastatic. gression (Fehrenbach et  al. 2019b). Two other
Izaaryene et  al. investigated the utility of studies by Aoki et  al. evaluated the correlation
DECT in order to assess therapeutic responses to between DECT parameters and metabolic uptake
RFA for lung neoplasia. The study included 70 in F-FDG PET-CT and the association with tumor
patients in which the enhancement values of all recurrence, finding strong correlations between
scars were measured without establishing a prior iodine concentration (IC), the maximum stan-
threshold of positivity. At the 1 month follow-up, dardized uptake value (SUVmax), and local
53 nodules were analyzed with DECT and four recurrence in NSCLC treated with stereotactic
nodules had recurred, all of which were detected body radiotherapy (SBRT) (Aoki et al. 2016a,b).
Thoracic Oncology 215

Compliance with Ethical Standards Cormode DP, Skajaa T, Fayad ZA, Mulder WJ (2009)
Nanotechnology in medical imaging: probe design
and applications. Arterioscler Thromb Vasc Biol
Conflict of Interest  The author declares that he has no
29:992–1000
conflicts of interest.
Cormode DP, Si-Mohamed S, Bar-Ness D, Sigovan M,
Naha PC, Balegamire J, Lavenne F, Coulon P, Roessl
E, Bartels M, Rokni M, Blevis I, Boussel L, Douek P
References (2017) Multicolor spectral photon-counting computed
tomography: in vivo dual contrast imaging with a high
count rate scanner. Sci Rep 7:4784
Ambrogi MC, Lucchi M, Dini P, Melfi F, Fontanini G,
den Harder AM, Bangert F, van Hamersvelt RW, Leiner
Faviana P, Fanucchi O, Mussi A (2006) Percutaneous
T, Milles J, Schilham AMR, Willemink MJ, de Jong
radiofrequency ablation of lung tumours: results in the
PA (2017) The effects of iodine attenuation on pul-
mid-term. Eur J Cardiothorac Surg 30:177–183
monary nodule volumetry using novel dual-layer
Aoki M, Akimoto H, Sato M, Hirose K, Kawaguchi
computed tomography reconstructions. Eur Radiol
H, Hatayama Y, Seino H, Kakehata S, Tsushima F,
27:5244–5251
Fujita H, Fujita T, Fujioka I, Tanaka M, Miura H,
Deniffel D, Sauter A, Dangelmaier J, Fingerle A,
Ono S, Takai Y (2016a) Impact of pretreatment
Rummeny EJ, Pfeiffer D (2019) Differentiating intra-
whole-tumor perfusion computed tomography and
pulmonary metastases from different primary tumors
18F-fluorodeoxyglucose positron emission tomog-
via quantitative dual-energy CT based iodine concen-
raphy/computed tomography measurements on
tration and conventional CT attenuation. Eur J Radiol
local control of non-small cell lung cancer treated
111:6–13
with stereotactic body radiotherapy. J Radiat Res
Diederich S, Theegarten D, Stamatis G, Lüthen R (2006)
57:533–540
Solitary pulmonary nodule with growth and contrast
Aoki M, Hirose K, Sato M, Akimoto H, Kawaguchi H,
enhancement at CT: inflammatory pseudotumour as an
Hatayama Y, Fujioka I, Tanaka M, Ono S, Takai Y
unusual benign cause. Br J Radiol 79:76–78
(2016b) Prognostic impact of average iodine density
Fehrenbach U, Kahn J, Böning G, Feldhaus F, Merz K,
assessed by dual-energy spectral imaging for pre-
Frost N, Maurer MH, Renz D, Hamm B, Streitparth
dicting lung tumor recurrence after stereotactic body
F (2019a) Spectral CT and its specific values in the
radiotherapy. J Radiat Res 57:381–386
staging of patients with non-small cell lung cancer:
Barber WC, Wessel JC, Nygard E, Iwanczyk JS (2015)
technical possibilities and clinical impact. Clin Radiol
Energy dispersive CdTe and CdZnTe detectors for
74:456–466
spectral clinical CT and NDT applications. Nucl
Fehrenbach U, Feldhaus F, Kahn J, Böning G, Maurer
Instrum Methods Phys Res A 784:531–537
MH, Renz D, Frost N, Streitparth F (2019b) Tumour
Baxa J, Vondráková A, Matoušková T, Růžičková O,
response in non-small-cell lung cancer patients
Schmidt B, Flohr T, Sedlmair M, Ferda J (2014) Dual-­
treated with chemoradiotherapy  - can spectral CT
phase dual-energy CT in patients with lung cancer:
predict recurrence? J Med Imaging Radiat Oncol
assessment of the additional value of iodine quanti-
63:641–649
fication in lymph node therapy response. Eur Radiol
Fornaro J, Leschka S, Hibbeln D, Butler A, Anderson N, Pache
24:1981–1988
G, Scheffel H, Wildermuth S, Alkadhi H, Stolzmann P
Baxa J, Matouskova T, Krakorova G, Schmidt B, Flohr
(2011) Dual- and multi-energy CT: approach to func-
T, Sedlmair M, Bejcek J, Ferda J (2016) Dual-phase
tional imaging. Insights Imaging 2:149–159
dual-energy CT in patients treated with erlotinib for
Gyobu T, Honda O, Kawata Y, Kikuyama A, Miki H,
advanced non-small cell lung cancer: possible benefits
Yanagawa M, Sumikawa H, Koyama M, Tomiyama
of iodine quantification in response assessment. Eur
N (2013) The effect of the virtual monochromatic
Radiol 26:2828–2836
spectral imaging for the metallic artifact and the pul-
Chae EJ, Song J-W, Seo JB, Krauss B, Jang YM, Song
monary nodule detection. J Comput Assist Tomogr
K-S (2008) Clinical utility of dual-energy CT in the
37:707–711
evaluation of solitary pulmonary nodules: initial expe-
Hou WS, Wu HW, Yin Y, Cheng JJ, Zhang Q, Xu JR
rience. Radiology 249:671–681
(2015) Differentiation of lung cancers from inflam-
Chen X, Xu Y, Duan J, Li C, Sun H, Wang W (2017)
matory masses with dual-energy spectral CT imaging.
Correlation of iodine uptake and perfusion parameters
Acad Radiol 22:337–344
between dual-energy CT imaging and first-pass dual-­
Hou W, Sun X, Yin Y, Cheng J, Zhang Q, Xu J, Li Y, Zhou
input perfusion CT in lung cancer. Medicine 96:e7479
W, Wu H (2016) Improving image quality for lung
Chen ML, Li XT, Wei YY, Qi LP, Sun YS (2019) Can
cancer imaging with optimal monochromatic energy
spectral computed tomography imaging improve the
level in dual energy spectral computed tomography. J
differentiation between malignant and benign pulmo-
Comput Assist Tomogr 40:243–247
nary lesions manifesting as solitary pure ground glass,
Izaaryene J, Vidal V, Bartoli JM, Loundou A, Gaubert
mixed ground glass, and solid nodules? Thorac Canc
JY (2017) Role of dual-energy computed tomogra-
10:234–242
phy in detecting early recurrences of lung tumours
216 P. Konietzke

treated with radiofrequency ablation. Int J Hyperth Lin JZ, Zhang L, Zhang CY, Yang L, Lou HN, Wang ZG
33:653–658 (2016) Application of gemstone spectral computed
Jaffer FA, Weissleder R (2004) Seeing within: molecu- tomography imaging in the characterization of soli-
lar imaging of the cardiovascular system. Circ Res tary pulmonary nodules: preliminary result. J Comput
94:433–445 Assist Tomogr 40:907–911
Jia Y, Xiao X, Sun Q, Jiang H (2018) CT spectral param- Lin LY, Zhang Y, Suo ST, Zhang F, Cheng JJ, Wu HW
eters and serum tumour markers to differentiate (2018) Correlation between dual-energy spectral CT
histological types of cancer histology. Clin Radiol imaging parameters and pathological grades of non-­
73:1033–1040 small cell lung cancer. Clin Radiol 73(412):e412.e1–
Jones GC, Kehrer JD, Kahn J, Koneru BN, Narayan R, e412.e7
Thomas TO, Camphausen K, Mehta MP, Kaushal A Liu L, Zhi X, Liu B, Zhang Y (2016) Utilizing gemstone
(2015) Primary treatment options for high-risk/medi- spectral CT imaging to evaluate the therapeutic effi-
cally inoperable early stage NSCLC patients. Clin cacy of radiofrequency ablation in lung cancer. Radiol
Lung Cancer 16:413–430 Med 121:261–267
Kauczor HU, Bonomo L, Gaga M, Nackaerts K, Peled Liu Z, Zhang Z, Chen C, Hong N (2018a) Value of vir-
N, Prokop M, Remy-Jardin M, von Stackelberg O, tual monochromatic spectral images with metal arti-
Sculier JP (2015) ESR/ERS white paper on lung can- fact reduction algorithm in dual-energy computed
cer screening. Eur Respir J 46:28–39 tomography-­guided microcoil localization of pulmo-
Kaup M, Scholtz JE, Engler A, Albrecht MH, Bauer RW, nary nodules. Medicine 97:e11562
Kerl JM, Beeres M, Lehnert T, Vogl TJ, Wichmann Liu G, Li M, Li G, Li Z, Liu A, Pu R, Cao H, Liu Y
JL (2016) Dual-energy computed tomography virtual (2018b) Assessing the blood supply status of the focal
monoenergetic imaging of lung cancer: assessment ground-glass opacity in lungs using spectral computed
of optimal energy levels. J Comput Assist Tomogr tomography. Korean J Radiol 19:130–138
40:80–85 MacMahon H, Austin JH, Gamsu G, Herold CJ, Jett JR,
Kawai T, Shibamoto Y, Hara M, Arakawa T, Nagai K, Naidich DP, Patz EF Jr, Swensen SJ (2005) Guidelines
Ohashi K (2011) Can dual-energy ct evaluate contrast for management of small pulmonary nodules detected
enhancement of ground-glass attenuation?: phan- on CT scans: a statement from the Fleischner Society.
tom and preliminary clinical studies. Acad Radiol Radiology 237:395–400
18:682–689 Maldonado F, Daniels CE, Hoffman EA, Yi ES, Ryu JH
Kim YN, Lee HY, Lee KS, Seo JB, Chung MJ, Ahn MJ, (2007) Focal organizing pneumonia on surgical lung
Park K, Kim TS, Yi CA (2012) Dual-energy CT in biopsy: causes, clinicoradiologic features, and out-
patients treated with anti-angiogenic agents for non-­ comes. Chest 132:1579–1583
small cell lung cancer: new method of monitoring Milne EN (1967) Circulation of primary and metastatic
tumor response? Korean J Radiol 13:702–710 pulmonary neoplasms. A postmortem microarterio-
Kim C, Kim D, Lee KY, Kim H, Cha J, Choo JY, Cho PK graphic study. Am J Roentgenol Radium Therapy,
(2018) The optimal energy level of virtual monochro- Nucl Med 100:603–619
matic images from spectral CT for reducing beam-­ Ohno Y, Nishio M, Koyama H, Miura S, Yoshikawa
hardening artifacts due to contrast media in the thorax. T, Matsumoto S, Sugimura K (2014) Dynamic
AJR Am J Roentgenol 211:557–563 ­contrast-­enhanced CT and MRI for pulmonary nodule
Kovacs DG, Rechner LA, Appelt AL, Berthelsen AK, assessment. AJR Am J Roentgenol 202:515–529
Costa JC, Friborg J, Persson GF, Bangsgaard JP, Ramnath N, Dilling TJ, Harris LJ, Kim AW, Michaud
Specht L, Aznar MC (2018) Metal artefact reduc- GC, Balekian AA, Diekemper R, Detterbeck FC,
tion for accurate tumour delineation in radiotherapy. Arenberg DA (2013) Treatment of stage III non-small
Radiother Oncol 126:479–486 cell lung cancer: diagnosis and management of lung
Leng S, Yu Z, Halaweish A, Kappler S, Hahn K, Henning cancer, 3rd ed: American College of Chest Physicians
A, Li Z, Lane J, Levin DL, Jorgensen S, Ritman E, evidence-based clinical practice guidelines. Chest
McCollough C (2016) Dose-efficient ultrahigh-­ 143:e314S–e340S
resolution scan mode using a photon counting detec- Ren Y, Jiao Y, Ge W, Zhang L, Hua Y, Li C, Zhai W, Tang
tor computed tomography system. J Med Imaging X, He W, Fang M, Zheng X (2018) Dual-energy
(Bellingham) 3:043504 computed tomography-based iodine quantitation for
Li GJ, Gao J, Wang GL, Zhang CQ, Shi H, Deng K response evaluation of lung cancers to chemoradio-
(2016) Correlation between vascular endothelial therapy/radiotherapy: a comparison with fluorine-18
growth factor and quantitative dual-energy spec- fluorodeoxyglucose positron emission tomography/
tral CT in non-­small-­cell lung cancer. Clin Radiol computed tomography-based positron emission
71:363–368 tomography/computed tomography response evalua-
Li M, Zhang L, Tang W, Jin Y-J, Qi L-L, Wu N (2019) tion criterion in solid tumors. J Comput Assist Tomogr
Identification of epidermal growth factor receptor 42:614–622
mutations in pulmonary adenocarcinoma using dual-­ Roessl E, Proksa R (2007) K-edge imaging in x-ray com-
energy spectral computed tomography. Eur Radiol puted tomography using multi-bin photon counting
29:2989–2997 detectors. Phys Med Biol 52:4679–4696
Thoracic Oncology 217

Rubin GD (2015) Lung nodule and cancer detection in Taguchi K, Iwanczyk JS (2013) Vision 20/20: single pho-
computed tomography screening. J Thorac Imaging ton counting x-ray detectors in medical imaging. Med
30:130–138 Phys 40:100901
Ruoslahti E (2002) Specialization of tumour vasculature. Travis WD, Brambilla E, Nicholson AG, Yatabe Y, Austin
Nat Rev Cancer 2:83–90 JHM, Beasley MB, Chirieac LR, Dacic S, Duhig E,
Salem A, Asselin MC, Reymen B, Jackson A, Lambin Flieder DB, Geisinger K, Hirsch FR, Ishikawa Y, Kerr
P, West CML, O’Connor JPB, Faivre-Finn C (2018) KM, Noguchi M, Pelosi G, Powell CA, Tsao MS,
Targeting hypoxia to improve non-small cell lung Wistuba I (2015) The 2015 world health organization
cancer outcome. J Natl Cancer Inst 110. https://doi. classification of lung tumors: impact of genetic, clini-
org/10.1093/jnci/djx160 cal and radiologic advances since the 2004 classifica-
Sekiguchi T, Ozawa Y, Hara M, Nakagawa M, Goto tion. J Thorac Oncol 10:1243–1260
T, Shibamoto Y (2019) Visibility of the hilar lymph Travis WD, Brambilla E, Geisinger KR (2016)
nodes using advanced virtual monoenergetic low-keV Histological grading in lung cancer: one system for
images for preoperative evaluation of lung cancer. Br J all or separate systems for each histological type? Eur
Radiol 92:20180734 Respir J 47:720
Shikhaliev PM (2005) Beam hardening artefacts in com- Vogelstein B, Papadopoulos N, Velculescu VE, Zhou S,
puted tomography with photon counting, charge inte- Diaz LA, Kinzler KW (2013) Cancer genome land-
grating and energy weighting detectors: a simulation scapes. Science 339:1546
study. Phys Med Biol 50:5813–5827 Wahl RL, Jacene H, Kasamon Y, Lodge MA (2009) From
Silvestri GA, Gonzalez AV, Jantz MA, Margolis ML, RECIST to PERCIST: evolving considerations for
Gould MK, Tanoue LT, Harris LJ, Detterbeck FC PET response criteria in solid tumors. J Nucl Med
(2013) Methods for staging non-small cell lung 50:122S
cancer: diagnosis and management of lung can- Wang G, Zhang C, Li M, Deng K, Li W (2014) Preliminary
cer, 3rd ed: American College of Chest Physicians application of high-definition computed tomographic
evidence-based clinical practice guidelines. Chest Gemstone Spectral Imaging in lung cancer. J Comput
143:e211S–e250S Assist Tomogr 38:77–81
Si-Mohamed S, Cormode DP, Bar-Ness D, Sigovan M, Wu L, Cao G, Zhao L, Tang K, Lin J, Miao S, Lin T, Sun J,
Naha PC, Langlois JB, Chalabreysse L, Coulon P, Zheng X (2018a) Spectral CT analysis of solitary pul-
Blevis I, Roessl E, Erhard K, Boussel L, Douek P monary nodules for differentiating malignancy from
(2017) Evaluation of spectral photon counting com- benignancy: the value of iodine concentration spatial
puted tomography K-edge imaging for determination distribution difference. Biomed Res Int 2018:4830659
of gold nanoparticle biodistribution in vivo. Nanoscale Wu F, Zhou H, Li F, Wang JT, Ai T (2018b) Spectral CT
9:18246–18257 imaging of lung cancer: quantitative analysis of spec-
Sun YS, Zhang XY, Cui Y, Tang L, Li XT, Chen Y, Zhang tral parameters and their correlation with tumor char-
XP (2013) Spectral CT imaging as a new quantitative acteristics. Acad Radiol 25:1398–1404
tool? Assessment of perfusion defects of pulmonary Xiao H, Liu Y, Tan H, Liang P, Wang B, Su L, Wang S, Gao
parenchyma in patients with lung cancer. Chin J Canc J (2015) A pilot study using low-dose Spectral CT and
Res 25:722–728 ASIR (Adaptive Statistical Iterative Reconstruction)
Sung H, Ferlay J, Siegel RL, Laversanne M, algorithm to diagnose solitary pulmonary nodules.
Soerjomataram I, Jemal A, Bray F (2021) Global can- BMC Med Imaging 15:54
cer statistics 2020: GLOBOCAN estimates of inci- Yang F, Dong J, Wang X, Fu X, Zhang T (2017) Non-­
dence and mortality worldwide for 36 cancers in 185 small cell lung cancer: spectral computed tomography
countries. CA Cancer J Clin 71(3):209–249 quantitative parameters for preoperative diagnosis of
Symons R, Cork TE, Sahbaee P, Fuld MK, Kappler S, metastatic lymph nodes. Eur J Radiol 89:129–135
Folio LR, Bluemke DA, Pourmorteza A (2017a) Yi CA, Lee KS, Kim EA, Han J, Kim H, Kwon OJ,
Low-­ dose lung cancer screening with photon- Jeong YJ, Kim S (2004) Solitary pulmonary nodules:
counting CT: a feasibility study. Phys Med Biol dynamic enhanced multi-detector row CT study and
62:202–213 comparison with vascular endothelial growth factor
Symons R, Krauss B, Sahbaee P, Cork TE, Lakshmanan and microvessel density. Radiology 233:191–199
MN, Bluemke DA, Pourmorteza A (2017b) Photon-­ Yoo SY, Kim Y, Cho HH, Choi MJ, Shim SS, Lee JK,
counting CT for simultaneous imaging of multiple Baek SY (2013) Dual-energy CT in the assessment of
contrast agents in the abdomen: an in vivo study. Med mediastinal lymph nodes: comparative study of virtual
Phys 44:5120–5127 non-contrast and true non-contrast images. Korean J
Symons R, Cork TE, Lakshmanan MN, Evers R, Davies-­ Radiol 14:532–539
Venn C, Rice KA, Thomas ML, Liu CY, Kappler S, Yu L, Leng S, McCollough CH (2012) Dual-energy
Ulzheimer S, Sandfort V, Bluemke DA, Pourmorteza CT-based monochromatic imaging. AJR Am J
A (2017c) Dual-contrast agent photon-counting com- Roentgenol 199:S9–s15
puted tomography of the heart: initial experience. Int J Yu Y, Wang X, Shi C, Hu S, Zhu H, Hu C (2019) Spectral
Cardiovasc Imaging 33:1253–1261 computed tomography imaging in the differential
218 P. Konietzke

diagnosis of lung cancer and inflammatory myofibro- organizing pneumonia: an analysis of consecutive
blastic tumor. J Comput Assist Tomogr 43:338–344 histopathologically confirmed 45 cases. Eur J Radiol
Yue D, Ru Xin W, Jing C, Fan Rong C, Li Fei S, Ai Lian 83:73–78
L, Hong LY (2017) Virtual monochromatic spectral Zhou W, Montoya J, Gutjahr R, Ferrero A, Halaweish A,
imaging for the evaluation of vertebral inconspicu- Kappler S, McCollough C, Leng S (2017) Lung nod-
ous osteoblastic metastases from lung. Acta Radiol ule volume quantification and shape differentiation
58:1485–1492 with an ultra-high resolution technique on a photon-­
Zhao F, Yan SX, Wang GF, Wang J, Lu PX, Chen B, Yuan counting detector computed tomography system. J
J, Zhang SZ, Wang YX (2014) CT features of focal Med Imaging (Bellingham) 4:043502
 Gastrointestinal Imaging:
Oncology (Liver, Pancreas, Bowel
Cancer, and Treatment Response)

Simon Lennartz and Nils Große Hokamp

Contents
1 Introduction  220
2  linical Applications of Dual-Energy CT for Oncologic Imaging
C
of the Liver  220
2.1  Imaging Protocols  220
2.2  Challenges in Imaging the Liver  221
2.3  Liver Lesion Detection and Delineation  221
2.4  Liver Lesion Characterization  224
2.5  Response Assessment of Hepatocellular Carcinoma  224
3  ual-Energy Imaging Applications for Pancreatic Imaging
D  224
3.1  Imaging Protocols  224
3.2  Imaging of Pancreatic Lesions  225
3.3  Imaging of Pancreatic Cancer  226
4  ncologic Imaging of the Gastrointestinal System
O
with Dual-Energy CT  226
4.1  Imaging Protocols  226
4.2  Upper GI Tract  227
4.3  Lower Intestinal Tract  227
5  ncologic Applications of DECT in the Abdomen outside
O
of Parenchymal Organs  229
6  he Role of Dual-Energy CT for Assessing Oncologic Treatment
T
Response  230
6.1  Assessment of Treatment Response After Locoregional Therapy  230
6.2  Assessment of Treatment Response After Antiangiogenic Treatment  231
References  232

Abstract

S. Lennartz · N. Große Hokamp (*) Oncologic diseases of the gastrointestinal

Institute for Diagnostic and Interventional Radiology,
University Hospital Cologne, Cologne, Germany
tract encompass a wide variety of tumors and
e-mail: [email protected]; nils.grosse-­ pathological conditions. Dual-Energy CT can
[email protected] be tremendously helpful in this instance, par-

© The Author(s), under exclusive license to Springer Nature Switzerland AG 2022 219
H. Alkadhi et al. (eds.), Spectral Imaging, Medical Radiology Diagnostic Imaging,
https://doi.org/10.1007/978-3-030-96285-2_14
220 S. Lennartz and N. Große Hokamp

ticularly with regard to visualization of subtle when there is no unenhanced acquisition avail-
tumor masses in parenchymal organs such as able. Moreover, iodine maps have been explored
the liver or the pancreas. In this context, low- as an alternate method for characterization of and
keV virtual monoenergetic images and iodine assessment of response to anticancer treatment.
maps play an important role. Other applica- This chapter will focus on the most important
tion fields of dual-energy CT for diagnosing DECT applications for imaging of the gastrointes-
oncologic diseases of the gastrointestinal tract tinal and hepatobiliary system as well as the pan-
include CT colonography, assessment of peri- creas with a focus on oncologic imaging. Further,
toneal lesions and evaluation of treatment it will include a review of current data on the
response. This chapter will provide a conclu- potential of quantitative characteristics of DECT
sive overview on the most important clinical for characterization and treatment response assess-
use cases as well as scientific concepts that are ment in oncologic diseases of the abdomen.
still pending clinical implementation.

2  linical Applications of Dual-

C
1 Introduction Energy CT for Oncologic
Imaging of the Liver
Dual-energy CT (DECT) is a clinically useful
technique for a wide range of oncologic imaging 2.1 Imaging Protocols
applications of the hepatobiliary system, the pan-
creas, and the gastrointestinal tract. Regarding In imaging of the liver, administration of i.v.-contrast
oncologic imaging of the abdomen, both the is usually required, while administration of oral
qualitative improvement of iodine and soft tissue contrast (positive or negative) depends on the clini-
contrast and the option for material quantification cal question and field of view. Pertaining to imaging
provided by DECT have been extensively investi- phases, the portal venous phase is the “standard”
gated and can be of clinical use. While the first phase at most institutions. Depending on the ques-
mentioned provides a more accurate lesion delin- tion in focus, early/late arterial (e.g. unknown
eation, lesion detection, and assessment of tumor hepatic malignancy, hepatocellular carcinoma,
infiltration, the second may be useful for quanti- cholangiocellular carcinoma, known cancer with
tative lesion characterization. possibly hypervascular metastasis, prior surgery) or
Despite substantial technical developments of late phases (e.g. unknown hepatic malignancy,
CT in the past decades including multidetector hepatocellular carcinoma, cholangiocellular carci-
CT and 3D volumetric rendering, lesion detec- noma) might be acquired in addition. Unenhanced
tion and delineation in parenchymal organs such examinations can provide incremental information
as the liver and pancreas in CT is still challenging in select cases (e.g. pre-/post local ablative therapy),
due to its inherently limited soft tissue contrast. however, are not considered standard in imaging of
This particularly accounts for small and/or inter- the liver (Prokop and Galanski 2002). The need for
mediate lesions with similar CT characteristics as a true-­unenhanced phase furthermore can be ques-
the surrounding organ parenchyma. Therefore, tioned if VNC from any DECT acquisition can be
the capability of DECT to increase the lesion-to-­ reconstructed. In a consensus paper published by
parenchyma contrast by means of low energy vir- Patel et al. in 2016, DECT acquisition is suggested
tual monoenergetic images as well as for the arterial phase for DSCT and kVp-switching
material-specific images such as iodine maps DECT, while this selection naturally is not required
adds clinical value in many diagnostic scenarios. for dual-layer detector DECT and photon counting
Moreover, the option of DECT-derived material detector CT.
quantification allows among others reconstruct- Specific use cases of VNC, low keV virtual
ing iodine maps as well as virtual unenhanced monoenergetic images, and material-specific
images. They can help characterizing inciden- maps for imaging of the liver are elucidated here-
tally detected ambiguous or intermediate lesions after and summarized in Table 1.
Gastrointestinal Imaging: Oncology (Liver, Pancreas, Bowel Cancer, and Treatment Response) 221

Table 1 Overview on key applications of different improvements in hard- and software as well as
DECT-derived reconstructions in liver imaging
the introduction of novel technical approaches to
Reconstruction Application DECT resulted in significant noise reductions in
Low keV virtual Detection and delineation of low keV virtual monoenergetic images which is
monoenergetic hypoattenuating and
images hyperattenuating liver lesions why these reconstructions can be used as clinical
Evaluation of vessels, e.g. prior “screening” tools, allowing for an optimal lesion-­
to surgery to-­parenchyma contrast without noticeable
Virtual Possible replacement for increase in image noise (Grant et al. 2014; Große
noncontrast true-unenhanced images for
Hokamp et al. 2019).
images assessment of lesion
enhancement in dedicated Hypoattenuating hepatic lesions are frequently
multiphasic protocols encountered incidentally in routine, portal venous
Iodine maps Liver lesion characterization phase CT examinations of the abdomen. While
many of these lesions fall in the category “too
small to characterize” and eventually turn out
2.2 Challenges in Imaging benign, the probability of underlying malignancy
the Liver or a small liver metastasis can be higher depend-
ing on the patient’s history (Fig.  1). Detecting
There are three key diagnostic challenges at CT hypoattenuating lesions in clinical routine is the
imaging of the liver: first, the accurate detection first step for further lesion characterization and
and delineation of lesions at oncologic staging or clinical decision making on the necessity of addi-
follow-up; second, the differential diagnosis (i.e. tional imaging (e.g. dedicated liver MRI or con-
accurate characterization) of these lesions; and trast-enhanced ultrasound) or follow-up
third, staging and response assessment of hepatic examinations. Many studies found that the
tumors, the last of which will be addressed in the improved contrast of hypoattenuating liver lesions
dedicated subsection below about DECT-enabled in low keV virtual monoenergetic images facili-
therapy response assessment. tates improvements in their detection and delinea-
tion (Husarik et al. 2015; Altenbernd et al. 2016a;
Hanson et al. 2018; Große Hokamp et al. 2018a).
2.3  iver Lesion Detection
L Conversely, it was described that in patients with
and Delineation liver steatosis, the advantage of low energy virtual
monoenergetic images for imaging hypoattenuat-
Many studies have elucidated the capability of ing lesions might be lower or even nullified
DECT to improve the contrast of hyper- and (Nattenmüller et  al. 2015). Opposed to this,
hypoattenuating liver lesions and to thereby Grosse Hokamp et al. demonstrated in an ex-vivo
allow for an improved lesion detection and delin- set-up that benefits from low keV imaging may be
eation (Robinson et al. 2010). While the underly- leveraged even in poorly attenuating livers and
ing principle of highlighting the iodine contrast mildly hypodense lesions; however, systematic
in DECT is the same for improving the assess- validation of these findings in vivo is necessary to
ment of both hypo- and hyperattenuating lesions, clarify this case (Große Hokamp et al. 2018b).
the contrast improvement of hyperattenuating Hyperattenuating lesions of the liver comprise
lesions is based on the increased attenuation of a wide spectrum of differential diagnoses ranging
the lesion itself, while for hypoattenuating from benign lesions with often characteristic
lesions, the higher contrast of the circumjacent appearance (e.g. focal nodular hyperplasia, hem-
organ parenchyma and the relatively lower iodine angioma) to liver metastasis or primary tumors
contrast enhancement of the lesion are exploited. such as hepatocellular carcinoma. Whereas
Whereas earlier generation DECT systems detection of hypervascular lesions may often be
showed a higher image noise in virtual monoen- effortless in case of a distinctive enhancement,
ergetic images at low energy levels, recent lesions with a more subtle enhancement or a
222 S. Lennartz and N. Große Hokamp

a b

c d

Fig. 1  Patient with new appearance of hypodense liver lent: (a) the contrast between the lesions and liver paren-
lesions suspicious for metastatic disease. While these chyma is gradually increased with decreasing keV level
lesions are very subtle in the conventional image equiva- from 70 keV (b) down to 40 keV (e)
Gastrointestinal Imaging: Oncology (Liver, Pancreas, Bowel Cancer, and Treatment Response) 223

a c

b d

Fig. 2 Patient with incidentally detected hyperdense dual-energy CT derived virtual monoenergetic images
lesion in the left liver lobe. The lesion is only subtle with 40 keV (b, c) allow for improved delineation by increas-
low contrast to circumjacent parenchyma (a, b), whereas ing the lesion-to-parenchyma contrast

smaller size might still be missed, hampering more accurate distinction between characteristic
accurate diagnosis. DECT can highlight the enhancement patterns.
iodine hyperenhancement of lesions in compari- To facilitate and optimize diagnostic assess-
son to the surrounding liver parenchyma in low ment of any focal lesions, dedicated window set-
keV virtual monoenergetic images or low kVp tings have been proposed for a long time.
acquisitions, respectively (Altenbernd et  al. Historically, this arose from the lacking possibil-
2011; Shuman et al. 2014; Große Hokamp et al. ity to adjust window settings on a printed film;
2018b). This effect facilitates an improved delin- however, quickly became implemented in the
eation and conspicuity of hypervascular lesions DICOM standard. For conventional images, a
of the liver (Fig. 2), more accurate measurements window center of 100 and a width of 150 have
of hyperattenuating lesion components, and a been proposed and commonly accepted (Mayo-­
224 S. Lennartz and N. Große Hokamp

Smith et  al. 1999); however, considering the could differentiate HCC from focal nodular
altered contrast characteristics in low keV virtual hyperplasia (Yu et al. 2013). The latter showed a
monoenergetic images it is obvious that adjust- significantly higher normalized iodine concentra-
ments become necessary. They further report that tion both when the lesions were normalized to
appropriate window settings can be semi-­ physiologically appearing liver parenchyma and
automatically obtained from simple attenuation to the abdominal aorta. In another study, it was
measurements using linear models (Große reported that iodine maps were more accurate in
Hokamp et al. 2018c), e.g. a center of 200 and a differentiating small hepatic hemangioma from
width of 405 are suggested when interpreting HCC as compared to qualitative differentiation
40 keV virtual monoenergetic images for hepatic using conventional CT (Lv et  al. 2011). Patel
lesions in portal venous phase. Naturally, such et al. found that malignant lesions were detected
settings have to be adjusted to the individual more accurately among incidentally found
examination protocol; however, this can easily be hypoattenuating liver lesions when using an
done using suggested conversion formulas (e.g.: iodine-based threshold as compared to a
center  =  HU_liver  +  15; width  =  6  ×  HU_ HU-based threshold. They reported a signifi-
liver + 12; with HU_liver being measured in the cantly higher iodine concentration of 1.7 mg/ml
respective reconstruction). Similar strategies have in hypoattenuating lesions with underlying
been suggested for imaging of other organ regions malignancy as opposed to a mean iodine concen-
(Hickethier et al. 2018). tration of 0.6  mg/ml in hypoattenuating lesions
that were benign (Patel et al. 2018). Last, Reimer
et al. recently reported that low keV imaging is
2.4 Liver Lesion Characterization superior in detecting washout phenomena in an
HCC population (Reimer et al. 2020).
Apart from the qualitative improvements in the
assessment of liver lesions that can be attained by
means of low keV virtual monoenergetic images, 2.5 Response Assessment
the material decomposition and quantification of Hepatocellular Carcinoma
capabilities of DECT further allow for a more
accurate characterization of hepatic lesions as As a tumor for which its enhancement patterns
compared to conventional, single-energy are among the most important diagnostic criteria,
CT.  However, it should be acknowledged that hepatocellular carcinoma has been subject to sev-
these approaches employing quantitative thresh- eral studies investigating the potential benefit of
olds mostly have not been validated across avail- using low keV virtual monoenergetic images and
able DECT scanner types and should therefore be iodine maps for response assessment particularly
constrained in regard to their clinical application after locoregional therapy. This topic will be
to the original scanner type on which the corre- addressed in Sect. 6.
sponding study was performed. In one study,
Kaltenbach et al. suggested that metastases from
hepatic neuroendocrine tumors may be distin- 3 Dual-Energy Imaging
guished from hepatocellular carcinoma by means Applications for Pancreatic
of (normalized) iodine uptake and attenuation Imaging
measurements in early arterial phase, with iodine-­
based measurements being superior (Kaltenbach 3.1 Imaging Protocols
et al. 2018); the metastases from neuroendocrine
tumors showed a significantly lower normalized Protocols for pancreatic imaging are subject of
iodine concentration. Yu et  al., in another study an ongoing discussion and most recommenda-
focused on discerning HCC from frequent differ- tions directly transfer to DECT imaging in sus-
ential diagnoses, found that iodine quantification pected pancreatic disease (Table 2). Fasting prior
Gastrointestinal Imaging: Oncology (Liver, Pancreas, Bowel Cancer, and Treatment Response) 225

Table 2  Overview on applications of different DECT-­ structions) in addition to blended or true conven-
derived reconstructions for imaging of the pancreas
tional images.
Reconstruction Application
Low keV virtual Detection and delineation of
monoenergetic incidental lesions
images Assessment of neuroendocrine 3.2 Imaging of Pancreatic Lesions
and other hypervascular tumors
of the pancreas In CT imaging of the pancreas, the accurate
Delineation of pancreatic detection of incidental lesions is important as
cancer
Evaluation of tumor extension some of them may require additional imaging,
and (vascular) infiltration in e.g. with MRI or follow-up examinations to
pancreatic cancer rule out malignancy. While lesions with a high
Iodine maps Differentiation of pancreatic contrast to the surrounding parenchyma such as
cancer and mass-forming
typical hypervascular neuroendocrine tumors
pancreatitis (explorative)
Virtual noncontrast Detection of calcifications in and poorly vascularized, clearly hypodense
chronic pancreatitis pancreatic adenomas may be easy to depict
even in single-­energy, contrast-enhanced CT,
many pancreatic lesions such as isoattenuating
to examination is commonly conducted. Most or mildly hypoattenuating pancreatic adenocar-
sites suggest negative oral contrast (1000– cinomas as well as hypovascular neuroendo-
1500 ml) including 250 ml immediately prior to crine tumors are more subtle and can easily be
image acquisition. Furthermore, administration missed, particularly when they are of smaller
of buscopan is often advocated for. Dependent on size. This accounts even more for incidental
the clinically suspected condition suggested con- detection in portal venous phase examinations
trast phases include unenhanced (pancreatitis), without a dedicated pancreatic phase in which
early and late arterial (suspected carcinoma, neu- the contrast even of typical lesions is often sub-
roendocrine tumor), and parenchymatous or por- optimal. DECT can enable improved depiction
tal venous phases (aforementioned and trauma); of such lesions following the same principle
however, the specific protocols settings (includ- that is exploited for liver imaging: increasing
ing timing) as well as contrast media application soft tissue and iodine enhancement to highlight
(including rate, volume, concentration) vary differences herein between the lesion itself and
largely between institutions (Prokop and Galanski the surrounding parenchyma (Table  2).
2002). Consequently, multiple studies using many dif-
Omitting an unenhanced phase can be consid- ferent DECT systems have concordantly shown
ered when another series is acquired in DECT that low keV virtual monoenergetic images
mode and therefore virtual noncontrast image improve conspicuity and diagnostic assessment
reconstructions are available. When examining a of focal pancreatic lesions (McNamara et  al.
patient on an emission-based DECT system, the 2015; Quiney et  al. 2015; El Kayal et  al.
phase in which DECT data is acquired needs to 2019)—this principle accounts for iso- to
be chosen. In this context, a balance between hypoattenuating lesions such as pancreatic ade-
radiation-dose, biological and technological pre- nocarcinoma (Bhosale et al. 2015; Quiney et al.
sets has to be found. It appears that most institu- 2015), cystic pancreatic lesions (Laukamp et al.
tions chose the arterial phase to serve as DECT 2021) as well as hypervascularized lesions such
acquisition (Mastrodicasa et al. 2019). as neuroendocrine tumors (Lin et al. 2012). Of
Image reconstructions used in assessment of note, it has been shown that low energy virtual
pancreatic disease include virtual noncontrast, monoenergetic images improve detection of
low energy virtual monoenergetic (e.g. 50 keV), small and isoattenuating pancreatic ductal ade-
and iodine density maps (possibly as fused recon- nocarcinomas and hypovascular/isoattenuating
226 S. Lennartz and N. Große Hokamp

neuroendocrine tumors as well (Lin et al. 2012; tral slopes may distinguish between these two
Patel et al. 2017). entities (Yin et al. 2015).
Compared to the unequivocal evidence on
improved qualitative lesion assessment which
has been reported for most available dual-energy 4 Oncologic Imaging
CT scanner types, data on quantitative differen- of the Gastrointestinal
tiation of pancreatic lesions is sparser. For exam- System with Dual-Energy CT
ple, Chu et al. suggested that iodine maps may be
helpful for distinguishing between solid and cys- Staging of malignancies of the gastrointestinal
tic pancreatic lesions. Moreover, virtual unen- tract often requires the use of different modalities
hanced images may be used to differentiate including CT or PET/CT for assessment of dis-
calcifications from linear iodine enhancement in tant metastasis and presurgical planning as well
cystic pancreatic lesions (Chu et al. 2012). as endoscopy for assessment of tumor infiltration
or metastatic spread to locoregional lymph nodes.
However, there are various potential applications
3.3 Imaging of Pancreatic Cancer of DECT that may benefit oncologic imaging of
the gastrointestinal system including improved
Pancreatic cancer greatly contributes to cancer tumor detection, e.g. of esophageal cancer or
mortality. The majority of pancreatic cancers neuroendocrine tumors, as well as DECT colo-
show metastatic spread at the time of diagnosis. nography (Table 3).
However, surgical resection of the tumor often
still yields the best chance of attaining long-
term survival. In assessing pancreatic cancer, 4.1 Imaging Protocols
DECT can help tackling three important chal-
lenges: detecting subtle primary tumors, delin- Various institutional protocols are available for
eating tumor margins, and assessing possible GI imaging. Little alterations from standard CT
infiltration of other anatomical structures (Gupta protocols are required with regard to luminal
et al. 2016). contrast. In this context it needs to be acknowl-
CT plays an important role in the presurgical edged that the bowel wall assessment is ham-
staging of pancreatic cancer, particularly with pered by administration of positive oral contrast.
regard to the accurate assessment of potential Besides, this being an issue well known from
infiltration of extrapancreatic structures such as conventional CT, the impairment naturally is
circumjacent vessels. Allowing abdominal sur- aggravated in low keV imaging. In the authors’
geons to evaluate resectability requires an accu- institution and supported by increasing evi-
rate assessment of it and to which degree vessels dence, positive oral contrast is therefore avoided
are affected by the tumor. Due to the highlighting whenever possible (Kammerer et  al. 2015).
of intravascular contrast, low energy virtual
monoenergetic images allow for an improved Table 3  Overview on applications of different DECT-­
assessment of vessel infiltration in pancreatic derived reconstructions for imaging of the GI system
cancer (Bellini et  al. 2017; Nagayama et  al. Reconstruction Application
2020). Low keV virtual Improved detection and
Another possible benefit of DECT for pancre- monoenergetic images delineation of focal masses
atic cancer imaging might be the capability to Iodine maps Improved detection,
particularly of small lesions
differentiate pancreatic adenocarcinoma from
Differentiation between
chronic mass-forming pancreatitis, an important different entities
imaging differential diagnosis. In this regard, a Virtual noncontrast Lesion characterization
pilot study by Yin et al. showed that normalized images Fecal tagging, particulalry
iodine concentration in combination with spec- in DECT colonography
Gastrointestinal Imaging: Oncology (Liver, Pancreas, Bowel Cancer, and Treatment Response) 227

Pertaining to the type of positive contrast located in the jejunum or duodenum, while ileal
medium, Gabbai et al. reported that differentiat- manifestations are rare. Despite missing litera-
ing iodine and barium is possible to some extent ture evidence in this context, the theoretical ben-
in an ex  vivo s­etting; however, evidence with efit from low keV virtual monoenergetic images
regard to in vivo applicability is missing (Gabbai and iodine map (overlays) translates to this
et al. 2015). application.
For evaluation of the GI tract, usually a single
contrast phase is sufficient; therefore, no decision
on DECT acquisition phase must be made. If, 4.3 Lower Intestinal Tract
however, an early arterial or angiographic phase
is required (e.g. prior to surgery), as elucidated Note, this section deals with DECT applications
earlier, most institutions perform the earlier for lower GI tumors only, while other diseases of
acquisition in dual-energy mode (if using an the lower GI are discussed in the chapter “Bowel
emission-based DECT system). Imaging.”
Again, DECT imaging, particularly low keV
virtual monoenergetic images and iodine maps
4.2 Upper GI Tract (overlays), is promising means to improve detec-
tion and delineation of tumors (Fulwadhva et al.
Assessing the upper GI tract naturally is not the 2016). In this respect, it was reported that iodine
primary domain of CT; however, particularly in maps and weighted average images derived from
oncologic disease, CT remains standard of care DECT facilitated detection of colorectal cancers
to assess tumor burden. Zopfs et  al. recently with a sensitivity of 96.7% even without bowel
reported that 40  keV virtual monoenergetic preparation or insufflation (Boellaard et al. 2013).
images hold value in assessing esophageal cancer However, these findings are still subject to larger-­
locoregionally. They demonstrated that these scale validation.
reconstructions can be helpful in qualitative In regard to quantitative use of iodine maps
assessment of the primary tumor and furthermore for lower intestinal tract imaging, researchers
in depiction of locoregional lymph nodes. Yet, from China recently reported that arterial phase
they also found that the inherited limitation of CT information in terms of iodine concentration
in detecting tumor infiltration is not overcome were found helpful in image-based differentia-
(Zopfs et al. 2021). Figure 3 neatly demonstrate tion between well- and poorly differentiated
this benefit in terms of qualitative assessment. colon cancers in 47 patients (Chuang-Bo et  al.
Figure 3c furthermore illustrates a possible ben- 2017). Al-Najami et  al. performed a proof-of-­
efit from iodine maps in this context; however, concept investigation on lymph node character-
their clinical impact in assessing esophageal can- ization in rectal cancer using small samples of
cer is subject of ongoing investigations. patients undergoing surgery. Patients were exam-
Similar findings account for gastric malig- ined with iodinated contrast media intraopera-
nancy. Here, however, imaging of gastrointestinal tively immediately prior to resection. The
stromal tumors (GIST) illustrates the potential specimen then underwent re-examination using a
iodine maps hold for assessing treatment response kVp-switching dual-energy CT. The authors sug-
(see Sect. 6). gest a cut-off of 2.58 mg/cc for their specific set-
As elucidated above, low keV imaging partic- ting and small cohort, resulting in a sensitivity
ularly is useful in hypervascular lesions and and specificity of 86% and 92%, respectively
therefore considered to be of particular benefit in (Al-Najami et al. 2016). Yet, such cut-offs have
assessing neuroendocrine tumors. Benefits of not been validated on a larger scale and therefore
these images in assessing pancreatic NET are should not be used for clinical decision making.
described in the corresponding section. Intestinal In this context, the authors would like to draw
manifestations of NET are most commonly attention to Sect. 6.
228 S. Lennartz and N. Große Hokamp

a b

Fig. 3  Patient with path-proven esophageal cancer. At as the iodine overlay image (c) whereas barely being per-
initial DECT staging, the contrast-enhancing semicircular ceivable in the conventional image (a). The localization
thickening of the distal esophagus can be well appreciated correlated well with primary tumor localization deter-
in the virtual monoenergetic images at 40 keV (b) as well mined with endoscopic ultrasound

4.3.1 Dual-Energy Colonography is tagged previously to the examination and then

CT colonography has been suggested as an virtually subtracted by DECT post-processing,
alternative to optical colonoscopy for screening facilitating accurate assessment of the bowel
and diagnosis of colorectal carcinoma. There wall (Cai et  al. 2013). Second, to differentiate
are two areas in which DECT may be useful for findings such as stool, lipomas or adenomas
improving the diagnostic assessment in CT from colorectal cancer by means of material-
colonography: First, by using it for a procedure specific images such as iodine maps (Schaeffer
known as “electronic cleansing,” in which stool et al. 2014).
Gastrointestinal Imaging: Oncology (Liver, Pancreas, Bowel Cancer, and Treatment Response) 229

5 Oncologic Applications (Lennartz et al. 2019b, 2020). Similar results were

of DECT in the Abdomen found for virtual monoenergetic images at lower
outside of Parenchymal keV levels (Darras et  al. 2019). However, it is
Organs
Table 4 Overview on applications of different DECT-­
Apart from improved lesion delineation or char- derived reconstructions for imaging of the abdominal cavity
acterization within parenchymal organs, DECT
Reconstruction Application
can also enhance the assessment of abdominal Low keV virtual Detection and delineation of
oncologic disease manifestations outside of monoenergetic small lesions
organs (Table  4). It has been shown that iodine images
overlay image may improve the assessment of Iodine maps Detection of peritoneal disease
peritoneal disease compared to a scenario where and differentiation from
postoperative changes
only conventional images are available (Fig.  4) Lymph node characterization

a b

Fig. 4  Patient with newly diagnosed nodular peritoneal tional polyenergetic image (a), they are clearly depicted in
metastases adjacent to the anterior abdominal wall. the virtual monoenergetic image at 40 keV (b) as well as
Whereas the lesions are relatively subtle in the conven- the iodine map (c)
230 S. Lennartz and N. Große Hokamp

worth noting that the attenuation of peritoneal important pillar of oncologic follow-up, the
lesions in contrast-enhanced CT may vary radiological assessment of treatment response
depending on factors such as the underlying dis- is still largely reliant on simple size measure-
ease, lesion size, or the individual patterns of dis- ments which often do not adequately reflect the
ease spread in the abdominal cavity. Therefore, tumor heterogeneity and the biological
the diagnostic benefits provided by DECT-derived response to novel treatment options outside the
virtual monoenergetic or iodine overlay images realm of traditional cytotoxic therapies. A
remain to be verified at a larger scale. Another plethora of novel imaging biomarkers has
field of application at which DECT might com- therefore been investigated in this regard,
plement the diagnostic assessment outside of ded- among which quantitative DECT has shown
icated parenchymal organs is the detection of promising results for assessing treatment
nodal disease. In this respect, an explorative study response. Here, iodine concentration measure-
on patients with known rectal cancer demon- ments performed on iodine maps is the most
strated that quantitative DECT parameters yielded commonly applied approach.
a comparable accuracy in determining N-stage as The ex vivo validity of iodine map-based mea-
MRI (Al-Najami et  al. 2017). In another study, surements has been thoroughly investigated and
Sun et al. demonstrated that iodine concentration confirmed for all clinically available scanner
and fat fraction derived from DECT could help types. Adapting these data into clinical routine
diagnosing lymph node metastasis in patients has led to mixed observations; most authors high-
with esophageal cancer: here, the fat fraction was light the dependency of iodine measurements on
much lower in metastatic nodes and showed accu- cardiac output. In recent, large scale studies our
racies of around 80% in differentiating metastatic group suggested to utilize normalized iodine con-
from non-metastatic nodes, while their short axis centration measurements as this reduces the
diameter was not significantly different (Sun et al. inter-individual variability, age, and gender-­
2020). Another clinically challenging scenario dependency and (which is of utmost importance
that may occur is that it is unclear whether a nodal for treatment response assessment) the intra-­
mass is due to metastatic spread from a particular individual consistency of quantitative iodine
tumor or if this mass is representing a nodal mani- measurements (Lennartz et al. 2019a; Zopfs et al.
festation of lymphoma. In this regard, Marin et al. 2020). The mentioned studies performed normal-
revealed that the iodine concentration as well as ization to the abdominal aorta and focused on
the fat fraction was significantly higher in lym- portal venous phase acquisitions. It needs to be
phomas as compared to lymph node metastasis emphasized that this approach still is investiga-
and that an iodine threshold of 2.0 mg/ml could tional while the clinical adoption is subject of
differentiate these two malignant diseases with a ongoing research.
sensitivity/specificity of 87% and 89%, respec- However, it is worth noting that few applica-
tively (Martin et al. 2018). It is important to note tions, particularly pertaining to treatment
that the results of most of these studies on mate- response assessment have been employed clini-
rial quantification for assessing extraorganic dis- cally yet as described in the consecutive
ease are pending validation at a larger scale. sections.

6  he Role of Dual-Energy CT
T 6.1 Assessment of Treatment
for Assessing Oncologic Response After Locoregional
Treatment Response Therapy

The rapid developments that took place in One important area in which DECT may be a
oncology over the past decade have revolution- promising approach is the response assessment
ized patient care in many areas. While being an of patients undergoing locoregional therapy of
Gastrointestinal Imaging: Oncology (Liver, Pancreas, Bowel Cancer, and Treatment Response) 231

hepatocellular carcinoma. DECT-derived iodine 6.2 Assessment of Treatment

maps reflect the iodine enhancement more accu- Response After
rately than conventional single-energy Antiangiogenic Treatment
­acquisitions. Therefore, they can serve as a sur-
rogate parameter for lesion perfusion (Skornitzke The capability of DECT to assess lesion perfu-
et al. 2018). In this respect, several studies have sion has led to several studies investigating its
shown that DECT-derived iodine maps facilitate potential use for assessing response by high-
depiction of presence or absence of residual lighting antiangiogenic treatment effects.
tumor perfusion of hepatocellular carcinoma Hellbach et al. reported that iodine quantifica-
more accurately than conventional images. For tion allowed for an improved characterization
response assessment of patients with radioembo- of response to tyrosine kinase inhibitors in
lization of hepatocellular carcinoma, Altenbernd patients with metastatic renal cell cancer: the
et al. described that more patients were classified iodine quantification of renal cell carcinoma
as having a stable disease as opposed to a pro- metastasis was significantly more sensitive in
gressive disease compared to an assessment detecting antiangiogenic treatment effects than
based on AASLD criteria (Altenbernd et  al. the corresponding Hounsfield unit measure-
2016b). In another pilot study, Bargellini et  al. ments (Hellbach et  al. 2017). Similarly,
described that volumetric iodine uptake after Schramm et al. and Apfaltrer et al. highlighted
Yttrium-90 radioembolization of hepatocellular a potential benefit of DECT in assessing
carcinoma was highly reproducible and that response to targeted treatment in gastrointesti-
patients that were defined as responders as per nal stroma tumors (Schramm et  al. 2011).
post-therapeutic reduction in volumetric iodine Correspondingly, Meyer et  al. reported that
showed a significantly higher overall survival DECT allowed an improved prediction of
compared to the patients deemed as responders response to tyrosine-kinase-inhibitor treatment
as per RECIST 1.1 or mrECIST (Bargellini et al. than established response evaluation criteria
2018). Pertaining to radiofrequency ablation, such as RECIST and Choi criteria (Meyer et al.
Vandenbroucke et al. outlined that DECT-derived 2013).
features may help distinguish successful from
unsuccessful ablative procedures by allowing an Compliance with Ethical Standards
improved differentiation of residual tumors from
post-interventional inflammatory changes Funding None.
(Vandenbroucke et  al. 2015). Zhang et  al.
reported that volumetric iodine concentration Disclosure of Interests Nils Große Hokamp receives
more accurately predicted therapeutic response speaker’s fees and research support from Philips
Healthcare. Nils Große Hokamp is consultant for Bristol-
after microwave ablation in a rabbit model Myers Squibb. Nils Große Hokamp, and Simon Lennartz
(Zhang et al. 2018). are on the editorial board of European Radiology.
Whereas most studies focusing on the use of
DECT in  locoregional therapy response assess- Studies involving human
ment were focused on the liver, Parakh et al. fur-
thermore reported that the additional use of Ethical Approval This chapter does not contain any
iodine maps helped to distinguish postsurgical studies with human participants performed by any of the
changes from local tumor recurrence in patients authors.
who underwent resection of pancreatic adenocar-
cinoma (Parakh et al. 2018).
232 S. Lennartz and N. Große Hokamp

References Chu AJ, Lee JM, Lee YJ et al (2012) Dual-source, dual-­
energy multidetector CT for the evaluation of pancre-
atic tumours. Br J Radiol 85. https://doi.org/10.1259/
Al-Najami I, Beets-Tan RGH, Madsen G, Baatrup G
bjr/26129418
(2016) Dual-energy CT of rectal cancer specimens: a
Chuang-Bo Y, Tai-Ping H, Hai-Feng D et  al (2017)
CT-based method for mesorectal lymph node charac-
Quantitative assessment of the degree of differen-
terization. Dis Colon Rectum 59:640–647. https://doi.
tiation in colon cancer with dual-energy spectral
org/10.1097/DCR.0000000000000601
CT. Abdom Radiol (New York) 42:2591–2596. https://
Al-Najami I, Lahaye MJ, Beets-Tan RGH, Baatrup G
doi.org/10.1007/s00261-­017-­1176-­6
(2017) Dual-energy CT can detect malignant lymph
Darras KE, Clark SJ, Kang H et al (2019) Virtual monoener-
nodes in rectal cancer. Eur J Radiol 90:81–88. https://
getic reconstruction of contrast-enhanced CT scans of the
doi.org/10.1016/j.ejrad.2017.02.005
abdomen and pelvis at 40 keV improves the detection of
Altenbernd J, Heusner TA, Ringelstein A et  al (2011)
peritoneal metastatic deposits. Abdom Radiol 44:422–
Dual-energy-CT of hypervascular liver lesions in
428. https://doi.org/10.1007/s00261-­018-­1733-­7
patients with HCC: investigation of image quality
El Kayal N, Lennartz S, Ekdawi S et al (2019) Value of
and sensitivity. Eur Radiol 21:738–743. https://doi.
spectral detector computed tomography for assess-
org/10.1007/s00330-­010-­1964-­7
ment of pancreatic lesions. Eur J Radiol 118:215–222.
Altenbernd J, Forsting M, Lauenstein T, Wetter A
https://doi.org/10.1016/j.ejrad.2019.07.016
(2016a) Improved image quality and detectabil-
Fulwadhva UP, Wortman JR, Sodickson AD (2016) Use
ity of hypovascular liver metastases on DECT
of dual-energy CT and iodine maps in evaluation of
with different adjusted window settings. RöFo  –
bowel disease. Radiographics 36:393–406. https://doi.
Fortschritte auf dem Gebiet der Röntgenstrahlen
org/10.1148/rg.2016150151
und der Bildgeb Verfahren 189:228–232. https://doi.
Gabbai M, Leichter I, Mahgerefteh S, Sosna J (2015)
org/10.1055/s-­0042-­115571
Spectral material characterization with dual-energy
Altenbernd J, Wetter A, Forsting M, Umutlu L (2016b)
CT: comparison of commercial and investigative
Treatment response after radioembolisation in
technologies in phantoms. Acta Radiol 56:960–969.
patients with hepatocellular carcinoma—an evalua-
https://doi.org/10.1177/0284185114545150
tion with dual energy computed-tomography. Eur J
Grant KL, Flohr TG, Krauss B et  al (2014) Assessment
Radiol Open 3:230–235. https://doi.org/10.1016/j.
of an advanced image-based technique to calculate
ejro.2016.08.002
virtual monoenergetic computed tomographic images
Bargellini I, Crocetti L, Turini FM et al (2018) Response
from a dual-energy examination to improve contrast-­
assessment by volumetric iodine uptake measurement:
to-­noise ratio in examinations using iodinated con-
preliminary experience in patients with intermediate-­
trast media. Investig Radiol 49:586–592. https://doi.
advanced hepatocellular carcinoma treated with
org/10.1097/RLI.0000000000000060
Yttrium-90 radioembolization. Cardiovasc Intervent
Große Hokamp N, Höink AJ, Doerner J et  al (2018a)
Radiol 41:1373–1383. https://doi.org/10.1007/
Assessment of arterially hyper-enhancing liver lesions
s00270-­018-­1962-­8
using virtual monoenergetic images from spectral
Bellini D, Gupta S, Ramirez-Giraldo JC et  al (2017)
detector CT: phantom and patient experience. Abdom
Use of a noise optimized monoenergetic algorithm
Radiol 43:2066–2074. https://doi.org/10.1007/
for patient-size independent selection of an opti-
s00261-­017-­1411-­1
mal energy level during dual-energy CT of the pan-
Große Hokamp N, Obmann VC, Kessner R et al (2018b)
creas. J Comput Assist Tomogr 41:39–47. https://doi.
Improved visualization of hypodense liver lesions in
org/10.1097/RCT.0000000000000492
virtual monoenergetic images from spectral detector
Bhosale P, Le O, Balachandran A et al (2015) Quantitative
CT: proof of concept in a 3D-printed phantom and
and qualitative comparison of single-source dual-­
evaluation in 74 patients. Eur J Radiol 109:114–123.
energy computed tomography and 120-kVp computed
https://doi.org/10.1016/j.ejrad.2018.11.001
tomography for the assessment of pancreatic ductal
Große Hokamp N, Obmann VC, Kessner R et al (2018c)
adenocarcinoma. J Comput Assist Tomogr 39:907–
Virtual monoenergetic images for diagnostic assess-
913. https://doi.org/10.1097/RCT.0000000000000295
ment of hypodense lesions within the liver: semiau-
Boellaard TN, Henneman ODF, Streekstra GJ et  al
tomatic estimation of window settings using linear
(2013) The feasibility of colorectal cancer detec-
models. J Comput Assist Tomogr 42:925–931. https://
tion using dual-energy computed tomography with
doi.org/10.1097/RCT.0000000000000794
iodine mapping. Clin Radiol 68:799–806. https://doi.
Große Hokamp N, Gilkeson R, Jordan MK et  al (2019)
org/10.1016/j.crad.2013.03.005
Virtual monoenergetic images from spectral detec-
Cai W, Kim SH, Lee JG, Yoshida H (2013) Informatics in
tor CT as a surrogate for conventional CT images:
radiology: dual-energy electronic cleansing for fecal-­
unaltered attenuation characteristics with reduced
tagging CT colonography. Radiographics 33:891–912.
image noise. Eur J Radiol 117:49–55. https://doi.
https://doi.org/10.1148/rg.333125039
org/10.1016/j.ejrad.2019.05.019
Gastrointestinal Imaging: Oncology (Liver, Pancreas, Bowel Cancer, and Treatment Response) 233

Gupta S, Wagner-Bartak N, Jensen CT et al (2016) Dual-­ Lin XZ, Wu ZY, Tao R et  al (2012) Dual energy spec-
energy CT of pancreatic adenocarcinoma: reproduc- tral CT imaging of insulinoma  – value in preop-
ibility of primary tumor measurements and assessment erative diagnosis compared with conventional
of tumor conspicuity and margin sharpness. Abdom multi-detector CT. Eur J Radiol 81:2487–2494. https://
Radiol 41:1317–1324. https://doi.org/10.1007/ doi.org/10.1016/j.ejrad.2011.10.028
s00261-­016-­0689-­8 Lv P, Lin XZ, Li J et  al (2011) Differentiation of small
Hanson GJ, Michalak GJ, Childs R et al (2018) Low kV hepatic hemangioma from small hepatocellular car-
versus dual-energy virtual monoenergetic CT imag- cinoma: recently introduced spectral CT method.
ing for proven liver lesions: what are the advantages Radiology 259:720–729. https://doi.org/10.1148/
and trade-offs in conspicuity and image quality? A radiol.11101425
pilot study. Abdom Radiol 43:1404–1412. https://doi. Martin SS, Czwikla R, Wichmann JL et al (2018) Dual-­
org/10.1007/s00261-­017-­1327-­9 energy CT-based iodine quantification to differentiate
Hellbach K, Sterzik A, Sommer W et  al (2017) Dual abdominal malignant lymphoma from lymph node
energy CT allows for improved characterization of metastasis. Eur J Radiol 105:255–260. https://doi.
response to antiangiogenic treatment in patients with org/10.1016/j.ejrad.2018.06.017
metastatic renal cell cancer. Eur Radiol 27:2532–2537. Mastrodicasa D, Delli Pizzi A, Patel BN (2019) Dual-­
https://doi.org/10.1007/s00330-­016-­4597-­7 energy CT of the pancreas. Semin Ultrasound
Hickethier T, Iuga A-I, Lennartz S et  al (2018) Virtual CT MRI 40:509–514. https://doi.org/10.1053/j.
monoenergetic images from a novel dual-layer spec- sult.2019.05.002
tral detector computed tomography scanner in portal Mayo-Smith WW, Gupta H, Ridlen MS et  al (1999)
venous phase: adjusted window settings depending on Detecting hepatic lesions: the added utility of CT liver
assessment focus are essential for image interpreta- window settings. Radiology 210:601–604. https://doi.
tion. J Comput Assist Tomogr 42:350–356. https://doi. org/10.1148/radiology.210.3.r99mr07601
org/10.1097/RCT.0000000000000711 McNamara MM, Little MD, Alexander LF et al (2015)
Husarik DB, Gordic S, Desbiolles L et al (2015) Advanced Multireader evaluation of lesion conspicuity in
virtual monoenergetic computed tomography of small pancreatic adenocarcinomas: complimen-
hyperattenuating and hypoattenuating liver lesions: tary value of iodine material density and low keV
ex-vivo and patient experience in various body sizes. simulated monoenergetic images using multipha-
Investig Radiol 50:695–702. https://doi.org/10.1097/ sic rapid kVp-­switching dual energy CT.  Abdom
RLI.0000000000000171 Imaging 40:1230–1240. https://doi.org/10.1007/
Kaltenbach B, Wichmann JL, Pfeifer S et al (2018) Iodine s00261-­014-­0274-­y
quantification to distinguish hepatic neuroendocrine Meyer M, Hohenberger P, Apfaltrer P et  al (2013)
tumor metastasis from hepatocellular carcinoma at CT-based response assessment of advanced gastro-
dual-source dual-energy liver CT. Eur J Radiol 105:20– intestinal stromal tumor: dual energy CT provides a
24. https://doi.org/10.1016/j.ejrad.2018.05.019 more predictive imaging biomarker of clinical benefit
Kammerer S, Höink AJ, Wessling J et al (2015) Abdominal than RECIST or Choi criteria. Eur J Radiol 82:923–
and pelvic CT: is positive enteric contrast still neces- 928. https://doi.org/10.1016/j.ejrad.2013.01.006
sary? Results of a retrospective observational study. Nagayama Y, Tanoue S, Inoue T et al (2020) Dual-layer
Eur Radiol 25:669–678. https://doi.org/10.1007/ spectral CT improves image quality of multipha-
s00330-­014-­3446-­9 sic pancreas CT in patients with pancreatic ductal
Laukamp KR, Tirumani SH, Lennartz S et  al (2021) adenocarcinoma. Eur Radiol 30:394–403. https://doi.
Evaluation of equivocal small cystic pancreatic org/10.1007/s00330-­019-­06337-­y
lesions with spectral-detector computed tomog- Nattenmüller J, Hosch W, Nguyen T-T et  al (2015)
raphy. Acta Radiol 62:172–181. https://doi. Hypodense liver lesions in patients with hepatic
org/10.1177/0284185120917119 steatosis: do we profit from dual-energy computed
Lennartz S, Abdullayev N, Zopfs D et  al (2019a) Intra-­ tomography? Eur Radiol 25:3567–3576. https://doi.
individual consistency of spectral detector CT-enabled org/10.1007/s00330-­015-­3772-­6
iodine quantification of the vascular and renal Parakh A, Patino M, Muenzel D et al (2018) Role of rapid
blood pool. Eur Radiol 29:6581–6590. https://doi. kV-switching dual-energy CT in assessment of post-­
org/10.1007/s00330-­019-­06266-­w surgical local recurrence of pancreatic adenocarci-
Lennartz S, Große Hokamp N, Abdullayev N et al (2019b) noma. Abdom Radiol (New York) 43:497–504. https://
Diagnostic value of spectral reconstructions in detect- doi.org/10.1007/s00261-­017-­1390-­2
ing incidental skeletal muscle metastases in CT stag- Patel BN, Alexander L, Allen B et al (2017) Dual-energy
ing examinations. Cancer Imaging 19:1–8. https://doi. CT workflow: multi-institutional consensus on stan-
org/10.1186/s40644-­019-­0235-­3 dardization of abdominopelvic MDCT protocols.
Lennartz S, Zopfs D, Abdullayev N et  al (2020) Iodine Abdom Radiol (NY) 42:676. https://doi.org/10.1007/
overlays to improve differentiation between perito- s00261-­016-­0966-­6
neal carcinomatosis and benign peritoneal lesions. Eur Patel BN, Rosenberg M, Vernuccio F et  al (2018)
Radiol. https://doi.org/10.1007/s00330-­020-­06729-­5 Characterization of small incidental indeterminate
234 S. Lennartz and N. Große Hokamp

hypoattenuating hepatic lesions: added value of ing biomarker to abdominal CT perfusion: determina-
single-­phase contrast-enhanced dual-energy CT mate- tion of appropriate trigger delays for acquisition using
rial attenuation analysis. Am J Roentgenol 211:571– bolus tracking. Br J Radiol 91. https://doi.org/10.1259/
579. https://doi.org/10.2214/AJR.17.19170 bjr.20170351
Prokop M, Galanski M (2002) Spiral and multislice com- Sun X, Niwa T, Ozawa S, et al (2020) Detecting lymph node
puted tomography of the body. Thieme, Leipzig metastasis of esophageal cancer on dual-energy com-
Quiney B, Harris A, McLaughlin P, Nicolaou S (2015) puted tomography. Acta Radiol 028418512098014.
Dual-energy CT increases reader confidence in the https://doi.org/10.1177/0284185120980144
detection and diagnosis of hypoattenuating pancre- Vandenbroucke F, Van Hedent S, Van Gompel G et  al
atic lesions. Abdom Imaging 40:859–864. https://doi. (2015) Dual-energy CT after radiofrequency ablation
org/10.1007/s00261-­014-­0254-­2 of liver, kidney, and lung lesions: a review of features.
Reimer RP, Große Hokamp N, Fehrmann Efferoth A Insights Imaging 6:363–379. https://doi.org/10.1007/
et  al (2020) Virtual monoenergetic images from s13244-­015-­0408-­y
spectral detector computed tomography facilitate Yin Q, Zou X, Zai X et al (2015) Pancreatic ductal ade-
washout assessment in arterially hyper-enhancing nocarcinoma and chronic mass-forming pancreatitis:
liver lesions. Eur Radiol. https://doi.org/10.1007/ differentiation with dual-energy MDCT in spectral
s00330-­020-­07379-­3 imaging mode. Eur J Radiol 84:2470–2476. https://
Robinson E, Babb J, Chandarana H, Macari M doi.org/10.1016/j.ejrad.2015.09.023
(2010) Dual source dual energy MDCT.  Investig Yu Y, Lin X, Chen K et al (2013) Hepatocellular carcinoma
Radiol 45:413–418. https://doi.org/10.1097/ and focal nodular hyperplasia of the liver: differentia-
RLI.0b013e3181dfda78 tion with CT spectral imaging. Eur Radiol 23:1660–
Schaeffer B, Johnson TRC, Mang T et  al (2014) Dual-­ 1668. https://doi.org/10.1007/s00330-­012-­2747-­0
energy CT colonography for preoperative “one-stop” Zhang L, Wang N, Mao J et  al (2018) Dual-energy CT–
staging in patients with colonic neoplasia. Acad derived volumetric iodine concentration for the assess-
Radiol 21:1567–1572. https://doi.org/10.1016/j. ment of therapeutic response after microwave ablation
acra.2014.07.019 in a rabbit model with intrahepatic VX2 tumor. J Vasc
Schramm N, Schlemmer M, Englhart E et  al (2011) Interv Radiol 29:1455–1461. https://doi.org/10.1016/j.
Dual energy CT for monitoring targeted therapies in jvir.2018.04.019
patients with advanced gastrointestinal stromal tumor: Zopfs D, Graffe J, Reimer RP et  al (2021) Quantitative
initial results. Curr Pharm Biotechnol 12:547–557. distribution of iodinated contrast media in body com-
https://doi.org/10.2174/138920111795164066 puted tomography: data from a large reference cohort.
Shuman WP, Green DE, Busey JM et  al (2014) Dual-­ Eur Radio 31(4):2340–2348. https://doi.org/10.1007/
energy liver CT: effect of monochromatic imaging on s00330-020-07298-3. Epub 2020 Sep 30. PMID:
lesion detection, conspicuity, and contrast-to-noise 32997173, PMCID: PMC7979665
ratio of hypervascular lesions on late arterial phase. Am Zopfs D, Große Hokamp N, Reimer R et al (2021) Value
J Roentgenol 203:601–606. https://doi.org/10.2214/ of spectral detector CT for pretherapeutic, locore-
AJR.13.11337 gional assessment of esophageal cancer. Eur J Radiol
Skornitzke S, Fritz F, Mayer P et al (2018) Dual-energy 134. https://doi.org/10.1016/j.ejrad.2020.109423
CT iodine maps as an alternative quantitative imag-
 Gastrointestinal Imaging: Liver Fat
and Iron Quantification

Malte Niklas Bongers

Contents
1 Clinical Background   235
2 Physical Background   237
3 Scan Protocol and Contrast Injection   238
4 Post-processing   239
5  iagnostic Evaluation and Scientific Evidence 
D  240
5.1  Liver Fat   240
5.2  Liver Iron   241
6 Conclusion   243
References   243

1 Clinical Background ease (NAFLD), alcoholic fatty liver disease, and

autoimmune diseases (Chundru et al. 2013). The
The liver, as the major metabolic organ, plays the main causes of diffuse liver disease differ signifi-
crucial role in both amino acid and carbohydrate cantly worldwide. In Asia and Africa, viral infec-
metabolism as well as in fat and iron balance in tions dominate, while in Europe and North
humans. The underlying pathophysiology of dif- America, NAFLD is the leading medical cause
fuse liver parenchymal diseases is typically due (Blachier et al. 2013).
to the dysfunction of one of these metabolic Due to the increasing incidence of diabetes,
pathways. obesity, and resulting metabolic syndrome world-
Diffuse liver parenchymal disease can gener- wide, the prevalence of NAFLD in particular is
ally be divided into storage disease, vascular dis- increasing dramatically. Two subentities of
ease, and inflammatory disease. The main causes NAFLD can be differentiated as follows. The
are hepatitis B or C, non-alcoholic fatty liver dis- non-alcoholic fatty liver is characterized by an
increased storage of triglycerides in the hepato-
M. N. Bongers (*) cytes and the so-called non-alcoholic steatohepa-
Department of Diagnostic and Interventional titis (NASH), in which an inflammatory reaction
Radiology, University Hospital of Tuebingen, prevails in addition to the increased fat storage.
Tuebingen, Germany In addition, chemotherapeutic drugs can also
e-mail: [email protected]

© The Author(s), under exclusive license to Springer Nature Switzerland AG 2022 235
H. Alkadhi et al. (eds.), Spectral Imaging, Medical Radiology Diagnostic Imaging,
https://doi.org/10.1007/978-3-030-96285-2_15
236 M. N. Bongers

cause increased fat storage and inflammatory matoses include numerous forms of anemia, such
reaction of the liver (chemotherapy-induced ste- as thalassemia, sickle cell anemia, sideroblastic
atohepatitis  =  CASH) (Meunier and Larrey anemia, myelodysplastic syndrome, aplastic ane-
2020). Any liver disease that directly or indirectly mia, and rare enzyme or protein deficiency syn-
induces an inflammatory response causes fibrosis dromes (Marx 2000). Mention should also be
of the liver parenchyma in the longer term. made of the newly discovered dysmetabolic iron
Increasing fibrosis may develop into liver cirrho- overload syndrome, which is associated with
sis with the typical consequences of portal hyper- metabolic syndrome characteristics such as obe-
tension, hepatic insufficiency, and increased risk sity and type 2 diabetes mellitus but also exces-
of developing hepatocellular carcinoma. The sive alcohol consumption and hepatitis C
time course of development from hepatic steato- (Fargion 1999). Iron overload has been observed
sis to liver fibrosis and then to cirrhosis is vari- in 15% of patients with metabolic syndrome, in
able and depends on the underlying cause. 50% of patients with NAFLD, and in over 40% of
Therefore, determination of liver fat content is patients with chronic hepatitis C infection.
crucial for prognostic assessment and for treat- Excessive liver iron overload can lead to tissue
ment planning. With adequate therapy of the damage and loss of organ function. As far as
underlying cause, regression of hepatic steatosis known until today, free iron in combination with
and even liver fibrosis is possible in the long term reactive oxygen species causes an increase in
(Byrne and Targher 2015). hydroxyl radicals. These lead to cell damage,
In addition, the liver is also the main producer liver fibrosis, and cirrhosis and increase the risk
of enzymes and proteins to maintain systemic of developing hepatocellular carcinoma.
iron balance. Physiologically, an adult person has To date, invasive liver biopsy is considered the
about 4–5 g of iron. Of this, 80% is bound in the gold standard in the diagnosis of diffuse liver
red blood cells in the form of hemoglobin. parenchymal disease. Besides the typical risks of
Hepatic ferritin and hemosiderin primarily serve invasive methods, such as bleeding and infection,
as proteins for iron storage (1–2  g) and can be liver biopsy results are only representative for the
mobilized from hepatocytes during increased biopsied parenchymal area and cannot necessar-
metabolic demand. Complex feedback mecha- ily be generalized to the entire organ. Furthermore,
nisms regulate intestinal iron absorption and iron liver biopsy is unsuitable as a screening and mon-
recycling from aged red blood cells in response itoring method, as repeatability is limited (Tapper
to systemic iron demand. Dysregulation of these and Lok 2017). Thus, the need for non-invasive
metabolic pathways leads to the so-called iron-­ methods to adequately assess the liver paren-
associated diseases, of which the iron storage dis- chyma with respect to diffuse liver parenchymal
eases are relevant to this article (Brissot and disease is evident.
Loreal 2016). Sonography, computed tomography, and mag-
In iron storage diseases, primary forms, such netic resonance imaging already play an impor-
as hereditary hemochromatosis, are distinguished tant role in the diagnosis of diffuse liver
from secondary forms, which are predominantly parenchymal disease. Increasingly, quantitative
associated with ineffective erythropoiesis. examination results can be obtained with these
Hereditary hemochromatosis is an autosomal methods.
recessive disorder and a common cause of hepatic In general, for the non-invasive determination
iron overload. Numerous genes encoding meta- of liver fat content, proton magnetic resonance
bolic processes of iron balance have now been spectroscopy (H-MRS) can be considered as a
identified, allowing the differentiation of 5 sub- non-invasive reference standard due to its high
types of this disease. In secondary hemochroma- accuracy. The disadvantage of H-MRS is the lim-
toses, repeated blood transfusions are used itation to the analysis of a cuboidal volume ele-
therapeutically, which can lead to hepatic iron ment and the need for repeated measurements in
overload in the long term. Secondary hemaochro- different parts of the liver to obtain a ­representative
Gastrointestinal Imaging: Liver Fat and Iron Quantification 237

result for the entire liver. However, chemical shift With the introduction of DECT and material-­
relaxometry of MRI now allows quantitative specific decomposition algorithms, it is now pos-
results to be obtained for the determination of sible to robustly determine liver iron content in
liver fat content even for the total organ. Here, the native DECT examinations. In particular, due to
so-called proton-density-fat fraction shows good the short examination time, but also in the case of
agreement with both H-MRS and liver biopsy occasional contraindications to MRI (pacemak-
(Reeder and Sirlin 2010). ers, etc.), DECT may be used in the future as an
Sonography is also suitable for determining alternative method for determining liver iron
the fat content of the liver. In standard abdominal content (Abadia et al. 2017).
sonography, this is done purely visually by com-
paring the echogenicity of the liver and kidney,
which is increased in fatty conditions. However, 2 Physical Background
quantitative methods for fat quantification in
sonography now also exist. Here, backscatter or In conventional single-energy CT (SECT), exam-
attenuation coefficients are calculated to draw inations are acquired with fixed tube voltages.
conclusions about the degree of fatty liver con- The detected X-ray attenuation is given in
version. Nevertheless, residual inaccuracy Hounsfield units and displayed as gray values. In
remains with sonographic methods due to non-­ these SECT images, there is a high structural
ideal interobserver agreement (Ferraioli and information content with respect to the material
Soares Monteiro 2019). being radiographed. However, since the gray val-
In single-energy computed tomography, a ues depicted only represent the linear attenuation
semiquantitative determination of liver fat con- coefficient of the radiographed material, the
tent is possible using the Hounsfield units, which material-specific information content is very low.
decrease with increasing fat content. However, Applied to the liver, this means that under physi-
since the Hounsfield units represent a sum atten- ological normal conditions in contrast-enhanced
uation coefficient, there is a strong influence, e.g. SECT, the linear attenuation coefficient is com-
by iodine-containing contrast medium or iron posed of the attenuation of the liver parenchyma
deposits that can lead to incorrect estimation of and the intravenously applied iodine. However,
the liver fat content (Kodama et al. 2007). With both the amount of iodine present in the liver
the introduction of dual-energy computed tomog- parenchyma and, in particular, the composition
raphy (DECT), it is now possible to accurately of the liver parenchyma can vary considerably.
determine liver fat content even in contrast-­ Here, for example, the acquired contrast medium
enhanced CT examinations based on material-­ phase and administered contrast medium quan-
specific attenuation properties (Fischer et  al. tity play a role. In addition, there are effects, such
2011; Hyodo et al. 2017a). as the deposition of iron or copper components
In non-invasive diagnosis of hepatic iron over- and fatty degeneration of the liver, which have an
load, methods of MR relaxometry have become opposite effect on the linear attenuation coeffi-
established over the last decades. Initial cient and may thus completely compensate each
approaches to determine liver iron content by other.
analysis of the transverse relaxation rate (also With the introduction of DECT, it is possible
called R2) were highly vulnerable to artifacts due to obtain material-specific information about the
to long acquisition times. However, the super- material being radiographed. The basic principle
paramagnetic properties of hemosiderin lead to of DECT imaging is based on the fact that differ-
focal inhomogeneity of the main magnetic field, ent materials have energy-dependent different
which can now be measured very precisely with mass attenuation coefficients when interacting
R2* relaxometry, allowing accurate quantifica- with X-rays (for details, see Part I: Physical
tion of liver iron content (Labranche et al. 2018). Implementation: Physical Background).
238 M. N. Bongers

Due to the availability of only two attenua- 3  can Protocol and Contrast

S
tion profiles, it is in general only possible to Injection
adequately separate a mixture of two materials
using DECT.  In order to be able to separate At the beginning of the clinical DECT era, quan-
three materials from each other, further criteria tification of liver fat content was only possible
are required to solve an equation with three using a two-compartment model based on native
unknowns on the basis of the information from DECT acquisitions. By implementing the three-­
only two attenuation profiles. Meanwhile, there material models, it is now possible to determine
are some slightly different approaches to this. liver fat content even in contrast-enhanced DECT
One solution is to assume that the sum of the acquisitions. The contrast medium phase used is
individual volumes of the three materials to be of secondary importance, but the greatest experi-
distinguished is equivalent to the volume of the ence is with DECT acquisitions in portal venous
mixture of the three materials, which is not true contrast medium phase. Further development of
in every case. A more generalizable solution is the DECT scanners, e.g. by introducing an addi-
based on the law of conservation of mass and tive tin filter in the dual-source technique to
describes the assumption that the sum of the reduce the soft radiation components in the poly-
three materials to be differentiated is equivalent chromatic X-ray spectrum, has led to a further
to the mass of the mixture of these three materi- improvement in spectral separation. Thus, for a
als (for details, see Part I: Physical second-generation dual-source scanner
Implementation: Dual-Energy Algorithms and (SOMATOM Definition Flash, Siemens
Post-processing Techniques). The two required Healthcare, Erlangen, Germany), it is recom-
attenuation profiles with high and low energy mended to examine the liver with a scanning pro-
can be acquired today with all common scanner tocol that includes 100  kV as low energy and
designs, such as dual-source DECT (dsDECT), 140 kV with additive tin filter (selective photon
fast kV switching DECT (fksDECT), and dual- shield, Sn140  kV) as high energy. For a third-­
layer DECT (dlDECT) (for details, see Part I: generation dual-source scanner (SOMATOM
Physical). Theoretically, the sequential DECT Force, Siemens Healthcare, Erlangen, Germany),
technique is also capable of generating the the standard protocol to be favored for examining
required attenuation profiles, but this is not suit- the liver, depending on the patient’s diameter, is
able for liver imaging due to the time latency 100 or 90 kV and 150 kV, also with additive tin
between acquisitions and resulting artifacts filter, resulting in dose neutrality compared with
from patient motion. SECT acquisition. For routine abdominal exami-
Depending on the DECT technology used, the nations, a total collimation of 0.6  ×  64  mm
material-specific information is obtained at the (second-­generation scanner) or 0.6  ×  128  mm
image domain level from the low- and high-­ (third-generation scanner) and a pitch of 0.6 with
energy images (dsDECT) or from the low- and a rotation time of 0.5 s are suitable.
high-energy sinograms (fksDECT and dlDECT) When using the latest dlDECT scanner (IQon
before image reconstruction. and CT 7500, Philips Healthcare, Best, The
The basic materials of interest should have Netherlands), the primary acquisition of the liver
clearly differentiated mass attenuation coeffi- is performed at 120  kV.  With the brand new
cients for adequate DECT differentiation. dlDECT scanner (CT 7500, Philips Healthcare),
Applied to the liver, this highlights the problem however, it is now also possible to perform DECT
that the simultaneous presence of iron overload examinations at 100 kV, which is preferable for
and intravenously applied iodine due to over- abdominal examinations of slim persons, for
lapping mass attenuation coefficients make example. Acquisition at 140 kV, which is possi-
simultaneous determination of liver fat and ble with both scanners, is not recommended for
liver iron content problematic in contrast- the liver. With a collimation of 128 × 0.625 mm
enhanced DECT. or 256 × 0.625 mm, a pitch between 0.5 and 1.7,
Gastrointestinal Imaging: Liver Fat and Iron Quantification 239

and a rotation time of 0.33 or 0.27 s, reconstruc- ately after a contrast-enhanced MRI examina-
tions with a slice thickness of 1–5 mm are com- tion, errors may occur in the quantification of
puted with a medium soft standard kernel (B). both liver fat and liver iron content due to the
When using interactive reconstruction algo- remaining gadolinium from the MRI contrast
rithms, level 1 is recommended for the model-­ medium in the body.
based method (IMR, Philips Healthcare) and
level 3–4 for the statistical method (iDose, Philips
Healthcare). 4 Post-processing
When using DECT devices with the latest
generation fast kV switching technology Independent of the primary acquisition based on
(Revolution Apex, GE Healthcare, Waukesha, dsDECT, dlDECT, or fksDECT, the final post-­
Wisconsin), a switch between the low voltage of processing for material differentiation is based
80  kV and the high voltage of 140  kV is per- on the information of a high- and a low-energy
formed every 25  ms. A collimation of image data set. In order to be able to separate and
128 × 0.625 mm with a pitch of 0.992 at a rota- effectively quantify materials using DECT, the
tion time of 0.5  s is suitable for examining the materials must be known and defined as accu-
liver. The device-specific noise index should be rately as possible. In detail, post-processing to
between 19 and 23 for a slice thickness of quantify liver fat content involves a three-­material
0.625 mm and between 13 and 15 for 2.5 mm. decomposition of fat, liver tissue, and iodine. For
All scan protocols should have automatic tube quantification of liver iron content, iodine has to
current modulation to accommodate the varying be replaced by iron in this analysis. Since iodine
body dimensions of patients. By default, we rec- and iron are high atomic number materials (Z is
ommend reconstructing slice thicknesses 53 and 26, respectively) compared to human
between 1 and 3 mm as a compromise between body tissue, there is an overlap in the three-­
spatial resolution and the need for PACS storage material decomposition preventing simultaneous
space. For the exclusive determination of liver fat differentiation of both using DECT.
and liver iron content, slice thicknesses of The underlying process of the modified three-­
3–10 mm are certainly acceptable, too. To ensure material decomposition is best understood by a
correct post-processing of the DECT images, graphical illustration (see Fig.  1). For this pur-
quantitative kernels (Q or Qr) are used in pose, the CT numbers obtained from the high-
dsDECT.  Post-processing of the DECT data is and low-energy data sets are plotted against each
mainly done in the raw data space for both other, with the ordinate representing the low-­
dlDECT and fksDECT, so no special kernels energy information. To quantify liver fat content,
need to be used here beyond the standard kernels. a coordinate point can be defined for pure fat that
In general, low and medium levels of vendor-­ has negative CT numbers (orange dot in Fig. 1).
specific reconstruction algorithms, both on an Another point for pure liver parenchyma can be
iterative basis and using machine learning tech- defined by positive CT numbers (blue dot in
niques, produce an image impression that most Fig.  1). On a direct line connecting these two
radiologists describe as familiar. points, the CT numbers of a continuous mixture
For the determination of liver iron content, the of these two materials (fat and liver tissue) can be
acquisition of DECT images in native technique expected. The closer the measured point on the
without intravenous application of iodine-­ line connecting fat and liver tissue to pure fat, the
containing contrast medium is required, since it higher the relative fat content. In contrast-­
is methodologically impossible to differentiate enhanced DECT, however, the X-ray attenuation
two materials with high atomic number from of intravenously applied iodine must also be
each other. taken into account. As mentioned above, in
In rare cases, when in clinical practice a DECT DECT post-processing, it is critical that the mate-
examination of the liver is performed immedi- rials to be quantified are precisely known. Thus,
240 M. N. Bongers

adapted to obtain iron in the three-material

Low energy

decomposition instead of iodine. The material-­

specific slope of iron is slightly flatter than that of
iodine. To quantify the iron content, the absolute

tio
value of the distance between the measurement

e ra
point and the intersection point on the line

n
Iodi
between fat and liver tissue has to be calculated.
tio
e ra

Iodine
n
Iodi

Liver
5 Diagnostic Evaluation
Fat
Fatty liver and Scientific Evidence
Fat content
5.1 Liver Fat
High energy
The first scientific paper on liver fat quantification
Fig. 1  Visualization of the DECT three-material decom- using DECT was published in 1991 by Raptopoulos
position in the form of a low-high energy diagram.
et al. At that time, they could show using sequen-
Calculating the amount of the distance between the green
dot, which represents the measured X-ray attenuation tial DECT technique that differentiation of fatty
without the iodine component, and the orange dot, which infiltration of the liver from low density liver
represents pure adipose tissue, allows the liver fat content lesions is possible, but this was only successful if
to be determined
the liver iron content was low (Raptopoulos et al.
1991). In subsequent years, numerous authors
for iodine, the material-specific DECT ratio, have published papers comparing the accuracy of
which is the slope of a straight line that can be DECT-based liver fat quantifications with tissue
obtained from the information of a DECT histology from liver biopsies, proton magnetic
scanned dilution series, must be known (dotted resonance spectroscopy (h-MRS), chemical shift
gray lines in Fig.  1). If the information from a relaxometry of MRI, and SECT (Fig. 2).
contrast-enhanced DECT is now plotted in the Here, the extent of fatty liver degeneration is
coordinate system described above, a point (red generally classified into histological grades. A
dot in Fig. 1) is created above the described con- liver fat content of 0–33% is considered as mild,
necting line between pure fat and liver tissue. 33–66% as moderate, and >66% as severe fatty
Starting from this point, a straight line with the liver infiltration (Brunt et al. 1999).
specific slope of iodine (dotted line in Fig. 1) can The correlation analyses performed show good
be drawn. The intersection point (green dot in agreement in the vast majority of studies.
Fig. 1) between this line and the line connecting Numerous study results from small animal models
pure fat and liver tissue reflects the measured are available. In rabbits, mice, and rats, excellent
point without the iodine attenuation. By deter- agreement with histology was repeatedly demon-
mining the absolute value of the distance between strated using sequential and fksDECT in native
this intersection point (green dot in Fig.  1) and studies (Wang et  al. 2003; Artz et  al. 2012; Sun
the point for pure fat (orange dot in Fig. 1), the et al. 2014). By further developing the algorithms
liver fat content can now be quantified. The abso- for material decomposition from two- to three-
lute value of the distance on the straight line material models (Mendonca et al. 2013), Hur et al.
between the determined intersection point (green in 2014 also succeeded in achieving high agree-
dot in Fig. 1) and the measured starting point (red ment with histology in contrast-­enhanced DECT
dot in Fig.  1) represents the amount of iodine examinations in the rabbit model when quantify-
from the applied contrast medium. ing liver fat content using fksDECT (Hur et  al.
When quantifying liver iron content, the 2014). Hyodo et  al. showed in an experimental
above-described post-processing has to be ex vivo work on three-­material decomposition by
Gastrointestinal Imaging: Liver Fat and Iron Quantification 241

a b c

Fig. 2  Color-coded fat maps of three clinical cases with tration with values ranging from 14% to 24%. (c) A
different degrees of fatty liver infiltration. (a) An over- 53-year-old female patient undergoing chemotherapy for
weight man aged 66 years shows no relevant hepatic fatty breast cancer presents with a marked, increasingly inho-
infiltration. (b) A 50-year-old female patient with obesity mogeneous fatty liver (CASH) by 40%
reveals moderate but inhomogeneous hepatic fatty infil-

fksDECT that object size has a small effect on the 5.2 Liver Iron
accuracy of liver fat quantification; however, iron
deposition (and other metals with high atomic Some studies are already available on DECT-­
number such as copper in Wilson’s disease) can based determination of liver iron content. Overall,
lead to a relevant underestimation of liver fat con- the scientific evidence is still low due to the lack
tent (Hyodo et al. 2017b). In a prospective study, of studies with large cohorts and systematic com-
the same authors demonstrated in a collective of parison with liver biopsy and the established
33 patients with NAFLD that three-material methods of MRI. To date, the greatest experience
decomposition in fksDECT for determination of is in the use of dsDECT to determine liver iron
liver fat content can achieve high agreement with content (Fig. 3).
both h-MRS and liver biopsy and, moreover, very For the first time, a study on the quantification
good reproducibility (Hyodo et al. 2017a). of hepatic iron content by dsDECT in eight
When comparing DECT-based liver fat quan- patients with primary hemochromatosis was pub-
tification with the two methods of MRI, volume-­ lished by Chapman and colleagues in 1980. They
selective h-MRS should be considered separately were able to demonstrate a strong correlation of
from chemical shift relaxometry. 0.993 between DECT-based values for the assess-
In native fksDECT studies, no superiority over ment of hepatic iron content and chemical analy-
SECT compared to H-MRS could be found in the ses of liver biopsies (Chapman et al. 1980).
quantification of liver fat content (Kramer et al. In 1988, Leigthon et al. described the clinically
2017). It should be noted that hepatic iron over- reliable quantification of liver iron in a study of 78
load can lead to a relevant underestimation of children suffering from thalassemia. Using the
liver fat content in both native SECT and DECT data obtained, they developed a method for cali-
(Hyodo et al. 2017b), which can be quantified by brating conventional CT scanners to allow acqui-
DECT post-processing as described in the second sition of DECT data sets (Leighton et al. 1988).
part of this chapter. Then, in 1992, it was shown by Nielsen et al.
In the meantime, however, experimental in a small animal model that although there was a
ex  vivo approaches exist to determine liver fat pronounced correlation between the bioptically
content in native and especially contrast-­ confirmed iron content of the liver and that from
enhanced DECT examinations sufficiently by DECT examinations, there is insufficient sensi-
means of further developed three-material mod- tivity of the DECT method at clinically relevant
els even in the presence of iron deposition iron concentrations around 1–3  mg/g wet liver
(Fischer et al. 2011; Ma et al. 2014). tissue (Nielsen et al. 1992).
242 M. N. Bongers

a b

Fig. 3  Color-coded iron maps of dsDECT calculated liver iron content of 1.4–2.2 mg/dl already after infusion
with manufacturer-specific prototype software (DE of a single red cell concentrate. (b) A 67-year-old patient
IronVNC; Syngo.Via Frontier; Siemens Healthineers, shows inhomogeneous liver iron overload of 5.6–8.1 mg/
Forchheim, Germany). (a) A 76-year-old female patient dl after recurrent blood transfusions due to acute myeloid
with myelodysplastic syndrome shows a slightly elevated leukemia

These early studies used CT acquisitions in Luo et al. showed in a prospective study that
sequential technique, which are prone to motion equivalent sensitivity and specificity values can
artifacts. be obtained by means of three-material decom-
The research group led by Oelckers and col- position in dsDECT compared with MR relax-
leagues undertook a vendor-independent model ometry for the clinically relevant manifestations
calculation in 1996 and showed that the DECT-­ of hepatic iron overload. However, with the small
based method for determining liver iron could in limitation that this does not prove true for low
theory detect even lower amounts of iron than iron overloads below the therapy-relevant thresh-
MR-based methods, but that there was an unclear old of 7  mg iron/g dry liver weight (Luo et  al.
susceptibility to error, e.g. due to artifacts or par- 2015).
allel prevailing liver fatty degeneration (Oelckers In a retrospective study, Werner et al. demon-
and Graeff 1996). strated that dsDECT can adequately quantify
The clinical introduction of the first genera- liver iron overload in a collective of patients with
tion dsDECT scanner in 2006 enabled simultane- hematologic diseases and that there is a strong
ous acquisition of the low- and high-energy correlation with serum ferritin and the number of
image data for the first time, minimizing patient transfused red blood cell concentrates (Werner
motion artifacts and opening a new era of DECT. et al. 2019).
In 2011, Joe et  al. analyzed iron-associated The fksDECT technology also succeeded in
X-ray attenuation of liver parenchyma for the determining the hepatic iron content, however,
first time based on dsDECT acquisitions. The according to a first study, coexisting liver fat
research group was able to show that the differ- seems to cause a slight underestimation of the
ence in attenuation values between low- and hepatic iron content (Xie et al. 2019).
high-energy correlates with the extent of hepatic Abadia et  al. showed that second-generation
iron overload and has an accuracy equal to MR dsDECT scanners even succeed in imaging local
relaxometry for the determination of liver iron in iron distribution in the liver parenchyma (Abadia
the clinically relevant ranges and, furthermore, is et al. 2017).
not relevantly influenced by the degree of liver A study by Ma et  al. in 2020 showed that
fatty degeneration (Joe et al. 2012). dlDECT technology can be used to present a
Gastrointestinal Imaging: Liver Fat and Iron Quantification 243

strong, linear correlation with R2* values from experimental status and are not yet commercially
MR relaxometry by calculating virtual monoen- available. Therefore, MRI methods are currently
ergetic images and quantifying the attenuation still favored in routine clinical diagnostics for
difference between 50 and 120 keV to determine quantification of liver iron.
liver iron content (Ma et al. 2020).
Due to the current lack of commercially avail- Compliance with Ethical Standards
able DECT post-processing algorithms for liver
iron quantification by all three major vendors, Ethical Approval All procedures performed in studies
MR methods should continue to be preferred to involving human participants were in accordance with the
ethical standards of the institutional and/or national
quantify iron overload in routine clinical diag- research committee and with the 1964 Helsinki declara-
nostics up to now. In the future, with the avail- tion and its later amendments or comparable ethical stan-
ability of the appropriate post-processing dards. Informed consent was obtained from all individual
software, DECT may also become a suitable participants included in the study.
alternative to MRI methods for determining liver
iron content, especially when contraindications
to MRI examination exist. References
Abadia AF, Grant KL, Carey KE et al (2017) Spatial dis-
6 Conclusion tribution of iron within the normal human liver using
dual-source dual-energy CT imaging. Investig Radiol
52(11):693–700
With all today available DECT scanner designs Artz NS, Hines CD, Brunner ST et al (2012) Quantification
of dual-source, fast-kV switching, and dual-layer of hepatic steatosis with dual-energy computed tomog-
technique, it is possible to accurately quantify the raphy: comparison with tissue reference standards and
quantitative magnetic resonance imaging in the ob/ob
liver fat content in vivo. With the advancement of mouse. Investig Radiol 47(10):603–610
post-processing algorithms from two- to three-­ Blachier M, Leleu H, Peck-Radosavljevic M et  al
material models, liver fat quantification is also (2013) The burden of liver disease in Europe: a
successful in contrast-enhanced CT examina- review of available epidemiological data. J Hepatol
58(3):593–608
tions. A very good agreement can be obtained in Brissot P, Loreal O (2016) Iron metabolism and related
comparison with both liver biopsy and MRI genetic diseases: a cleared land, keeping mysteries. J
methods, with H-MRS being the non-invasive Hepatol 64(2):505–515
reference standard. The software required to Brunt EM, Janney CG, Di Bisceglie AM et  al (1999)
Nonalcoholic steatohepatitis: a proposal for grad-
evaluate DECT examinations is commercially ing and staging the histological lesions. Am J
available from the device manufacturers and Gastroenterol 94(9):2467–2474
approved for routine clinical diagnostics. Byrne CD, Targher G (2015) NAFLD: a multisystem dis-
DECT-based determination of liver iron meth- ease. J Hepatol 62(1 Suppl):S47–S64
Chapman RW, Williams G, Bydder G et  al (1980)
odologically requires acquisition of native DECT Computed tomography for determining liver iron
examinations, which is a rarity in routine daily content in primary haemochromatosis. Br Med J
use and therefore requires dedicated examination 280(6212):440–442
planning. Numerous studies have shown that reli- Chundru S, Kalb B, Arif-Tiwari H et  al (2013) MRI of
diffuse liver disease: the common and uncommon eti-
able determination of liver iron content is possi- ologies. Diagn Interv Radiol 19(6):479–487
ble with all currently available DECT scanners Fargion S (1999) Dysmetabolic iron overload syndrome.
and that there is good agreement with results Haematologica 84(2):97–98
from liver biopsies and MRI methods. The great- Ferraioli G, Soares Monteiro LB (2019) Ultrasound-based
techniques for the diagnosis of liver steatosis. World J
est experience is in the use of dsDECT scanners; Gastroenterol 25(40):6053–6062
for dlDECT and fksDECT, further studies are Fischer MA, Gnannt R, Raptis D et  al (2011)
needed for methodological establishment. In Quantification of liver fat in the presence of iron and
general, the DECT post-processing algorithms iodine: an ex-vivo dual-energy CT study. Investig
Radiol 46(6):351–358
for quantification of liver iron are still at an
244 M. N. Bongers

Hur BY, Lee JM, Hyunsik W et al (2014) Quantification assessment of iron overload in myelodysplastic syn-
of the fat fraction in the liver using dual-energy com- dromes and aplastic anemia. Jpn J Radiol 38(4):374–381
puted tomography and multimaterial decomposition. J Marx JJ (2000) Primary and secondary haemochromato-
Comput Assist Tomogr 38(6):845–852 sis. Transfus Sci 23(3):183–184
Hyodo T, Yada N, Hori M et  al (2017a) Multimaterial Mendonca PR, Lamb P, Kriston A et al (2013) Contrast-­
decomposition algorithm for the quantification of independent liver-fat quantification from spectral CT
liver fat content by using fast-kilovolt-peak switch- exams. Med Image Comput Comput Assist Interv
ing dual-energy CT: clinical evaluation. Radiology 16(Pt 1):324–331
283(1):108–118 Meunier L, Larrey D (2020) Chemotherapy-associated
Hyodo T, Hori M, Lamb P et  al (2017b) Multimaterial steatohepatitis. Ann Hepatol 19(6):597–601
decomposition algorithm for the quantification of Nielsen P, Engelhardt R, Fischer R et  al (1992)
liver fat content by using fast-kilovolt-peak switching Noninvasive liver-iron quantification by computed
dual-energy CT: experimental validation. Radiology tomography in iron-overloaded rats. Investig Radiol
282(2):381–389 27(4):312–317
Joe E, Kim SH, Lee KB et  al (2012) Feasibility and Oelckers S, Graeff W (1996) In situ measurement of iron
accuracy of dual-source dual-energy CT for nonin- overload in liver tissue by dual-energy methods. Phys
vasive determination of hepatic iron accumulation. Med Biol 41(7):1149–1165
Radiology 262(1):126–135 Raptopoulos V, Karellas A, Bernstein J et al (1991) Value
Kodama Y, Ng CS, Wu TT et al (2007) Comparison of CT of dual-energy CT in differentiating focal fatty infiltra-
methods for determining the fat content of the liver. tion of the liver from low-density masses. AJR Am J
AJR Am J Roentgenol 188(5):1307–1312 Roentgenol 157(4):721–725
Kramer H, Pickhardt PJ, Kliewer MA et  al (2017) Reeder SB, Sirlin CB (2010) Quantification of liver fat
Accuracy of liver fat quantification with advanced CT, with magnetic resonance imaging. Magn Reson
MRI, and ultrasound techniques: prospective com- Imaging Clin N Am 18(3):337–357. ix
parison with MR spectroscopy. AJR Am J Roentgenol Sun T, Lin X, Chen K (2014) Evaluation of hepatic steato-
208(1):92–100 sis using dual-energy CT with MR comparison. Front
Labranche R, Gilbert G, Cerny M et al (2018) Liver iron Biosci (Landmark Ed) 19:1377–1385
quantification with MR imaging: a primer for radiolo- Tapper EB, Lok AS (2017) Use of liver imaging and
gists. Radiographics 38(2):392–412 biopsy in clinical practice. N Engl J Med 377(8):
Leighton DM, de Campo JF, Matthews R, Sephton RG 756–768
(1988) Dual energy CT estimation of liver iron Wang B, Gao Z, Zou Q, Li L (2003) Quantitative diagno-
content in thalassaemic children. Australas Radiol sis of fatty liver with dual-energy CT. An experimental
32(2):214–219 study in rabbits. Acta Radiol 44(1):92–97
Luo XF, Xie XQ, Cheng S et al (2015) Dual-energy CT Werner S, Krauss B, Haberland U et al (2019) Dual-energy
for patients suspected of having liver iron overload: CT for liver iron quantification in patients with haema-
can virtual iron content imaging accurately quantify tological disorders. Eur Radiol 29(6):2868–2877
liver iron content? Radiology 277(1):95–103 Xie T, Li Y, He G et al (2019) The influence of liver fat
Ma J, Song ZQ, Yan FH (2014) Separation of hepatic iron deposition on the quantification of the liver-iron frac-
and fat by dual-source dual-energy computed tomog- tion using fast-kilovolt-peak switching dual-energy
raphy based on material decomposition: an animal CT imaging and material decomposition technique: an
study. PLoS One 9(10):e110964 in vitro experimental study. Quant Imaging Med Surg
Ma Q, Hu J, Yang W, Hou Y (2020) Dual-layer detector 9(4):654–661
spectral CT versus magnetic resonance imaging for the
 Bowel Imaging

Markus M. Obmann

Contents
1 Clinical Background   246
2 Physical Background   246
3 Scan Protocol Considerations   247
4 Postprocessing   248
5 Diagnostic Evaluation and Scientific Evidence   249
5.1   hronic Inflammatory Conditions 
C  249
5.2  Acute Infectious Conditions   249
5.3  Vascular Bowel Conditions   249
5.4  CT Colonography   250
5.5  Contrast Materials   252
6 Conclusion   252
References   252

Abstract from intrinsically dense feces. In infectious

and inflammatory conditions monoenergetic
In the past dual-energy CT of the bowel has images can be used to improve the visualiza-
played a lesser role compared to solid abdom- tion of inflammation associated contrast
inal organs. However, several applications for media uptake. In the same way vascular con-
bowel imaging have emerged in which dual-­ trast can be enhanced in acute or chronic
energy CT provides additional value. Virtual bowel ischemia using dual-energy CT. Dual-
non-contrast images can distinguish hyper- energy reconstructions may be used to reduce
dense intraluminal bowel contents as contrast the influence of peristalsis artifacts. Initial
enhancing such as tumors or acute bleeding studies show increased sensitivity for polyp
detection in dual-energy CT colonography
and it may be used to improve suboptimal
M. M. Obmann (*) fecal tagging.
Clinic for Radiology and Nuclear Medicine,
University Hospital Basel, Basel, Switzerland
e-mail: [email protected]

© The Author(s), under exclusive license to Springer Nature Switzerland AG 2022 245
H. Alkadhi et al. (eds.), Spectral Imaging, Medical Radiology Diagnostic Imaging,
https://doi.org/10.1007/978-3-030-96285-2_16
246 M. M. Obmann

1 Clinical Background bowel wall hemorrhage. In cases of chronic isch-

emic colitis dual-energy CT may aid the detec-
Historically the diagnostic assessment of the tion of decreased bowel wall enhancement. In the
luminal gastrointestinal tract has been challeng- same way dual-energy can increase the conspicu-
ing. Due to its ready availability cross-sectional ity of duodenal or gastric ulcers. Gastrointestinal
imaging has become the modality of choice for bleeding is common in the emergency setting and
the initial evaluation of gastrointestinal symp- dual-energy CT offers ways to either reduce radi-
toms with abdominal CT being the workhorse in ation dose or improve the readers confidence in
the acute setting. So far dual-energy CT has making the correct diagnosis.
played a lesser role in the evaluation of the lumi- Colorectal carcinoma is the second most com-
nal gastrointestinal tract compared to the solid mon cancer in the western world (Siegel et  al.
abdominal organs. This may be because of spe- 2018). Several screening methods are currently
cific challenges the bowel poses to CT imaging, in use, with CT colonography being increasingly
such as bowel wall peristalsis in combination used due to improved cost-effectiveness and
with gas–tissue interfaces. While these need to be lower complication rates (van der Meulen et al.
kept in mind when evaluating images, dual-­ 2018).
energy CT offers several unique advantages Another type of bowel cancer is gastrointesti-
which benefit bowel imaging (Yeh et  al. 2018). nal stromal tumors (GIST). Quantitative bio-
Clinical conditions of the bowel can broadly be markers for treatment response are desired
categorized into inflammatory, infectious, vascu- especially as targeted therapies are used and
lar, or neoplastic. dual-energy provides quantitative iodine metrics,
Chronic inflammatory conditions such as which have been shown to correlate with tumor
ulcerative colitis or Crohn’s disease are a source response (Schramm et  al. 2011; Meyer et  al.
of lifelong morbidity and acute episodes may be 2013). While not necessarily of gastrointestinal
complicated by abscess formation or fistulas origin, peritoneal metastases can affect the bowel
(Torres et al. 2017). Assessment of disease sever- and lead to bowel obstruction. However, in the
ity is important for acute and long-term treatment early stages peritoneal implants may be hard to
decisions and imaging plays a key role in disease detect and dual-energy CT may aid the detection
monitoring (Deepak et  al. 2017; Baker et  al. of these metastases (Darras et al. 2019).
2018). The most common acute infectious condi- While in the past oral contrast was often given
tion in the abdomen that requires surgery is for CT evaluation of abdominal problems, this
appendicitis. Appendicitis itself is usually confi- practice has been changing, and currently oral
dently diagnosed with conventional CT, as are contrast is used to a lesser extent (Kielar et  al.
major complications such as perforation or 2016; Kessner et al. 2017). However, there may
abscess formation (Tsuboi et al. 2008). However, be a renaissance of oral contrast media for two
gangrenous appendicitis which is harder to iden- reasons. Photon-counting CT is able to more pre-
tify on conventional CT may rapidly develop cisely differentiate currently available oral and
complications and a more severe clinical course intravenous contrasts, i.e. gadolinium, iodine,
(Romano et al. 2014). bismuth. Furthermore, new dual-energy CT spe-
Acute bowel ischemia is a life-threatening dis- cific contrast materials are being developed.
ease characterized by high morbidity and mortal-
ity. In conventional CT multiphasic CT is
employed to assess the vasculature and bowel 2 Physical Background
wall enhancement (Aschoff et  al. 2009). Dual-­
energy CT can be used to increase vascular con- Most clinical questions for bowel imaging relate
trast for mesenteric vessel evaluation, improve to the uptake of intravenous contrast media in
the evaluation of bowel wall enhancement as well CT. Iodine has the favorable property of a higher
as generate virtual non-contrast images to exclude k-edge (33.2  keV) compared to soft tissues
Bowel Imaging 247

(0.01–0.53  keV), which can be leveraged using associated artifacts may be reduced or exagger-
dual-energy CT as there is a significantly higher ated due to spatial and temporal differences
differential absorption by the lower energetic between the high- and low energy datasets.
photons due to the photoelectric effect (Murray
et al. 2019).
Reconstructing low energy virtual monoener- 3 Scan Protocol
getic images, the iodine-based contrast can be Considerations
increased, allowing for better vessel assessment
in portal venous phase images, and possibly skip- When acquiring dual-energy CT scans of the
ping an additional arterial phase, saving radiation bowel the complete abdomen is imaged from the
dose (Hickethier et al. 2018; Shaqdan et al. 2018). diaphragm to the pelvic symphysis. To reduce
Furthermore, not only the intravascular contrast respiratory motion artifacts, patients should be
is increased, but also the bowel wall contrast. instructed to hold their breath, as all current dual-­
On conventional CT, the HU value of a voxel energy CT scanners allow for acquisition of the
reflects the total attenuation disregarding how whole abdomen in one breath hold.
photons interact with the material in the voxel. For specific clinical questions acquisition of a
Therefore, voxels with completely different non-contrast series may be of interest (Geffroy et
materials may show identical HU values, due to al. 2014) and for vessel evaluation an arterial
different relative densities of the materials. With contrast injection phase series is desirable. There
dual-energy CT radio-dense structures, such as is mounting evidence that with dual-energy CT
active bleeding versus dense intraluminal con- both of these can be replaced using virtual
tents may be further characterized using material ­non-­contrast images and low keV virtual monoen-
decomposition images. ergetic images (Obmann et al. 2021a; Lennartz et
Intraluminal fecal material is a special chal- al. 2021).
lenge in CT colonography. The current gold stan- Hence, for bowel imaging acquisition of a
dard in CT colonography preparation comprises single portal venous phase series is
fecal tagging, where iodine or barium containing recommended.
oral contrast and cathartic medication is ingested Two main challenges need to be kept in mind
by the patient in the days prior to the scan, to when imaging bowel with CT, bowel peristalsis
remove as much fecal material as possible before and gas–tissue interfaces. As bowel peristalsis is
the exam. Not only can polyps be distinguished involuntary it cannot be changed by patient instruc-
from the contrast tagged residual fecal material, tions as compared to breathing artifacts. While
but also the contrast coats hard to spot flat lesions glucagon or other peristalsis reducing agents are in
aiding their detection (Kim et  al. 2014, 2016). use for MRI imaging (Froehlich et al. 2009), they
However, to enable virtual endoscopic views the are not regularly used in CT bowel imaging, as
tagged material needs to be removed using elec- artifacts are less severe in conventional CT acqui-
tronic cleansing (Bräuer et  al. 2018). While in sitions. As peristalsis artifacts can for some scan-
conventional CT electronic cleansing was purely ners be pronounced in dual-­energy reconstructions
based on HU thresholds, dual-energy CT offers (Obmann et al. 2021b), gantry rotation time should
the opportunity to use spectral cleansing, identi- be kept low to reduce potential artifacts (Shah
fying iodine with a higher precision compared to et  al. 2018). As the bowel lumen often contains
a pure HU-threshold base approach (Eliahou gas, gas–tissue artifacts are commonly seen in the
et al. 2010; Cai et al. 2012, 2013; Tachibana et al. form of windmill artifacts due to the high HU dif-
2015). ference between gas and tissue (Fleischmann and
Other specific challenges to the luminal gas- Boas 2011; Boas and Fleischmann 2012). To limit
trointestinal tract are bowel peristalsis in combi- these artifacts pitch should be kept at 1 or lower.
nation with gas–tissue interfaces. Depending on In terms of radiation dose, multiple studies
the technical solution of dual-energy CT used have shown the feasibility of dose neutral dual-­
248 M. M. Obmann

Table 1  Scan parameters for regular dual-energy CT bowel imaging

Scanner Dual-layer Rapid-kVp-switching Dual-source Split-filter
CTDIvol [mGy] 15 15 15 15
Tube voltage [kVp] 120 80/140 A:100 120
B:140
Additional filtration – – Sn for B Au and Sn
Tube current [mA] 511 606 A:609 617
B:471
Exposure time [s] 0.325 0.250 0.285 0.285
Collimation [mm] 64 × 0.625 80 × 0.625 128 × 0.6 64 × 0.6
Pitch 1.015 0.992 0.8 0.25
Rotation time [s] 0.33 0.5 0.28 0.28

energy CT acquisition compared to conventional enhanced in the low keV images, inflammatory
CT (Euler et  al. 2016; Uhrig et  al. 2016; Grajo changes are also accentuated. This can also be
and Sahani 2018). Therefore, a radiation dose of leveraged to improve suboptimal fecal tagging
around 15 mGy may serve as a target value and in dual-energy CT colonography studies. In the
be adapted to the institution’s standards and same way subtle bowel wall enhancement dif-
patient population. Exemplary acquisition set- ferences due to ischemic changes can be
tings for several different dual-energy CT scan- depicted. However, to analyze bowel wall
ners for bowel imaging are provided in Table 1. enhancement or the lack thereof is more com-
For low-dose CT colonography examinations monly done using iodine maps. These can either
radiation dose levels are usually lower at around be viewed as a separate image series or a col-
4 mGy (Berrington de Gonzalez et al. 2010), and ored fusion overlay on top of their correspond-
the same low-dose levels from conventional CT ing material decomposition pair, i.e. virtual
colonography should be aimed for when using unenhanced images or water images. Either
dual-energy CT. way, iodine images should always be analyzed
together with the corresponding material
decomposition pair to be able to identify densi-
4 Postprocessing ties on iodine maps as true iodine. Other materi-
als such as calcium or bismuth also appear dense
For initial clinical review of images, a 120 kVp-­ on iodine maps and might therefore be mistaken
equivalent reconstruction is used. This may be for iodine. This is solved using the virtual unen-
either be a linear blend of the high- and low hanced images, on which they appear equally
energy dataset (dual-source and split-filter dual-­ dense. This is of special importance when evalu-
energy CT scanners), a virtual monoenergetic ating for bowel wall ischemia and intramural
image between 60 and 70  keV (rapid-kVp-­ hemorrhage, where the intrinsic density of the
switching dual-energy CT scanners), or a true mural hemorrhage may appear isodense to nor-
120 kVp acquisition (detector based dual-energy mally enhancing bowel on venous phase images.
CT scanners). Evaluation of vascular structures The combination of virtual unenhanced and
is preferably done on low keV virtual monoen- iodine images is able to identify the hemor-
ergetic image reconstructions between 40 and rhage, dense on virtual unenhanced images, ver-
50  keV depending on the scanner type and its sus contrast enhancement of the bowel wall,
noise characteristics. As iodine contrast is dense on iodine images.
Bowel Imaging 249

5 Diagnostic Evaluation attenuation of iodine containing structures

and Scientific Evidence increased but also density of fat containing tis-
sues is decreased. This is especially useful in the
5.1 Chronic Inflammatory abdomen where mesenteric fat stranding can be
Conditions the most obvious feature of disease (Fig. 2). For
acute appendicitis it has been shown that both
While in chronic inflammatory bowel disease low keV virtual monoenergetic images and iodine
studies have shown the improved contrast-to-­ overlay images may facilitate the differentiation
noise characteristics using low keV virtual between acute uncomplicated and gangrenous
monoenergetic images (Lee et al. 2018), quanti- appendicitis (Elbanna et  al. 2018). The lack of
tative metrics such as iodine concentration or appendiceal mural iodine uptake is indicative of
effective atomic number (Z) have been investi- gangrenous appendicitis (Fig. 3).
gated as potential biomarkers for disease burden
and therapy response (Villanueva Campos et al.
2018; Taguchi et al. 2018b; De Kock et al. 2019), 5.3 Vascular Bowel Conditions
with quantitative iodine concentration emerging
as a reliable biomarker correlating to the clini- When evaluating for acute bowel ischemia the
cally widely used Crohn’s Disease activity index mesenteric vasculature needs to be assessed.
score (Van Hees et al. 1980; Kim et al. 2017). All Dual-energy CT improves visualization of
currently clinical available dual-energy platforms abdominal vessels using low keV virtual monoen-
enable the quantification of mural iodine uptake ergetic images (Shaqdan et  al. 2018). This can
(Fig. 1). either be used to reduce the amount of intrave-
nous contrast used or to skip a dedicated arterial
phase acquisition (Rajiah et al. 2018; Hickethier
5.2 Acute Infectious Conditions et al. 2018; Lourenco et al. 2018). Evaluating the
bowel wall is simplified using iodine (fusion
Increased signal-to-noise ratio of low keV virtual overlay) images (Fulwadhva et al. 2016), where
monoenergetic images can help identify active perfusion deficits are easily identified. In the
inflammatory processes of the bowel, not only is same way low keV images can be used to accen-

a b c

Fig. 1  Dual-source dual-energy CT in portal venous with the highest iodine uptake (arrow) is visualized in
phase of a patient with acute exacerbation of known axial plane on the iodine fusion overlay image. (c)
Crohn’s disease. (a) 120 kVp-equivalent mixed image Freehand region-of-interest measurement of the quantita-
shows multiple stenoses with increased inflammatory tive iodine uptake of the inflamed bowel segment
iodine uptake of the bowel wall (arrows). (b) The lesion
250 M. M. Obmann

a b c

Fig. 2  Dual-source dual-energy CT in portal venous (50  keV) virtual monoenergetic image increases the
phase of a patient with acute uncomplicated, phlegmon- iodine-based attenuation while lowering the attenuation of
ous appendicitis. (a) The 120-kVp-equivalent blended fatty structures, accentuating the inflammation process.
image shows a wall thickened vermiform appendix with (c) Iodine fusion overlay images show preserved iodine
surrounding fat stranding (arrow). (b) The low keV uptake of the appendiceal wall

a b c

Fig. 3  Dual-source dual-energy CT in portal venous virtual monoenergetic image shows a decreased attenua-
phase of a patient with pathology proven acute gangre- tion of the appendiceal wall. (c) Iodine fusion overlay
nous appendicitis. (a) 120-kVp-equivalent blended image images show a lack iodine uptake of the appendiceal wall,
shows a wall-thickend vermiform appendix with sur- confirmative of gangrenous appendicitis, compare also to
rounding fat stranding (arrow). (b) The low keV (50 keV) Fig. 2c

tuate subtle differences in bowel wall enhance- tious colitis may be hard to distinguish. Dual-
ment (Potretzke et  al. 2014). A pitfall for energy CT can reveal the true diagnosis by
conventional arterial and venous phase CT imag- showing the decreased vascular supply to the
ing is bowel wall hemorrhage, which due to its affected segments (Fig. 5).
high attenuation might be misconstrued as mural Gastrointestinal bleeding can be detected with
enhancement (Rondenet et  al. 2018). With the dual-energy without the need for an additional
combination of VNC images and iodine images non-contrast phase scan, as virtual non-contrast
of dual-energy CT, true non-contrast images can images have been able to perform comparably
be simulated and a confident diagnose of bowel (Sun et al. 2015; Trabzonlu et al. 2020).
wall hemorrhage indicative of transmural necro-
sis can be made (Fig. 4).
Patients with ischemic colitis usually present 5.4 CT Colonography
with a less acute course of disease. In the major-
ity of cases the underlying cause is non-occlusive Dual-energy CT has been shown to benefit intra-
mesenteric ischemia (NOMI) (Taourel et  al. venous contrast enhanced CT colonography stud-
2008). The main imaging finding is bowel wall ies, differentiating impacted stool from polyps
thickening and the differential diagnosis of infec- (Karcaaltincaba et al. 2009; Taguchi et al. 2017).
Bowel Imaging 251

a b

c d

Fig. 4  Split-filter dual-energy CT in portal venous phase both the thickened and the non-thickened bowel segments.
of a patient with a closed-loop small bowel obstruction. (a) (d) Virtual non-contrast images confirm the absence of
120-kVp-equivalent mixed images show dilated small iodine uptake, as the density of the bowel wall seen in (a)
bowel loops with reduced mural enhancement (white is mapped to the virtual non-contrast image, indicative of
arrows) and partially thickened bowel wall (yellow arrow). bowel wall hemorrhage rather than iodine perfusion. (b)
(c) Iodine images show complete lack of iodine uptake in Iodine fusion overlay shows the same findings

a b c

Fig. 5  Dual-source dual-energy CT in portal venous focal absent bowel wall enhancement, in keeping with
phase of a patient with ischemic colitis. (a) 120kVp-­ ischemic colitis. (c) Same is visualized on the iodine
equivalent images show wall thickening of the descending fusion overlay images
colon (white arrows). (b) Iodine maps show reduced and
252 M. M. Obmann

Furthermore multiple studies have shown that Compliance with Ethical Standards
low keV virtual monoenergetic images can
improve contrast of fecal tagging CT colonogra- Disclosure of Interests The manuscript has not been
submitted or published elsewhere. None of the paper’s
phy studies (Taguchi et al. 2018a; Obmann et al. contents have been previously published in any form or
2020). language (partial or complete). The presented data, text
and theories are solely of the author.
There are no financial or other author disclosures.
5.5 Contrast Materials

Current contrast agents (i.e. iodine, barium and References

gadolinium) are hard to differentiate using dual-­
Aschoff AJ, Stuber G, Becker BW et al (2009) Evaluation
energy CT due to the proximity of their k-edges. of acute mesenteric ischemia: accuracy of biphasic
However, experimental studies have shown the mesenteric multi-detector CT angiography. Abdom
potential of multi-bin photon-counting CT to Imaging 34:345–357. https://doi.org/10.1007/
separate k-edges of multiple currently available s00261-­008-­9392-­8
Baker ME, Fletcher JG, Al-Hawary M, Bruining D (2018)
contrast agents, either for CT colonography to Interdisciplinary updates in Crohn’s disease reporting
further differentiation between fecal tagged nomenclature, and cross-sectional disease monitoring.
material and polyps (Muenzel et al. 2017) or to Radiol Clin N Am 56:691–707
differentiate oral from intravenous contrast mate- Berrington de Gonzalez A, Kim KP, Yee J (2010) CT
colonography: perforation rates and potential radia-
rial (Ren et al. 2020). In the past it has been sug- tion risks. Gastrointest Endosc Clin N Am 20:279–291
gested that other than the currently available Boas FE, Fleischmann D (2012) CT artifacts: causes and
contrast materials, high Z-elements may be used reduction techniques. Imaging Med 4:1–19. https://
for novel CT contrast agents (FitzGerald et  al. doi.org/10.2217/iim.12.13
Bräuer C, Lefere P, Gryspeerdt S et  al (2018) CT colo-
2016). Because of their distinctly different spec- nography: size reduction of submerged colorectal
tral properties compared to current iodinated polyps due to electronic cleansing and CT-window
contrast materials dual-contrast exams in a single settings. Eur Radiol 1–9. https://doi.org/10.1007/
dual-energy CT scan may become possible in the s00330-­018-­5416-­0
Cai W, Kim SH, Lee JG, Yoshida H (2012) Virtual
future (Mongan et  al. 2012, 2013; Rathnayake colon tagging for electronic cleansing in dual-energy
et al. 2016). fecal-tagging CT colonography. In: Proceedings
of the annual international conference of the IEEE
Engineering in Medicine and Biology Society,
EMBS. IEEE, pp 3736–3739
6 Conclusion Cai W, Kim SH, Lee J-G, Yoshida H (2013) Informatics in
radiology: dual-energy electronic cleansing for fecal-­
So far, the gastrointestinal tract has lagged behind tagging CT colonography. Radiographics 33:891–912.
in dual-energy CT application compared to other https://doi.org/10.1148/rg.333125039
Darras KE, Clark SJ, Kang H et al (2019) Virtual monoen-
abdominal organs. However, due the improve- ergetic reconstruction of contrast-enhanced CT scans
ments in dual-energy CT scanners with faster of the abdomen and pelvis at 40 keV improves
acquisitions and thus lower artifacts bowel imag- the detection of peritoneal metastatic deposits.
ing is becoming more and more mainstream. Abdom Radiol 44:422–428. https://doi.org/10.1007/
s00261-­018-­1733-­7
Similar to other organs most dual-energy CT De Kock I, Delrue L, Lecluyse C et  al (2019)
applications in bowel imaging revolve around Feasibility study using iodine quantification on
iodinated contrast material. Yet, as novel devel- dual-energy CT enterography to distinguish nor-
opments such as dual-energy CT colonography mal small bowel from active inflammatory Crohn’s
disease. Acta Radiol 60:679–686. https://doi.
or dual-energy specific contrast agents are on the org/10.1177/0284185118799508
horizon the importance of dual-energy for bowel Deepak P, Park SH, Ehman EC et al (2017) Crohn’s dis-
imaging will increase. ease diagnosis, treatment approach, and management
Bowel Imaging 253

paradigm: what the radiologist needs to know. Abdom really required? Acad Radiol 24:840–845. https://doi.
Radiol 42:1068–1086. https://doi.org/10.1007/ org/10.1016/j.acra.2017.01.013
s00261-­017-­1068-­9 Kielar AZ, Patlas MN, Katz DS (2016) Oral contrast for CT
Elbanna KY, Mohammed MF, Chahal T et  al (2018) in patients with acute non-traumatic abdominal and pel-
Dual-energy CT in differentiating nonperforated gan- vic pain: what should be its current role? Emerg Radiol
grenous appendicitis from uncomplicated appendi- 23:477–481. https://doi.org/10.1007/s10140-­016-­1403-­4
citis. Am J Roentgenol 1–7. https://doi.org/10.2214/ Kim DH, Hinshaw JL, Lubner MG et al (2014) Contrast
AJR.17.19274 coating for the surface of flat polyps at CT colonogra-
Eliahou R, Azraq Y, Carmi R et  al (2010) Dual- phy: a marker for detection. Eur Radiol 24:940–946.
energy based spectral electronic cleansing in non- https://doi.org/10.1007/s00330-­014-­3095-­z
cathartic computed tomography colonography: Kim DH, Matkowskyj KA, Lubner MG et al (2016) Serrated
an emerging novel technique. Semin Ultrasound, polyps at CT colonography: prevalence and characteris-
CT MRI 31:309–314. https://doi.org/10.1053/j. tics of the serrated polyp spectrum. Radiology 280:455–
sult.2010.05.005 463. https://doi.org/10.1148/radiol.2016151608
Euler A, Obmann MM, Szucs-Farkas Z et  al (2016) Kim YS, Kim SH, Ryu H, Han JK (2017) Iodine quan-
Comparison of single-source split-filter dual-energy tification on spectral detector-based dual-energy CT
abdominal CT with single-energy abdominal CT enterography: correlation with Crohn’s disease activ-
using automatic tube voltage modulation: assessment ity index. In: RSNA
of image quality and radiation dose. In: Radiological Lennartz S, Laukamp KR, Tandon Y, et al (2021)
Society of North America 2016 scientific assembly Abdominal vessel depiction on virtual triphasic spec-
and annual meeting, Chicago tral detector CT: initial clinical experience. Abdom
FitzGerald PF, Colborn RE, Edic PM et  al (2016) CT Radiol (New York) 46:3501. https://doi.org/10.1007/
image contrast of high-Z elements: phantom imag- S00261-021-03001-2
ing studies and clinical implications. Radiology 278: Lee SM, Kim SH, Ahn SJ et  al (2018) Virtual monoen-
723–733. https://doi.org/10.1148/radiol.2015150577 ergetic dual-layer, dual-energy CT enterography:
Fleischmann D, Boas FE (2011) Computed tomography – optimization of KeV settings and its added value for
old ideas and new technology. Eur Radiol 21:510–517. crohn’s disease. Eur Radiol 28:2525–2534. https://doi.
https://doi.org/10.1007/s00330-­011-­2056-­z org/10.1007/s00330-­017-­5215-­z
Froehlich JM, Daenzer M, von Weymarn C et al (2009) Lourenco PDM, Rawski R, Mohammed MF et al (2018)
Aperistaltic effect of hyoscine N-butylbromide versus Dual-energy CT iodine mapping and 40-keV monoen-
glucagon on the small bowel assessed by magnetic ergetic applications in the diagnosis of acute bowel
resonance imaging. Eur Radiol 19:1387–1393. https:// ischemia. Am J Roentgenol 211:564–570. https://doi.
doi.org/10.1007/s00330-­008-­1293-­2 org/10.2214/AJR.18.19554
Fulwadhva UP, Wortman JR, Sodickson AD (2016) Use Meyer M, Hohenberger P, Apfaltrer P et  al (2013)
of dual-energy CT and iodine maps in evaluation of CT-based response assessment of advanced gastro-
bowel disease. Radiographics 36:393–406. https://doi. intestinal stromal tumor: dual energy CT provides a
org/10.1148/rg.2016150151 more predictive imaging biomarker of clinical benefit
Geffroy Y, Boulay-Coletta I, Jullès M-C, et al (2014) than RECIST or Choi criteria. Eur J Radiol 82:923–
Increased unenhanced bowel-wall attenuation at mul- 928. https://doi.org/10.1016/j.ejrad.2013.01.006
tidetector CT is highly specific of ischemia compli- Mongan J, Rathnayake S, Fu Y et al (2012) In vivo dif-
cating small-bowel obstruction. Radiol 270:159–167. ferentiation of complementary contrast media at
https://doi.org/10.1148/radiol.13122654 dual-energy CT. Radiology 265:267–272. https://doi.
Grajo JR, Sahani DV (2018) Dual-energy CT of the abdo- org/10.1148/radiol.12120692
men and pelvis: radiation dose considerations. J Am Mongan J, Rathnayake S, Fu Y et al (2013) Extravasated
Coll Radiol 15:1128–1132. https://doi.org/10.1016/j. contrast material in penetrating abdominopelvic trauma:
jacr.2017.08.012 dual-contrast dual-energy CT for improved diagnosis –
Hickethier T, Byrtus J, Hauger M et al (2018) Utilization preliminary results in an animal model. Radiology
of virtual mono-energetic images (MonoE) derived 268:738–742. https://doi.org/10.1148/radiol.13121267
from a dual-layer spectral detector CT (SDCT) for Muenzel D, Bar-Ness D, Roessl E et al (2017) Spectral pho-
the assessment of abdominal arteries in venous con- ton-counting CT: initial experience with dual–­contrast
trast phase scans. Eur J Radiol 99:28–33. https://doi. agent K-edge colonography. Radiology 283:723–728.
org/10.1016/j.ejrad.2017.12.007 https://doi.org/10.1148/radiol.2016160890
Karcaaltincaba M, Karaosmanoglu D, Akata D et  al Murray N, Darras KE, Walstra FE et al (2019) Dual-energy
(2009) Dual energy virtual ct colonoscopy with dual CT in evaluation of the acute abdomen. Radiographics
source computed tomography: initial experience. 39:264–286. https://doi.org/10.1148/rg.2019180087
RoFo Fortschritte auf dem Gebiet der Rontgenstrahlen Obmann MM, An C, Schaefer A et  al (2020) Improved
und der Bildgeb Verfahren 181:859–862. https://doi. sensitivity and reader confidence in CT colonogra-
org/10.1055/s-­0028-­1109569 phy using dual-layer spectral CT: a phantom study.
Kessner R, Barnes S, Halpern P et  al (2017) CT for Radiology 297:99–107. https://doi.org/10.1148/
acute nontraumatic abdominal pain-is oral contrast radiol.2020200032
254 M. M. Obmann

Obmann MM, Punjabi G, Obmann VC, et al (2021a) images and iodine map for active gastrointestinal
Dual-energy CT of acute bowel ischemia. Abdom bleeding: image quality, radiation dose and diagnos-
Radiol. https://doi.org/10.1007/s00261-021-03188-4 tic performance. Eur J Radiol 84:884–891. https://doi.
Obmann MM, Sun Y, An C, et al (2021b) Bowel peristal- org/10.1016/j.ejrad.2015.01.013
sis artifact on dual-energy CT: In-Vitro study of the Tachibana R, Näppi JJ, Kim SH, Yoshida H (2015)
influence of different dual-energy CT platforms and Electronic cleansing for dual-energy CT colonography
enteric contrast agents. Am J Roentgenol. https://doi. based on material decomposition and virtual mono-
org/10.2214/ajr.21.26345 chromatic imaging. In: Hadjiiski LM, Tourassi GD
Potretzke TA, Brace CL, Lubner MG et al (2014) Early (eds) Proceedings of SPIE – the International Society
small-bowel ischemia: dual-energy CT improves for Optical Engineering, p 94140Q
conspicuity compared with conventional CT in a Taguchi N, Oda S, Imuta M et al (2017) Dual-energy CT
swine model. Radiology 275:119–126. https://doi. colonography using dual-layer spectral detector com-
org/10.1148/radiol.14140875 puted tomography: feasibility of virtual monochro-
Rajiah P, Rong R, Martinez-Rios C et  al (2018) Benefit matic imaging for electronic cleansing. In: RSNA,
and clinical significance of retrospectively obtained Chicago
spectral data with a novel detector-based spectral com- Taguchi N, Oda S, Imuta M et  al (2018a) Dual-energy
puted tomography  – initial experiences and results. computed tomography colonography using dual-layer
Clin Imaging 49:65–72. https://doi.org/10.1016/j. spectral detector computed tomography: utility of
clinimag.2017.10.019 virtual monochromatic imaging for electronic cleans-
Rathnayake S, Mongan J, Torres AS et al (2016) In vivo ing. Eur J Radiol 108:7–12. https://doi.org/10.1016/J.
comparison of tantalum, tungsten, and bismuth enteric EJRAD.2018.09.011
contrast agents to complement intravenous iodine for Taguchi N, Oda S, Kobayashi T et al (2018b) Advanced
double-contrast dual-energy CT of the bowel: enteric parametric imaging for evaluation of Crohn’s disease
contrast comparison for dual-energy CT of the bowel. using dual-energy computed tomography enterog-
Contrast Media Mol Imaging 11:254–261. https://doi. raphy. Radiol Case Reports 13:709–712. https://doi.
org/10.1002/cmmi.1687 org/10.1016/j.radcr.2018.04.002
Ren L, Rajendran K, Fletcher JG et al (2020) Simultaneous Taourel P, Aufort S, Merigeaud S et  al (2008) Imaging
dual-contrast imaging of small bowel with iodine of ischemic colitis. Radiol Clin N Am 46:909–924.
and bismuth using photon-counting-­ detector com- https://doi.org/10.1016/j.rcl.2008.06.003
puted tomography: a feasibility animal study. Torres J, Mehandru S, Colombel JF, Peyrin-Biroulet L
Investig Radiol 55:688–694. https://doi.org/10.1097/ (2017) Crohn’s disease. Lancet 389:1741–1755
RLI.0000000000000687 Trabzonlu TA, Mozaffary A, Kim D, Yaghmai V (2020)
Romano A, Parikh P, Byers P, Namias N (2014) Simple Dual-energy CT evaluation of gastrointestinal bleed-
acute appendicitis versus non-perforated gangre- ing. Abdom Radiol 45. https://doi.org/10.1007/
nous appendicitis: is there a difference in the rate of s00261-­019-­02226-­6
post-operative infectious complications? Surg Infect Tsuboi M, Takase K, Kaneda I et  al (2008) Perforated
15:517–520. https://doi.org/10.1089/sur.2013.106 and nonperforated appendicitis: defect in enhancing
Rondenet C, Millet I, Corno L et  al (2018) Increased appendiceal wall – depiction with multi–detector row
unenhanced bowel-wall attenuation: a specific CT. Radiology 246:142–147. https://doi.org/10.1148/
sign of bowel necrosis in closed-loop small-bowel radiol.2461051760
obstruction. Eur Radiol 28:4225–4233. https://doi. Uhrig M, Simons D, Kachelrieß M et al (2016) Advanced
org/10.1007/s00330-­018-­5402-­6 abdominal imaging with dual energy CT is feasible
Schramm N, Schlemmer M, Englhart E et  al (2011) without increasing radiation dose. Cancer Imaging
Dual energy CT for monitoring targeted therapies in 16:15. https://doi.org/10.1186/s40644-­016-­0073-­5
patients with advanced gastrointestinal stromal tumor: van der Meulen MP, Lansdorp-Vogelaar I, Goede SL et al
initial results. Curr Pharm Biotechnol 12:547–557. (2018) Colorectal cancer: cost-effectiveness of colo-
https://doi.org/10.2174/138920111795164066 noscopy versus CT colonography screening with par-
Shah R, Khoram R, Lambert JW et  al (2018) Effect of ticipation rates and costs. Radiology 162359. https://
gantry rotation speed and scan mode on peristalsis doi.org/10.1148/radiol.2017162359
motion artifact frequency and severity at abdomi- Van Hees PAM, Van Elteren P, Van Lier HJJ, Van
nal CT.  Abdom Radiol 43:2239–2245. https://doi. Tongeren JHM (1980) An index of inflammatory
org/10.1007/s00261-­018-­1497-­0 activity in patients with Crohn’s disease. Gut 21:279–
Shaqdan KW, Parakh A, Kambadakone AR, Sahani DV 286. https://doi.org/10.1136/gut.21.4.279
(2018) Role of dual energy CT to improve diagno- Villanueva Campos AM, Tardáguila de la Fuente G,
sis of non-traumatic abdominal vascular emergen- Utrera Pérez E et al (2018) Value of dual-energy CT
cies. Abdom Radiol 1–16. https://doi.org/10.1007/ enterography in the analysis of pathological bowel
s00261-­018-­1741-­7 segments in patients with Crohn’s disease. Radiologia
Siegel RL, Miller KD, Jemal A (2018) Cancer statis- 60:223–229. https://doi.org/10.1016/j.rx.2018.01.004
tics, 2018. CA Cancer J Clin 68:7–30. https://doi. Yeh BM, Obmann MM, Westphalen AC et al (2018) Dual
org/10.3322/caac.21442 energy computed tomography scans of the bowel: ben-
Sun H, Hou XY, Xue HD et al (2015) Dual-source dual-­ efits, pitfalls, and future directions. Radiol Clin N Am
energy CT angiography with virtual non-enhanced 56:805–819
 Role of Dual-Energy Computed
Tomography (DECT) in Acute
Abdomen

Saira Hamid, Muhammad Umer Nasir,
Aneta Kecler-Pietrzyk, Adnan Sheikh,
Nicolas Murray, Faisal Khosa, and Savvas Nicolaou

Contents
1 Introduction   257
2 Role of DECT in Non-traumatic Acute Abdomen   257
3 Bowel Ischemia   257
4 Bowel Perforation   258
5 Diverticulitis   258
6 Ruptured Abdominal Aortic Aneurysm   260
7 Gastrointestinal (GI) Hemorrhage   260
8 Gangrenous Cholecystitis   261
9 Gangrenous Appendicitis   262
10 Pancreatitis   264
11 Urinary Tract Calculi   265
12 Ovarian Torsion   266
13 DECT in the Acute Traumatic Abdomen   266
13.1  Blunt Abdominal Trauma   266
13.2  Pneumoperitoneum   267
13.3  Hemoperitoneum   267
14 Active Extravasation on DECT   267
15 Visceral Injuries on DECT   268
16 Splenic Injuries   268
17 Hepatic Injuries   268
18 Pancreatic Injuries   269

S. Hamid (*) · M. U. Nasir · A. Kecler-Pietrzyk ·

A. Sheikh · N. Murray · F. Khosa · S. Nicolaou
Vancouver General Hospital, University of British
Columbia, Vancouver, BC, Canada
e-mail: [email protected]

© The Author(s), under exclusive license to Springer Nature Switzerland AG 2022 255
H. Alkadhi et al. (eds.), Spectral Imaging, Medical Radiology Diagnostic Imaging,
https://doi.org/10.1007/978-3-030-96285-2_17
256 S. Hamid et al.

19 Urinary Tract Injuries   269

20 Diaphragmatic Injuries   270
21 Bowel and Mesenteric Injuries   271
22 Vascular Injuries   272
23 Bone Injuries   272
24 Penetrating Abdominal Trauma   272
25 Conclusions   273
References   273

Abstract dependent on MDCT.  The hemodynamically

stable patient should receive a multidetector
Acute abdominal and pelvic pain can have a computed tomography (MDCT) scan with IV
wide range of etiologies, from self-limiting contrast in the assessment of visceral and vas-
conditions to acute pathologies requiring cular injuries. DECT can increase the conspi-
emergent surgery. Emergency radiology plays cuity of traumatic solid organ and hollow
a pivotal role in the routine diagnosis and out- visceral damage by utilizing its post-­
comes of patients with these disorders. processing applications like selective iodine
Computed tomography (CT) with evolved imaging, virtual monoenergetic imaging.
protocols and technological advancements in
particular with advent of dual-energy CT
remains the first diagnostic modality of choice Abbreviations
in patients with non-traumatic surgical abdo-
men, including appendicitis, cholecystitis, CT Computed tomography
bowel ischemia, bowel obstruction, visceral DECT Dual-energy computed tomography
perforation, bowel hemorrhage, and abdomi- ED Emergency department
nal aortic aneurysm rupture. Renal calculi are GI Gastrointestinal
also frequently encountered non-surgical IV Intravenous
causes of acute abdominal pain. MDCT Multidetector computed tomography
Conversely of the spectrum, abdominal MVC Motor vehicle collisions
trauma is one of the leading causes of death VNC Virtual non- contrast
and morbidity. Mechanism of injury classifies
abdominal trauma into blunt and penetrating.
Blunt abdominal trauma usually results from
falls from heights, motor vehicle collisions Learning Objectives
(MVC), and assaults. Penetrating abdominal • To emphasize the role of dual-energy com-
trauma arises from stab wounds and gunshot puted tomography (DECT) in the non-­
wounds. In both blunt and abdominal trauma, traumatic and traumatic emergencies of the
imaging should exclude life-threatening inju- abdomen.
ries after careful examination and primary sur- • To highlight the significance of an optimized
vey. Ultrasound is rapid and portable imaging DECT protocol imperative for an accurate and
modality making it a good option for abdomi- prompt diagnosis.
nal trauma screening. However, fast scan is • To demonstrate the spectrum of the imaging
limited to look for hemoperitoneum, and eval- findings in acute traumatic and non-­traumatic
uation of intra-abdominal organ injuries is disorders of the abdomen.
Role of Dual-Energy Computed Tomography (DECT) in Acute Abdomen 257

1 Introduction 2  ole of DECT in Non-

R
traumatic Acute Abdomen
Acute abdominal pain is one of the most common
presentations in the emergency department (ED) Dual-energy CT has emerged as a promising tool
and comprises approximately 15% of all emer- in imaging the patients presenting with nontrau-
gent cases (Kamin et  al. 2003). Acute onset of matic causes of acute abdomen and has multiple
severe abdominal pain raises the concern for potential clinical applications. Use of the dual-­
intra-abdominal emergencies, such as perforated energy CT acquisition allows characterization of
solid viscera, bowel ischemia/bleeding, or acute various materials depending on their differential
visceral inflammation. Therefore, based on the attenuation and thus helps to identify the material
imaging findings and diagnosis, patients may composition (Grajo et al. 2016). Hence, selective
require immediate intervention or follow-up. identification of the iodine content can provide
Evaluation of abdominal pain in the ED requires information about the visceral enhancement,
a thorough approach, based on patient’s history, including decreased or no enhancement in case
physical examination, laboratory tests and imag- of vascular compromise or hyperenhancement in
ing studies. Despite extensive evaluation, mostly inflammation. Besides identifying iodine, its
half of these patients still remain undiagnosed, quantitative analysis can help differentiate
however with latest imaging advancements this between the inflammatory and neoplastic pathol-
number has decreased significantly. Abdominal ogies. Virtual nonenhanced images can lower the
radiographs have limited diagnostic value. patient dose by eliminating the need of perform-
However, because of the easy availability, they ing a non-contrast phase. DECT with VNC
can be used as a first line of investigation to guide images can help differentiate hemorrhage from
the treatment and further investigations. tissue enhancement or mineralization.
Prompt diagnosis and characterization of Furthermore, the ability of dual-energy CT to
abdominal injuries are essential in polytraumatic substantially reduce metallic prosthesis related
patients to avoid mortality. The most commonly artifact is an added advantage to unmask the
injured intra-abdominal organ is the spleen, fol- underlying anatomical and pathological details.
lowed by the liver and the genitourinary tract
(Soto and Anderson 2012). Computed tomogra-
phy (CT) helps distinguish injuries that require 3 Bowel Ischemia
immediate surgical or angiographic management.
DECT for abdominal and pelvic trauma is an Bowel ischemic can be seen as a consequence of
emerging application (Jeremy et al. 2018). DECT arterial embolism, venous thrombosis, hypoper-
allows us to acquire images at different Kev and fusion, vasculitis and small bowel obstruction,
improve the detection and assessment of organ particularly a closed-loop obstruction (Fig.  1).
lacerations and decreased organ perfusion by Hypoenhancement or lack of bowel wall enhance-
obtaining low Kev images. Moreover, creating ment is considered a direct sign of bowel isch-
virtual non-contrast (VNC) images by subtract- emia (Wallace et al. 2016; Potretzke et al. 2015a).
ing the iodine and iodine overlay images helps to Other cross-sectional imaging findings include
look for active bleeding, and many studies have direct visualization of the thrombus in the mesen-
shown the accuracy of iodine maps and VNC teric arterial or venous circulation. Non-specific
images with solid organs (Glazer et  al. 2014; imaging findings may include bowel distension,
Ascenti et al. 2013; Wortman et al. 2016). In this bowel wall edema, hyperenhancement of the
review article, we will discuss the spectrum of mucosa, circumferential mural thickening, and
abdominal injuries in a motor vehicle collision surrounding fat stranding. In advanced cases,
with emphasis on the CT protocols, imaging find- pneumatosis, portal/mesenteric venous gas, peri-
ings, and illustrations. tonitis/ascites, and free gas can be found. Iodine
258 S. Hamid et al.

a b c

d e f g

Fig. 1  An 82-year-old female with acute onset of abdom- non-contrast image (e), this appearance is consistent with
inal pain and elevated lactate. (a, d) Unenhanced axial and mural hemorrhage. (b, f) Post-contrast coronal and axial
coronal CT images show multiple mildly dilated small CT images show no attenuation difference or enhance-
bowel loops in the right lower quadrant with two transi- ment on the venous phase on comparison to the VNC or
tion points (yellow arrow) signifying closed loop obstruc- unenhanced images. This feature is seen as no iodine
tion. Mural hyper-attenuation (red arrow) on the uptake on iodine overlay coronal and axial images (c, g)
unenhanced images can be accurately seen on the virtual

overlay imaging and virtual monoenergetic imag- contrast is the investigation of choice. The direct
ing at lower energy levels can detect and differen- evidence of bowel perforation on CT is demon-
tiate mural hypoperfusion from normally stration of free air and fluid within peritoneal
perfused bowel wall (Wallace et  al. 2016; cavity or concealed extraluminal collection
Potretzke et  al. 2015a). Early detection of the ( h t t p s : / / r a d i o p a e d i a . o rg / a r t i c l e s / b ow e l -­
bowel ischemia is crucial to prevent development perforation-­summary). Dual-energy CT is help-
of the gangrenous bowel and can significantly ful in identifying underlying causes like bowel
impact the patient outcome (Fig. 2). ischemia, bowel related tumors, perforated diver-
ticulitis (Fig.  4), gangrenous appendicitis, and
cholecystitis. Treatment is dependent on the iden-
4 Bowel Perforation tification of underlying cause followed by surgi-
cal intervention.
Abdominal pain in hollow viscus perforation is
usually very severe and diffuse rather than local-
ized to one quadrant. Other associated symptoms 5 Diverticulitis
include fever, nausea, vomiting, dizziness, and
shortness of breath. Bowel perforation can be Diverticulosis is the herniation of the mucosa and
secondary to myriads of pathologies including the submucosa through the muscular layer of the
appendicitis, diverticulitis, stomach ulcers, bowel, which if develops inflammation results in
inflammatory bowel disease, ischemia, ingestion diverticulitis. The abdominal pain is usually
of foreign body, or long-standing bowel tumor localized to the respective abdominal quadrant,
(Fig. 3). Iatrogenic causes may include post colo- mostly left lower quadrant. Apart from the
noscopy or endoscopy bowel perforation (https:// abdominal pain, other symptoms include nausea,
radiopaedia.org/articles/bowel-­p erforation-­ vomiting, fever, and localized tenderness. If left
summary). Bowel perforation is an acute surgical untreated, the complications of diverticular dis-
emergency and CT scan with intravenous (IV) ease include focal colitis, lower gastrointestinal
Role of Dual-Energy Computed Tomography (DECT) in Acute Abdomen 259

a b c

Fig. 2  A 75-year-old female with end stage COPD who signifying ischemia (red arrow). (c) Perioperative picture
presented with abdominal pain (a). Coronal post-contrast shows the blackish discoloration (blue arrow) of the
CT image shows fluid filled ascending colon with ques- ascending colon signifying ischemic insult leading to
tionable mural enhancement (yellow arrow). (b) Coronal hemicolectomy
iodine overlay image confirms the lack of iodine uptake

a b

Fig. 3  A 63-year-old male with known history of GIST locule adjacent to it (yellow arrow), confirmed on iodine
presented with marked epigastric pain. (a) DECT post-IV map (b). (c) Coronal reformates demonstrate heteroge-
contrast in portal venous phase shows 1 cm focal defect nous partially calcified lesion involving the lesser curva-
along the lesser curvature of the stomach with small gas ture of the stomach (asterisk)
260 S. Hamid et al.

a b

Fig. 4  A 44-year-old female with suprapubic pain and vesical collection on the background of acute diverticulitis.
elevated inflammatory markers. (a) Coronal image show- (b) Sinus tract (yellow arrow) is highlighted with periph-
ing communication between the sigmoid colon and supra- eral iodine update on the overlay maps

bleeding, infection, abscess, perforation, perito- abdominal aneurysm can present with diffuse
nitis, fistula formation, and obstruction. CT with abdominal discomfort, back pain, or in advanced
IV contrast is the imaging modality of choice for cases with pulsatile abdominal mass, especially
both acute complicated and uncomplicated diver- in thin lean patients. Rupture of the abdominal
ticulitis. Colonoscopy is advised in older patients aneurysm can cause life-threating hemorrhage
and patients with long-standing history of diver- and in these patients a sudden drop of blood pres-
ticulitis, after resolution of the symptom to sure should be considered a red flag. CT can have
exclude any underlying pathology. Common typical signs of impending rupture such as aortic
findings on CT scan include segmental bowel wall edema, presence of gas, interrupted periph-
wall thickening, peri-colonic fat stranding, local- eral calcifications, aorto-enteric fistulas, and in
ized fluid, and air (Fig. 4). In complicated diver- cases of ruptured aneurysm signs like retroperito-
ticulitis, CT scan findings include localized neal hematoma and visualization of direct defect
bowel perforation with abscess formation, fistu- in wall can be seen (Rakita et al. 2007).
lous tracts, with extravasation of gas and fluid in
the peritoneal cavity in cases of perforation.
Acute complicated diverticulitis with bowel per- 7 Gastrointestinal (GI)
foration is a surgical emergency because of the Hemorrhage
associated risk of peritonitis (Suzanne Albrecht
2010). A common presentation for the patients with sus-
pected gastrointestinal hemorrhage is diffuse
severe abdominal pain and hematemesis. Though
6  uptured Abdominal Aortic
R it is imperative to obtain a careful history and
Aneurysm physical examination, a well-coordinated
approach between the referring physician and the
Abdominal aortic aneurysms can have an indo- emergency radiologist prior to the CT scan is cru-
lent course with slow rate of progression in size cial for the optimized protocol mandatory to
and no obvious symptoms. Acutely enlarging diagnose the GI hemorrhage. The oral contrast
Role of Dual-Energy Computed Tomography (DECT) in Acute Abdomen 261

should not be used in these patients, as the high 8 Gangrenous Cholecystitis

attenuation intraluminal contrast can obscure
identification of the GI hemorrhage. Cholecystitis is one of the most common causes
Dual-energy CT can be valuable in the setting of pain in the right upper quadrant. Usually, ultra-
of the GI hemorrhage, resulting in improved sound is the first investigation of choice. Typical
diagnostic accuracy (Sun et al. 2013). The ability ultrasound findings in the acute cholecystitis
to generate the virtual non-contrast (VNC) include gallstones, pericholecystic fluid, gall-
images from the post-contrast dual-energy CT bladder wall thickening, and positive Murphy’s
can effectively reduce the radiation dose by elim- sign. However, it is noteworthy that the ultra-
inating the need to perform non-contrast images sound cannot accurately differentiate between
(Fig. 1). Dual-energy CT can help to distinguish the gangrenous and simple cholecystitis, which is
high attenuation enteric contents from intralumi- decisive in the management of the patient
nal iodinated contrast extravasation using the (Ratanaprasatporn et al. 2018). CT scan can fur-
VNC reconstructions. The iodine overlay color ther demonstrate cholecystitis related complica-
map can easily detect and increase the diagnostic tions like perforation/abscess formation,
confidence to detect the presence of intraluminal hemorrhage, gas, stone lodged in the gallbladder
Iodine (Fig. 5). neck or cystic duct, and gallbladder wall gan-

a b

c d e

Fig. 5  A 64-year-old male presented with symptoms of tual non-contrast subtraction images complete subtraction
intestinal obstruction and hematemesis. (a) Axial is noted suggestive of active bleed rather than calcification
sequence of DECT abdomen with intravenous contrast (yellow arrow). (d, e) show iodine quantification with
demonstrates high density material in second part of the approximately similar iodine density in second part of
duodenum in the dependant part (yellow arrow). (b) On duodenum and aorta
DECT color iodine map, iodine uptake is seen. (c) On vir-
262 S. Hamid et al.

grene (Kim et  al. 2012; Murray et  al. 2019). 9 Gangrenous Appendicitis
Moreover, dual-energy CT can demonstrate the
areas of absent wall enhancement consistent with Acute appendicitis is one of the most common
gangrenous cholecystitis, more conspicuously as causes of acute right lower abdominal pain lead-
compared to the conventional CT (Fig. 6). These ing to surgery. In North America, incidence of
findings can alter the surgical approach with a appendicitis ranges between 75 and 83 cases per
need for an open rather than a laparoscopic cho- 1000,000 population (Sulu et  al. 2010).
lecystectomy. Dual-energy CT can be helpful in Diagnostic modalities including ultrasound, CT,
identifying the gallstones which are isodense to and MRI in selective patient population (young
the bile (Sulu et  al. 2010). Prior studies have females of childbearing age or pregnancy) have
shown that gallstones are better visualized on an important role in the diagnosis of acute appen-
monochromatic low KeV (Fig.  7). Moreover, dicitis and its related complications.
impacted stone in cystic duct resulting in Mirizzi Long-standing, progressive transmural inflam-
syndrome and gallstone in small bowel causing mation causes ischemia and necrosis of the appen-
the gallstone ileus can be easily identified by dix, thus resulting in gangrenous appendicitis
using the dual-energy gallstone application which is prone to complications such as perfora-
(Figs. 8 and 9). tion, abscess formation, and sepsis. It is important

a b

Fig. 6  A 73-year-old male presented with abdominal bladder wall, and a few areas of decreased wall enhance-
pain, jaundice, weight loss. (a) Axial CT abdomen shows ment are concerning for gangrenous cholecystitis (yellow
markedly dilated gallbladder with hyperdense content, arrow). (b) Color coded iodine map shows absent wall
extensive pericholecystic fluid, mild sloughing of the gall- enhancement suggestive of gangrenous (yellow arrow)
Role of Dual-Energy Computed Tomography (DECT) in Acute Abdomen 263

a b

c d

Fig. 7  A 43-year-old male patient presented with abdom- getic axial image reconstruction shows the correlative low
inal discomfort to emergency department. (a) 120  keV attenuation of the calculus (red arrow) signifying lipid
axial CT image shows decompressed gall bladder with rich content of the calculus. (d) 190 KeV axial monoener-
questionable intraluminal high density. (b) Dual-energy getic image shows relative hyperattenuation of the calcu-
color coded overlay image shows the presence of 2  cm lus (blue arrow)
calculus in the gall bladder. (c) 40 KeV virtual monoener-

a b c

Fig. 8  A 65-year-old female with fever and elevated biliary dilatation consistent with Mirizzi syndrome. (b)
LFTs. Multiple large gallbladder stones. (a) Coronal Iodine overlay map shows iodine uptake of the gall blad-
reformatted image of CT shows gallbladder cauli associ- der wall consistent with inflammation (red arrow). (c)
ated with circumferential gallbladder thickening with Dual-energy color coded overlay image shows variable
trace of pericholecystic fluid and fat stranding (yellow composition of the calculi with red, yellow, and green
arrow). These findings are associated with intrahepatic colors
264 S. Hamid et al.

a b c

d e f

Fig. 9  An 80-year-old male with severe abdominal pain duodenal dilatation. (b, e) Iodine overlay map highlights
and deranged LFTs. (a, d) Contrast enhanced axial CT the fistulous communication (yellow arrow). (c, f) Dual-
mages show fistula extending from the gallbladder neck to energy color coded overlay images confirm the gallstone
the first part of duodenum. Large laminated impacted gall- within the 3rd part of duodenum (red arrow)
stone in the proximal 3rd part of duodenum with upstream

to identify gangrenous appendicitis prior to sur- cal diagnosis, however, it can be difficult to differ-
gery as the rate of postoperative complications is entiate it from other causes of upper abdominal
relatively higher than uncomplicated appendicitis. pain, especially if the serum amylase and lipase lev-
Thick-walled, dilated, fluid-filled appendix along els are non-contributory. Pancreatitis can be fre-
with hyperenhancement of the mucosa and edem- quently seen as a complication of long-­standing
atous wall are the most common CT findings to gallstones or alcohol abuse. Other etiologies include
diagnose acute appendicitis. Dual-energy CT can trauma, interventional pancreato-biliary proce-
detect the presence of transmural necrosis of the dures, drugs, tumors, and hereditary pancreatitis.
wall of the appendix on the iodine overlay images A standard venous phase CT abdomen is used
and on low 40-KeV virtual monoenergetic images. in cases with low clinical suspicion; however, a
The difference between a subtle non enhancing dedicated multiphasic pancreatic protocol should
gangrenous segment of the appendix with low be used in cases with high clinical suspicion and
iodine uptake from a normal enhancing mucosa of for the follow-up of known pancreatitis. Usually,
the appendix is crucial to identify because if left no oral contrast is required and only IV contrast
untreated, it can lead to perforation (Fig. 10). The administration is preferred. CT findings range
ability of the dual-­energy CT to distinctly differ- from a normal or nearly normal pancreas, mild
entiate the gangrenous mucosa from the normal focal, or diffuse peripancreatic inflammation to
enhancing mucosa clearly adds value in patient marked pancreatic edema and associated necro-
management (Elbanna et al. 2018). sis. While diagnosing pancreatitis, careful atten-
tion should be given to the related complications
such as necrosis, fluid collections, erosions,
10 Pancreatitis thrombosis of SMV or splenic artery pseudoan-
eurysms. In the initial few days, CT findings do
Acute pancreatitis is a common abdominal emer- not necessarily correlate with the patient out-
gency and has a broad spectrum of presentations come, but later in the disease course, the CT find-
and outcomes. Although acute pancreatitis is a clini- ings have greater prognostic significance.
Role of Dual-Energy Computed Tomography (DECT) in Acute Abdomen 265

a c d

Fig. 10  An 81-year-old male with acute onset of right represent appendicolith (red arrow). There is a segment of
abdominal pain overnight. (a, b) Axial and coronal post-­ poor mucosal enhancement seen as discontinuous iodine
contrast images show thickening of the appendix (yellow uptake on the iodine overlay maps (c, d) in the anterosu-
arrow) with surrounding fat stranding and free fluid which perior wall suggesting gangrenous changes (blue arrow)
is in keeping with appendicitis. Central high-density foci

There are two main types of acute pancreati- identification of urinary tract calculi, size of cal-
tis: interstitial edematous versus necrotizing. CT culi, overall calculus burden, associated compli-
scoring systems are frequently used to character- cations while accurately identifying an alternate
ize acute pancreatitis and provide prognostic diagnosis as well.
information. Scoring is dependent on the number The majority of renal calculi are radiopaque
of collections, pseudocyst formation, necrosis, with a few exemptions such as medication
and associated complications (White et al. 1986). induced calculi (Indinavir, sulfonamide, cipro-
CT-guided aspiration is very useful when floxacin, etc.). Dual-energy CT can identify the
indicated, to distinguish sterile from infected
­ specific composition of the calculus and espe-
necrosis/fluid collections. Dual-energy CT helps cially can differentiate non-uric acid from the
to identify necrosis of the pancreatic tissue by uric acid calculi (Hidas et  al. 2010). Moreover,
detecting the iodine content. other related findings like hydroureteronephrosis,
periureteric/perinephric fat stranding, and ure-
teric wall enhancement can be seen (Fig. 11). The
11 Urinary Tract Calculi presence of phleboliths in the pelvis can some-
times make it difficult to diagnose the distal
Patients presenting with acute flank pain is usu- ureteric calculi. The provision of multiplanar
­
ally suspected to have urinary tract calculi and sagittal and coronal reconstructions and identifi-
CT KUB (Kidney, ureter, and bladder) is cation of a peripheral rim of soft tissue created by
requested. Administration of IV contrast is usu- the ureteral wall around the calculus can help in
ally not indicated in the routine cases, and it is differentiating the ureteral calculi from the
reserved for those cases in which there is a high phleboliths. CT can demonstrate a wide variety
clinical suspicion of pyelonephritis. CT KUB has of alternative diagnoses in patients presenting
the highest sensitivity and specificity for the with renal colic without urinary tract calculus.
266 S. Hamid et al.

a b c

d e f

Fig. 11  A 68-year-old male with left flank pain. (a, d) ter. (b, c, e) Coronal and axial DECT images show red color
Noncontrast coronal and axial images show high density cal- coding of the calculi (yellow arrows in b and e). (f) Graphic
culi in the interpolar region of the left kidney and distal ure- representation of the uric acid composition of the calculi

12 Ovarian Torsion 13 DECT in the Acute Traumatic

Abdomen
Ovarian torsion occurs when the ovary twists
around its ligamentous attachments. It can occur 13.1 Blunt Abdominal Trauma
at any age but is most commonly seen in women
of childbearing age. It is mostly associated with Blunt abdominal trauma is the main cause of
benign and malignant tumors. Direct CT findings mortality among all age groups. CT has become
include thickening of the fallopian tube, the the primary imaging examination for evaluating
twisted structure of the adnexa and on dual-­ blunt abdominal trauma. Fast scan plays a vital
energy iodine maps, reduced iodine uptake can role in the detection of hemoperitoneum in hemo-
be seen suggestive of disruption in the blood flow dynamically unstable patients. Identifying severe
secondary to the torsion. However, CT findings abdominal injuries is often challenging, and
are mostly non-specific and presence of the indi- many injuries may not become visible during the
rect findings including edematous swollen ovary, initial period. The presence of a seat belt protects
free fluid, and dilated fallopian tube usually sup- the chest and head but increases the chances of
port the diagnosis (Iraha et al. 2017) (Fig. 12). abdominal injuries.
Role of Dual-Energy Computed Tomography (DECT) in Acute Abdomen 267

a b

Fig. 12  A 74-year-old female presented with severe right swirl sign (yellow arrow). (c) On dual-energy analysis, no
lower quadrant pain. (a) On axial CT post-contrast portal iodine uptake is noted in adnexa suggestive of absent per-
venous phase images, septated cystic structure in the right fusion (yellow arrow)
adnexa (yellow arrow). (b) Saggital reformates shows

13.2 Pneumoperitoneum uation of 30–45  HU, and potential causes are

injury to the liver, spleen, bowel, or mesentery.
On MDCT, pneumoperitoneum is described as the Dot signs can be seen adjacent to the source of
presence of free air under the diaphragm or small hemorrhage, which can help identify the site of
gas locules trapped between the layers of the mes- bleeding.
entery, and using lung window settings can help
identify small amounts of intraperitoneal gas. In
the background of trauma, pneumoperitoneum 14 Active Extravasation
raises concerns for bowel perforation, and any area on DECT
of localized gas locules adjacent to abnormal-
looking bowel may indicate the site of injury. The importance of active extravasation of con-
trast has been demonstrated in multiple studies,
and usually, its presence indicated urgent surgical
13.3 Hemoperitoneum intervention. For example, in cases of active
extravasation of contrast along with splenic
In trauma patients, the presence of free fluid in injury, the preferred treatment is splenectomy
the peritoneal cavity likely represents blood (Dreizin and Munera 2012; Federle et al. 1998).
called hemoperitoneum having an average atten- Similarly, in patients with bowel and mesenteric
268 S. Hamid et al.

injuries, open laparotomy is indicated when improve sensitivity for detection of hypovascular
active extravasation of contrast is present. lesions (Yamada et al. 2012; Sudarski et al. 2014;
Moreover, studies done on pelvic trauma suggest Robinson et  al. 2010). Iodine quantification in
embolization and laparotomy in patients with the injures and uninjured parenchyma can be
active extravasation of contrast and pelvic frac- done using iodine-selective images. Another sig-
tures (Federle et al. 1998). nificant advantage of iodine-selective imaging of
In hemoperitoneum cases, 5–10  min of solid-organ injury is its ability to allow parenchy-
delayed scans are usually performed to look for mal hematomas to be distinguished from organ
active extravasation of contrast. If the attenuation lacerations. Hematomas are hyperdense on non-
or size of the hemoperitoneum increases on enhanced images, and subcapsular hematoma is a
delayed scans, it is consistent with active well-defined collection of blood “high attenua-
­extravasation (Marmery et al. 2007). DECT can tion” that appears as an indentation over an organ.
help in better evaluation of the active extravasa- DECT virtual non-contrast images can help iden-
tion of contrast. There is no CT evidence of active tify hyperattenuating blood products in patients
extravasation of contrast if high attenuation mate- with hematoma.
rial seen on VNC images does not correspond
with iodine density images. Moreover, in patients
with pelvic fractures, it is difficult to differentiate 16 Splenic Injuries
between high attenuation material and fracture
fragments. DECT VNC images help to differenti- The splenic injuries in the majority of cases are
ate between the two (Fig. 11). managed conservatively. Early identification of
injuries that requires surgical or angiographic
intervention is of critical importance. The AAST
15 Visceral Injuries on DECT grading system is dependent on the size of the
splenic laceration, active extravasation, pseudoa-
The surgical injury grading scale for each organ neurysm, and involvement of vascular pedicle.
was created by the American Association for the Surgical therapy is required for higher-grade
Surgery of Trauma (AAST) (Dreizin and Munera injuries (generally AAST grade III and higher).
2012). However, in AAST guidelines, CT find- The delayed phase image acquisition is useful for
ings in acute trauma like active extravasation and the definitive characterization of splenic vascular
vascular injury were not considered (Federle injury as active hemorrhage or contained vascu-
et al. 1998). Therefore, various CT grading sys- lar injury as contrast washes out in the latter.
tems have been designed by incorporating and CT features of splenic trauma including the
modifying the AAST guidelines to predict better parenchymal hypoattenuation, presence of active
which patients will not respond to conservative hemorrhage, a pseudoaneurysm, or an arteriove-
treatment and will require surgery (Marmery nous (AV) fistula. Pseudoaneurysms often can be
et al. 2007; Shanmuganathan et al. 2017). managed conservatively, whereas arteriovenous
For both types of trauma, traumatic solid fistulas generally require embolization (Federle
organ injuries manifest as geographic areas of et al. 1998; Marmery et al. 2007).
parenchyma that are relatively hypoattenuating,
with organ lacerations generally hypoattenuating
linear or branching in appearance. A contusion is 17 Hepatic Injuries
a vague, poorly defined hypodense area in a solid
organ. DECT improves the detection of lacera- The AAST liver injury scale is commonly used to
tion on low Kev images as the contrast between assess the severity of hepatic injuries. The liver
the hypovascular laceration and organ paren- injury grading scale is based on the location, size
chyma increases. This concept is taken from the of liver lacerations, hematomas, or devascular-
fact that at low Kiloelectron volt images, there is ization in higher-grade injuries.
Role of Dual-Energy Computed Tomography (DECT) in Acute Abdomen 269

a b c

d e f

Fig. 13  A 23-year-old male with blunt abdominal injury Conventional angiography image post-gel foam emboli-
after crashing into a fence while skating. (a) Axial post-­ zation of distal right hepatic artery. (c) Follow-up axial
contrast CT image shows central high density material iodine overlay CT image after embolization shows no
within a large hepatic laceration (yellow arrow). (b) pooling of iodinated material. (e, f) Follow-up CT after
Iodine map shows accumulation of iodinated material 4  weeks shows low attenuation collection in the right
within this high density focus signifying active hemor- hepatic lobe likely suggesting biloma
rhage, orange color coding (yellow arrow). (d)

CT findings include hypoattenuating lacera- injuries are difficult to characterize; however,

tions, subcapsular or intraparenchymal hemato- they become more conspicuous on peak enhance-
mas, the involvement of intrahepatic ducts and ment of the pancreatic tissue (Patel et al. 2013).
vessels, active extravasation, and the presence of Lacerations that involve more than 50% of the
large volume hemoperitoneum. DECT is helpful pancreatic thickness usually cause pancreatic
to better characterize lacerations and hypoper- ductal injury (Figs. 14 and 15). Pancreatic ­injuries
fused segments (Fig.  13) (Yamada et  al. 2012; can be subtle and difficult to identify in the early
Sudarski et al. 2014; Robinson et al. 2010). posttraumatic period; therefore, particular atten-
Major hepatic vein injury requires surgery, tion should be given to peripancreatic fluid or fat
whereas active extravasation can usually be stranding. DECT low KEV images can better
treated with endovascular intervention. Delayed delineate pancreatic duct (Rekhi et  al. 2010).
complications are the formation of pseudoaneu- DECT has shown improve visualization of hypo-
rysms, hepatic abscess, and bilomas. vascular pancreatic lesions and, therefore, can be
differentiated with lacerations (Rekhi et al. 2010;
Macari et al. 2010; Marin et al. 2010; Patel et al.
18 Pancreatic Injuries 2013; McNamara et al. 2015). Pancreatic injuries
may progress to pancreatitis.
Pancreatic trauma has higher mortality than other
solid-organ injuries (Venkatesh and Wan 2008) –
the most common site of pancreatic injuries in 19 Urinary Tract Injuries
the neck and body. Direct signs of pancreatic
injury on CT scan include contusions and lacera- The severity of renal trauma is based on the size
tion, appearing as focal areas of hypoattenuation and location of renal laceration according to the
or linear low attenuation defects. Pancreatic duct AAST grading system. Isolated renal injury
270 S. Hamid et al.

a b

c d

Fig. 14 A 25-year-old female with blunt abdominal attenuation hematoma in left superior perinephric location
trauma secondary to mountain biking accident. (a) Axial shows iodine uptake suggesting active hemorrhage (blue
CT image shows linear hypoattenuation through the pan- arrow). (c) Coronal CT image shows complex laceration
creatic tail and high density hematoma in left perinephric of the left kidney. (d) Iodine overlay map shows devascu-
location (blue arrow). (b) Axial image with iodine overlay larization of a small fragment of the anterior inferior pole
shows the laceration conspicuously confirming complete (yellow arrow) and filling defect in the renal pelvis (red
transection of the pancreatic tail (yellow arrow). High arrow), showing clot

results in contained hemorrhage within the 2011). DECT is specifically helpful to look for
Gerota fascia and perirenal fascia. Delayed excre- active renal extravasation and arteriovenous fistu-
tory CT imaging is necessary to assess the col- las requiring embolization (Dinkel et al. 2002).
lecting system and search for leaks, usually
approximately 5  min after contrast material
administration. Retrograde urethrography or 20 Diaphragmatic Injuries
voiding cystourethrography can be performed to
evaluate urethral injuries, whereas, for bladder In blunt abdominal trauma, the sudden increase
injuries, CT cystography can be done. in intra-abdominal pressure can result in dia-
CT findings, including avulsion of the pedicle phragmatic injury. CT findings include disconti-
or absent enhancement of the kidney, are a high nuity of diaphragm, herniation, and constriction
risk of renal devascularization and usually require of abdominal viscera into the thorax, and depen-
surgical treatment (Fig.  14) (McGuire et  al. dent position of the herniated viscera along the
Role of Dual-Energy Computed Tomography (DECT) in Acute Abdomen 271

a b

Fig. 15  A 29-year-old male patient with blunt abdominal full thickness of pancreatic parenchyma (yellow arrow).
trauma. (a) Axial post-contrast CT image shows low (c) T2 weighted MRI mage confirms the full thickness
attenuation in the pancreatic body which is consistent laceration (yellow arrow) with absence of surrounding
with laceration in settings of trauma. (b) Axial image of pancreatic fluid collection favoring an intact main pancre-
iodine overlay map shows the lack of iodine uptake in the atic duct

posterior chest wall (dependent viscera sign) hematoma, peritoneal extravasation of intrave-
(Iochum et  al. 2002). DECT can be helpful in nous contrast-enhanced blood, focal discontinu-
identifying focal defects and reduced enhance- ity of the bowel wall (transection), focal wall
ment of the constricted viscera. thickening, abnormal bowel wall enhancement,
and intramural hematoma (Brofman et al. 2006).
DECT can detect the change in bowel wall
21 Bowel and Mesenteric enhancement and thus extremely helpful in sus-
Injuries pected bowel injury. Iodine map images can
increase the visibility of iodine content in the
Injuries to the hollow viscera and mesentery bowel wall, and VNC images can improve the
occur in approximately 5% of patients with blunt visualization of intramural hemorrhage (Hamilton
abdominal trauma. CT findings of bowel injuries et al. 2008).
are subtle, and it is the most frequently missed In some cases, diffuse bowel wall thickening
diagnosis. The most commonly involved bowel usually may not be secondary to trauma but may
segments are the proximal jejunum and the distal be related to the hypoperfusion complex. Other
ileum. Specific signs of bowel injury include CT features of the hypoperfusion complex
272 S. Hamid et al.

include a collapse of the inferior vena cava, evaluate the lumen (Hamid et  al. 2020; Korn
small aortic diameter, peripancreatic fluid, et al. 2011).
splenic hyperenhancement, and liver heteroge-
neity (Joseph et  al. 2013; Landry et  al. 2016;
Sivit et al. 1994; Potretzke et al. 2015b; Mirvis 23 Bone Injuries
et al. 1994). DECT can detect intramural hem-
orrhage, mucosal bowel wall hyperenhancement Fractures are associate with both blunt and pen-
and differentiate ingested high attenuating etrating trauma. DECT has an added advantage
material from blood products in these patients. in assessing the age and extent of the fracture,
Complications of bowel injuries include fistu- especially in older patients and in cases with sub-
lous formations. tle findings. Bone marrow edema is a biomarker
of acute occult fracture (Wang et  al. 2013). In
patients with abdominal and pelvic trauma, vir-
22 Vascular Injuries tual noncalcium images can be used to
differentiate chronic fractures from acute
­
Retroperitoneal injuries, including major vascu- (Fig. 16) and non-displaced occult fractures.
lar injuries, are challenging to identify without
performing CT.
In hepatic lacerations, it is important to 24 Penetrating Abdominal
determine the extent of laceration and involve- Trauma
ment of hepatic veins or inferior vena cava.
Patients with vascular involvement have a Penetrating injuries such as blast injuries and gun-
higher likelihood of hemodynamic instability. shot wounds are much more complex injuries and
Similarly, in cases of splenic and renal trauma, are difficult to evaluate clinically. Trauma from
the presence of vascular injury/avulsion of the blast injuries can result in retained material in the
pedicle increases the grading scale and usually soft tissues or solid organs. Moreover, gunshot
requires intervention (Uyeda et al. 2014; Hamid wounds may have entry and exit wounds. CT is
et al. 2020). helpful in recognizing the path of the bullet and
Abdominal aortic injuries are uncommon but identifying solid and hollow organ injuries. Solid-
highly lethal. Common CT finding includes organ injuries are staged using the AAST criteria.
active extravasation, intramural, and retroperito- CT is not sensitive in detecting bowel injuries.
neal hematoma. Specific signs on DECT are large However, some of the indirect signs include free
hematoma or active extravasation of contrast-­ air and free fluid (Navsaria et al. 2009).
enhanced blood, whereas non-specific signs Management of stab wounds depends on the
include small pseudoaneurysms, intimal flaps, or location and depth of penetration. In an anterior
even thrombosis (Vlahos et al. 2012). stab wound, if it penetrates through the deep lay-
Low-kiloelectron volt monoenergetic images ers of the abdominal wall, laparotomy is indi-
can improve vascular contrast-to-noise ratios in cated by the higher likelihood of bowel injury.
the abdomen and pelvis (Vlahos et  al. 2012). Posterior stab wounds confined to the paraspinal
Therefore, vascular injuries in patients with musculature with no extension into the perito-
abdominal trauma can be evaluated using low-­ neum may not require surgical exploration. CT
kilo electron volt virtual monoenergetic images. can identify the tract of the wound through gas
In patients with severe atherosclerotic disease, locules dissecting through the muscles and the
the calcium subtraction technique can help presence of the hemoperitoneum.
Role of Dual-Energy Computed Tomography (DECT) in Acute Abdomen 273

a b c

d e

Fig. 16  A 72 female with backpain. (a) Frontal radio- ture (yellow arrow). (d, e) Coronal reformatted color
graph shows S-shaped thoracolumbar scoliosis with the coded DECT images demonstrate associated marrow
caudal convex left lumbar curve having a Cobb angle of edema, color coded as green (red arrow), and extension of
62 degrees. No obvious fracture noted. (b, c) Coronal the fracture into the end plates (blue arrow)
reformatted CT images of the lumbar spine show L1 frac-

Disclosure of Interests  All authors declare they have no

25 Conclusions conflict of interest.

Dual-energy CT, with its unique ability to iden- Ethical Approval  This article does not contain any stud-
tify the material composition, qualitative and ies with human participants performed by any of the
authors.
quantitative assessment of the iodine content, and
the provision of virtual non-contrast imaging, is a
game-changer in diagnosing the patients present-
ing with acute abdomen in emergency settings. It References
increases diagnostic confidence by accentuating
Ascenti G, Mileto A, Krauss B et al (2013) Distinguishing
the difference in iodine uptake between patho- enhancing from nonenhancing renal masses with
logical and normal tissues. dual-source dual-energy CT: iodine quantification ver-
sus standard enhancement measurements. Eur Radiol
Compliance with Ethical Standards 23(8):2288–2295
Brofman N, Atri M, Hanson JM, Grinblat L, Chughtai
T, Brenneman F (2006) Evaluation of bowel
Funding  No funding was required for the study.
274 S. Hamid et al.

and mesenteric blunt trauma with multidetector evaluation of abdominal pain. Emerg Med Clin North
CT. Radiographics 26(4):1119–1131 Am 21:61–72
Dinkel HP, Dansuer H, Triller J (2002) Blunt renal trauma: Kim JE, Lee JM, Baek JH, Han JK (2012) Initial assessment
minimally invasive management with microcatheter of dual-energy CT in patients with gallstones or bile duct
embolization – experience in nine patients. Radiology stones: can virtual nonenhanced images replace true
223(3):723–730 nonenhanced images? Am J Roentgenol 198(198):817–
Dreizin D, Munera F (2012) Blunt polytrauma: evalua- 824. https://doi.org/10.2214/AJR.11.6972
tion with 64-section whole-body CT angiography. Korn A, Bender B, Thomas C et  al (2011) Dual energy
Radiographics 32(3):609–631 CTA of the carotid bifurcation: advantage of plaque
Elbanna KY, Mohammed MF, Chahal T, Khosa F, Ali subtraction for assessment of grade of the stenosis and
IT, Berger FH, Nicolaou S (2018) Dual-energy CT morphology. Eur J Radiol 80(2):e120–e1e5
in differentiating nonperforated gangrenous appendi- Landry BA, Patlas MN, Faidi S et al (2016) Are we miss-
citis from uncomplicated appendicitis. Am J Radiol ing traumatic bowel and mesenteric injuries? Can
211:776–782 Assoc Radiol J 67(4):420–425
Federle MP, Courcoulas AP, Powell M, Ferris JV, Peitzman Macari M, Spieler B, Kim D et  al (2010) Dual-source
AB (1998) Blunt splenic injury in adults: clinical and dualenergy MDCT of pancreatic adenocarcinoma: ini-
CT criteria for management, with emphasis on active tial observations with data generated at 80 kVp and
extravasation. Radiology 206(1):137–142 at simulated weighted-average 120 kVp. AJR Am J
Glazer DI, Maturen KE, Kaza RK et  al (2014) Adrenal Roentgenol 194(1):W27–W32
incidentaloma triage with single-source (fast-­ Marin D, Nelson RC, Barnhart H et al (2010) Detection of
kilovoltage switch) dual-energy CT.  AJR Am J pancreatic tumors, image quality, and radiation dose
Roentgenol 203(2):329–335 during the pancreatic parenchymal phase: effect of a
Grajo JR, Patino M, Prochowski A, Sahani DV (2016) low-tube-voltage, high-tube-current CT technique  –
Dual energy CT in practice: basic principles and appli- preliminary results. Radiology 256(2):450–459
cations. Appl Radiol 2:6–12 Marmery H, Shanmuganathan K, Alexander MT, Mirvis
Hamid S, Nicolaou S, Khosa F, Andrews G, Murray SE (2007) Optimization of selection for nonopera-
N, Abdellatif W, Qamar SR (2020) Dual-energy tive management of blunt splenic injury: comparison
CT: a paradigm shift in acute traumatic abdomen. of MDCT grading systems. AJR Am J Roentgenol
Can Assoc Radiol J 71(3):371–387. https://doi. 189(6):1421–1427
org/10.1177/0846537120905301. Epub 2020 Mar McGuire J, Bultitude MF, Davis P, Koukounaras J, Royce
11.PMID: 32157893 PL, Corcoran NM (2011) Predictors of outcome for
Hamilton JD, Kumaravel M, Censullo ML, Cohen AM, blunt high grade renal injury treated with conservative
Kievlan DS, West OC (2008) Multidetector CT evalua- intent. J Urol 185(1):187–191
tion of active extravasation in blunt abdominal and pel- McNamara MM, Little MD, Alexander LF, Carroll LV,
vic trauma patients. Radiographics 28(6):1603–1616 Beasley TM, Morgan DE (2015) Multireader evalua-
Hidas G, Eliahou R, Duvdevani M, Coulon P, Lemaitre L, tion of lesion conspicuity in small pancreatic adeno-
Gofrit ON, Pode D, Sosna J (2010) Determination of carcinomas: complimentary value of iodine material
renal stone composition with dual-energy CT: in vivo density and low keV simulated monoenergetic images
analysis and comparison with X-ray diffraction. using multiphasic rapid kVp-switching dual energy
Radiology 257(2):394–401 CT. Abdom Imaging 40(5):1230–1240
h t t p s : / / r a d i o p a e d i a . o r g / a r t i c l e s / Mirvis SE, Shanmuganathan K, Erb R (1994) Diffuse
bowel-­perforation-­summary small-bowel ischemia in hypotensive adults after blunt
Iochum S, Ludig T, Walter F’d’r, Sebbag H, Grosdidier trauma (shock bowel): CT findings and clinical signifi-
G, Blum AG (2002) Imaging of diaphragmatic cance. AJR Am J Roentgenol 163(6):1375–1379
injury: a diagnostic challenge? Radiographics Murray N, Darras KE, Walstra FE, Mohammed MF,
22(1):S103–SS18 McLaughlin PD, Nicolaou S (2019) Dual-energy CT
Iraha Y, Okada M, Iraha R, Azama K, Yamashiro T, in evaluation of the acute abdomen. Radiographics
Tsubakimoto M, Aoki Y, Murayama S (2017) CT 39:264–286
and MR imaging of gynecologic emergencies. Navsaria PH, Nicol AJ, Krige JE et  al (2009) Selective
Radiographics 37:1569–1586 nonoperative management of liver gunshot injuries.
Jeremy R, Wortman JWU, Fulwadhva UP, Sodickson Ann Surg 249(4):653–656
AD (2018) Dual-energy CT for abdominal and pelvic Patel BN, Thomas JV, Lockhart ME, Berland LL, Morgan
trauma. Radiographics 38(2):586–602 DE (2013) Single-source dual-energy spectral multi-
Joseph DK, Kunac A, Kinler RL et al (2013) Diagnosing detector CT of pancreatic adenocarcinoma: optimiza-
blunt hollow viscus injury: is computed tomography tion of energy level viewing significantly increases
the answer? Am J Surg 205(4):414–418 lesion contrast. Clin Radiol 68(2):148–154
Kamin RA, Nowicki TA, Courtney DS, Powers RD Potretzke TA, Brace CL, Lubner MG, Sampson LA,
(2003) Pearls and pitfalls in the emergency department Willey BJ, Lee FT Jr (2015a) Early small-bowel isch-
Role of Dual-Energy Computed Tomography (DECT) in Acute Abdomen 275

emia: dal-energy CT improves conspicuity compared Sulu B, Günerhan Y, Palanci Y, Isler B, Caglayan K (2010)
with conventional CT in a swine model. Radiology Epidemiological and demographic features of appen-
275:119–126 dicitis and influences of several environment factors.
Potretzke TA, Brace CL, Lubner MG, Sampson LA, Ulus Travma Acil Cerrahi Derg 16:38–42
Willey BJ, Lee FT Jr (2015b) Early small-bowel isch- Sun H, Xue H-D, Wang Y-N, Qian J-M, Yu J-C, Zhu F,
emia: dual-energy CT improves conspicuity compared Zhu H-D, Jin Z-Y, Li X-G (2013) Dual-source dual-­
with conventional CT in a swine model. Radiology energy computed tomography angiography for active
25(1):119–126 gastrointestinal bleeding: a preliminary study. Clin
Rakita D, Newatia A, Hines JJ, Siegel DN, Friedman B Radiol 68(2):139–147
(2007) Spectrum of Ct findings in rupture and impend- Suzanne Albrecht P (2010) Management of diverticular
ing rupture of abdominal aortic aneurysms, impending disease. Gastroenterology 35:12
rupture of abdominal aortic aneurysms. Radiographics Uyeda JW, LeBedis CA, Penn DR et al (2014) Active hem-
27:497–507 orrhage and vascular injuries in splenic trauma: utility
Ratanaprasatporn L, Uyeda JW, Wortman JR, Richardson of the arterial phase in multidetector CT.  Radiology
I, Sodickson AD (2018) Multimodality imaging, 270(1):99–106
including dual-energy CT in the evaluation of gall- Venkatesh SK, Wan JM (2008) CT of blunt pancreatic
bladder disease. Radiographics 38:75–89 trauma: a pictorial essay. Eur J Radiol 67(2):311–320
Rekhi S, Anderson SW, Rhea JT, Soto JA (2010) Imaging Vlahos I, Chung R, Nair A, Morgan R (2012) Dual-energy
of blunt pancreatic trauma. Emerg Radiol 17(1):13–19 CT: vascular applications. AJR Am J Roentgenol
Robinson E, Babb J, Chandarana H, Macari M (2010) 199:s87–s97
Dual source dual energy MDCT: comparison of 80 Wallace AB, Raptis CA, Mellnick VM (2016) Imaging of
kVp and weighted average 120 kVp data for conspicu- bowel ischemis. Curr Radiol Rep 4:29
ity of hypo-vascular liver metastases. Investig Radiol Wang CK, Tsai JM, Chuang MT, Wang MT, Huang KY,
45(7):413–418 Lin RM (2013) Bone marrow edema in vertebral
Shanmuganathan K, Mirvis SE, Boyd-Kranis R, Takada T, compression fractures: detection with dual-energy
Scalea TM (2017) Nonsurgical management of blunt CT. Radiology 269(2):525–533
splenic injury: use of CT criteria to select patients for White EM, Simeone JF, Mueller PR, Simeone JF, Butch
splenic arteriography and potential endovascular ther- RJ, Warshaw AL, Neff CC, Nardi GL, Ferrucci JT
apy. Radiology 1:75–82 Jr (1986) Pancreatic necrosis: CT manifestations.
Sivit CJ, Eichelberger MR, Taylor GA (1994) CT in Radiology 158(2):343
children with rupture of the bowel caused by blunt Wortman JR, Bunch PM, Fulwadhva UP, Bonci GA,
trauma: diagnostic efficacy and comparison with Sodickson AD (2016) Dual-energy CT of incidental
hypoperfusion complex. AJR Am J Roentgenol findings in the abdomen: can we reduce the need for
163(5):1195–1198 follow-up imaging? AJR Am J Roentgenol 6:W1–W11
Soto JA, Anderson SW (2012) Multidetector CT of blunt Yamada Y, Jinzaki M, Tanami Y, Abe T, Kuribayashi S
abdominal trauma. Radiology 265(3):678–693 (2012) Virtual monochromatic spectral imaging for
Sudarski S, Apfaltrer P, Nance JW Jr et al (2014) Objective the evaluation of hypovascular hepatic metastases:
and subjective image quality of liver parenchyma and the optimal monochromatic level with fast kilovoltage
hepatic metastases with virtual monoenergetic dual-­ switching dual-energy computed tomography. Investig
source dual-energy CT reconstructions: an analysis Radiol 47(5):292–298
in patients with gastrointestinal stromal tumor. Acad
Radiol 21(4):514–522
 Spectral Computed Tomography
Imaging of the Adrenal Glands

Matthias Benndorf, August Sigle,
and Fabian Bamberg

Contents
1 Introduction   277
2 Lipid-Rich Adenoma and Virtual Non-contrast Imaging   278
3 Lipid-Poor Adenoma, Washout Analysis, and Chemical Shift MRI   281
4 Quantification of Iodine   281
5 Summary   283
References   283

Abstract lipid-rich adenoma is lower compared to unen-

hanced scans, since VNC-CT images tend to
Spectral computed tomography offers addi- systematically overestimate HU of lipid-rich
tional diagnostic information for the evaluation adenomas. The American College of
of adrenal lesions in contrast enhanced exami- Radiology’s white paper on the management of
nations. Virtual non-contrast (VNC-CT) images incidental adrenal lesions explicitly mentions
can be used as a substitute for true unenhanced dual-energy CT and the potential usage of
images to confidently diagnose lipid-rich ade- VNC-CT with a threshold of ≤10 HU for diag-
noma with a threshold of ≤10 HU, making nosis. In this chapter, we provide an overview
additional unenhanced scans or work-up unnec- and comparison of the diagnostic accuracy of
essary in these cases. However, sensitivity for unenhanced CT, VNC-CT, CT washout calcu-
lations, and chemical shift MRI for the diagno-
sis of adrenal adenoma versus non-adenoma.
M. Benndorf (*) · F. Bamberg
Department of Radiology, Medical Center -
University of Freiburg, Faculty of Medicine,
University of Freiburg, Freiburg, Germany
e-mail: [email protected]; 1 Introduction
[email protected]
A. Sigle With the increasing use of computed tomography
Department of Urology, Medical Center - University
(CT) as a widely available imaging modality in
of Freiburg, Faculty of Medicine, University of
Freiburg, Freiburg, Germany emergency diagnosis and diagnostic body imag-
e-mail: [email protected] ing for benign and malignant diseases, the detec-

© The Author(s), under exclusive license to Springer Nature Switzerland AG 2022 277
H. Alkadhi et al. (eds.), Spectral Imaging, Medical Radiology Diagnostic Imaging,
https://doi.org/10.1007/978-3-030-96285-2_18
278 M. Benndorf et al.

tion of incidental findings has increased corresponding Hounsfield units (HU) of the
substantially in the last decades. On the one hand, lesion measured in unenhanced CT (Korobkin
the adrenal glands are a common site for inciden- et al. 1996). This observation has led to the clas-
tally detected lesions. Estimates of the percent- sification of adrenal adenomas as either lipid-rich
age of patients with incidentally detected adrenal (defined in unenhanced CT as having mean
lesions range between 3% and 8% (Barrett et al. HU ≤ 10) or lipid-poor (mean HU > 10). Around
2009; Bovio et al. 2006; Mansmann et al. 2004; 70% of adrenal adenomas are of the lipid-rich
Mayo-Smith et  al. 2017). For example, Barrett type (Boland et al. 1998).
et  al. report that incidental adrenal lesions are The diagnosis of an (lipid-rich) adenoma can
detected in 3.8% of patients who undergo CT for be made with 71% sensitivity and 98% specific-
diagnosis in trauma (Barrett et  al. 2009). The ity by using a HU threshold of ≤10 to indicate
majority of incidental lesions of the adrenal adenoma, measured in unenhanced CT (Boland
glands proves to be hormonally inactive adenoma et al. 1998). The high specificity means that only
(Mansmann et al. 2004). On the other hand, the rarely the diagnosis adenoma based on this HU
adrenal glands are also a site that is predisposed threshold is wrong. With virtual non-contrast
to harbor metastasis in a variety of malignancies. images derived from contrast enhanced spectral
In 9% of living patients with non-small cell lung CT acquisitions (VNC-CT), there is the opportu-
cancer adrenal metastases are evident, with num- nity to reliably classify a proportion of otherwise
bers rising to 25–39% in autopsy series indeterminate adrenal lesions as (lipid-rich)
(Bazhenova et al. 2014). adenomas.
Because lesions of the adrenal glands are In a systematic review and meta-analysis of
common, accurate differentiation of benign and the diagnostic accuracy of VNC-CT for adrenal
malignant lesions is crucial. Different imaging adenoma, Connolly et al. report a pooled sensi-
modalities can be employed to classify adrenal tivity of 54% for (lipid-rich) adenoma (Connolly
lesions—with dedicated conventional CT offer- et al. 2017). They do not derive a pooled estimate
ing a reliable diagnostic tool with unenhanced for specificity, since no false positive findings
images and washout calculation in contrast occur in the included five studies (Ho et al. 2012;
enhanced images (Johnson et al. 2009a). In this Kim et  al. 2013; Botsikas et  al. 2014; Gnannt
chapter, we provide an overview of the diagnostic et al. 2012; Helck et al. 2014), all of which use a
accuracy of spectral CT compared with conven- dual-source dual-energy technique for image
tional dedicated CT and MRI for the diagnosis of acquisition of spectral CT.  Notably, Connolly
adrenal lesions (mainly, to distinguish adenomas et  al. report a pooled sensitivity of 57% for
from metastases). We highlight the possibility to ­unenhanced CT (Connolly et  al. 2017), being
avoid further work-up of incidental adrenal lower than the oftentimes quoted 71% (Boland
lesions in a proportion of cases when spectral et al. 1998). Connolly et al. reason that selection
imaging is implemented as standard of care in bias might be responsible for this deviation
diagnostic, contrast enhanced CT examinations. (Connolly et al. 2017). All of the included studies
employ a threshold of 10 HU on VNC-CT for the
diagnosis of adenoma. Figure  1 provides an
2 Lipid-Rich Adenoma example for the potential of VNC-CT to diagnose
and Virtual Non-contrast lipid-rich adenomas.
Imaging VNC-CT HU measurements of adrenal lesions
tend to systematically result in higher HU values
Adenomas originate from the cortex of the adre- compared to measurements in unenhanced CT
nal gland. The cells of adrenal adenomas often (Kim et al. 2013; Botsikas et al. 2014; Nagayama
contain a considerable amount of intracytoplas- et al. 2020). Kim et al. report an average of 11.7
matic fat (Korobkin et al. 1996). The higher the HU of lipid-rich adenomas when measured in
amount of intracytoplasmatic fat, the lower the VNC-CT derived from early phase contrast
Spectral Computed Tomography Imaging of the Adrenal Glands 279

a b

c d

Fig. 1  An 80-year-old female patient with left buccal adenoma (b). One month later, liver MRI is performed for
Merkel-cell carcinoma is referred for staging CT. In the evaluation of an unclear focal liver lesion (not shown). In
dual-energy, portal venous phase scan of the abdomen, a this examination, chemical shift MRI demonstrates a
2 cm lesion is observed in the right adrenal gland (a), with marked signal drop in opposed-phase images (c) com-
mean HU of 83. In the derived VNC-CT the lesion dem- pared to in-phase images (d), confirming the VNC-CT
onstrates HU ≤ 10, prompting the diagnosis of a lipid-rich diagnosis of a lipid-rich adenoma

enhanced images, compared to 0.7 HU in unen- contrast phases are taken for VNC-CT calcula-
hanced CT (Kim et  al. 2013). Botsikas et  al. tion compared to earlier contrast phases (Kim
report a mean difference of 4 HU for all lesions et al. 2013; Botsikas et al. 2014). By using a dual-­
analyzed (Botsikas et  al. 2014). Small, statisti- layer CT detector for acquisition of spectral CT,
cally non-significant differences in HU measure- Nagayama et al. report a mean difference of 11
ment are reported by Ho et al. and Gnannt et al. HU between measurements of lipid-rich adeno-
(difference of 1.8 HU and 1.1 HU, respectively) mas in VNC-CT and unenhanced CT (Nagayama
(Ho et al. 2012; Gnannt et al. 2012). There is a et al. 2020). This phenomenon is not restricted to
trend towards smaller differences when delayed the adrenal glands, but pertains to a variety of
280 M. Benndorf et al.

abdominal organs (Durieux et al. 2018). Notably, mas and this difference not necessarily extending
the same systematic difference of HU values to metastases) there is no generally agreed upon
between measurements in VNC-CT and unen- adapted HU threshold for measurements in
hanced CT does not automatically extend to adre- VNC-CT at which to consider a lesion a lipid-­
nal metastases (Nagayama et al. 2020). Figure 2 rich adenoma. With the threshold of ≤10 HU in
provides an example of different HU obtained in VNC-CT, a smaller proportion of cases can be
VNC-CT and unenhanced CT. classified as lipid-poor adenoma compared to
For these reasons (systematic difference assessment with unenhanced CT. Given that the
between VNC-CT and unenhanced CT of adeno- HU difference is generally positive between

a b

Fig. 2  A 70-year-old male patient with non-small cell age of 8 HU (b). In a true unenhanced CT acquired earlier
lung cancer referred for staging. An adrenal lesion of the lesion measures 4 HU (c). In this lipid-rich adenoma,
1.9 × 1.2 cm is observed in the left adrenal gland, with 64 the tendency of spectral CT to overestimate HU of lipid-­
HU in the dual-energy, portal venous phase scan (a). In rich adenoma in VNC-CT is demonstrated
the derived VNC-CT, the lesion is measured with an aver-
Spectral Computed Tomography Imaging of the Adrenal Glands 281

VNC-CT and unenhanced CT, specificity should Diagnostic accuracy of washout analysis for the
not be impaired, and the common ≤10 HU diagnosis of adenoma is generally very good
threshold for diagnosis of adenoma can be (Johnson et al. 2009b); in a summary by Johnson
applied to VNC-CT (Mayo-Smith et al. 2017). et al. 6/7 studies report a sensitivity between 96%
and 100% and 6/7 studies report a specificity
between 95% and 100% (Johnson et al. 2009b).
3 Lipid-Poor Adenoma, The technique is considered the gold standard
Washout Analysis, method of evaluation for adrenal lesions by some
and Chemical Shift MRI authors (Albano et  al. 2019)—since higher HU
values of adenomas on unenhanced scans do not
Around 30% of adenomas belong to the lipid-­ affect the high sensitivity of washout analysis as
poor type (compare for above). This means they compared to csMRI (Seo et al. 2014; Warda et al.
cannot reliably be diagnosed with unenhanced, or 2016).
VNC, CT (HU in unenhanced CT  >  10). There A possible application of spectral CT in this
are two commonly employed imaging techniques setting is the replacement of the unenhanced scan
that can further classify this type of lesions. by VNC-CT for absolute percent washout calcu-
Firstly, chemical shift MRI (csMRI) is able to lation. Kim et al. study the performance of abso-
quantify the amount of intracytoplasmatic fat lute washout percentage using VNC-CT as
(Adam et al. 2016). There is a characteristic drop unenhanced scan and report a sensitivity of 100%
in signal intensity in opposed-phase images com- and a specificity of 87.5% (Kim et  al. 2013).
pared to in-phase images (Adam et al. 2016) in a Washout analysis that employs the unenhanced
large proportion of adrenal adenomas, compare scan has a reported sensitivity of 100% and a
to Fig.  1, c and d. Several metrics can quantify specificity of 93.8% in this study (Kim et  al.
the degree of signal loss: the Adrenal Signal 2013). Botsikas et al. report that washout calcu-
Intensity Index (ASII) uses signal intensities lated with VNC-CT is higher when VNC-CT is
from in-phase and opposed-phase images only, derived from early (portal venous) scans com-
whereas other metrics normalize the signal drop pared to VNC-CT derived from delayed scans
to the signal of liver, spleen, and skeletal muscle (Botsikas et  al. 2014). At this stage, additional
(Ream et al. 2015). csMRI is highly sensitive and studies seem warranted to investigate this
specific for the diagnosis for adrenal adenoma— approach.
a systematic review and meta-analysis report a
pooled sensitivity of 94% and a specificity of
95% (Platzek et al. 2019). However, there is evi- 4 Quantification of Iodine
dence that the higher the HU of the lesion in
unenhanced CT, the worse the sensitivity of Adrenal adenomas typically enhance avidly in
csMRI (Haider et  al. 2004). Therefore, csMRI early (60–90s) phases after intravenous contrast
cannot be generally recommended as a work-up agent injection, and there is evidence that abso-
test in lesions >30 HU in unenhanced CT (Adam lute contrast enhancement (defined as difference
et al. 2016; Haider et al. 2004). in HU between enhanced and unenhanced scans)
Secondly, contrast enhanced CT with washout is higher in early phases in adenomas than in
calculation is routinely performed to further clas- non-adenomas (Korobkin et al. 1996; Szolar and
sify indeterminate adrenal lesions. For washout Kammerhuber 1998; Peña et al. 2000; Foti et al.
evaluation, an unenhanced, an early phase (60– 2010). However, the absolute HU value in
90s) and a delayed phase (10 min or 15 min) scan enhanced early phases is not a meaningful diag-
are acquired. The absolute washout percentage is nostic variable, since here HU values are quite
derived from all three phases, whereas for rela- similar for adenomas and non-adenomas
tive washout percentage early phase and delayed (Korobkin et al. 1996; Szolar and Kammerhuber
phase images suffice (Johnson et  al. 2009b). 1998; Peña et al. 2000; Foti et al. 2010). Spectral
282 M. Benndorf et al.

CT offers the possibility of material decomposi- noma (Nagayama et al. 2020)—being inferior to
tion (Patino et al. 2016) and thus to directly mea- the performance of the unenhanced scan. This is
sure the high absolute contrast enhancement of improved to 95% sensitivity and 95% specificity
adenomas in early phase images by quantifying when normalization to VNC-CT HU is performed
the corresponding iodine concentration. (iodine-density to VNC ratio) (Nagayama et  al.
For iodine-density (concentration, given in 2020), compare to Fig. 3. Confirmatory studies of
mg/ml), Nagayama et  al. report a sensitivity of this promising finding are pending. In accordance
78% and a specificity of 71% for adrenal ade- with the presented literature that adrenal adeno-

a b

c d

Fig. 3  A 59-year-old female patient with non-small cell iodine-density to VNC ratio of 5.4. This is suggestive of a
lung cancer is referred for staging CT.  A lesion of metastasis, according to the article by Nagayama et  al.
2.2 × 1.3 cm is observed in the right adrenal gland. The (2020). The lesion demonstrates rapid growth in a follow-
lesion measures 98 HU in the portal venous dual-energy ­up examination performed three months later and there-
CT (a) and 39 in the derived VNC image (b). The iodine- fore is considered a metastasis (d)
density of the lesion is 2.1 mg/ml (c), resulting in an
Spectral Computed Tomography Imaging of the Adrenal Glands 283

mas show higher absolute contrast enhancement nal gland: principles, pitfalls, and applica-
compared to metastases in early phase images tions. RadioGraphics 36:414–432. https://doi.
org/10.1148/rg.2016150139
(Korobkin et al. 1996; Szolar and Kammerhuber Albano D, Agnello F, Midiri F, Pecoraro G, Bruno A,
1998; Peña et al. 2000; Foti et al. 2010), adeno- Alongi P, Toia P, Di Buono G, Agrusa A, Sconfienza
mas tend to have higher iodine-density in this LM, Pardo S, La Grutta L, Midiri M, Galia M (2019)
study (Nagayama et al. 2020). The opposite rela- Imaging features of adrenal masses. Insights Imaging
10:1. https://doi.org/10.1186/s13244-­019-­0688-­8
tionship is reported by Martin et  al. and Mileto Barrett TW, Schierling M, Zhou C, Colfax JD, Russ S,
et  al., here adenomas demonstrate significantly Conatser P, Lancaster P, Wrenn K (2009) Prevalence of
lower iodine-density compared to non-adenomas incidental findings in trauma patients detected by com-
(Martin et  al. 2018; Mileto et  al. 2014). At the puted tomography imaging. Am J Emerg Med 27:428–
435. https://doi.org/10.1016/j.ajem.2008.03.025
moment, there is not enough evidence to recom- Bazhenova L, Newton P, Mason J, Bethel K, Nieva J,
mend a diagnostic threshold of iodine-density to Kuhn P (2014) Adrenal metastases in lung cancer:
accurately classify adrenal lesions as adenoma or clinical implications of a mathematical model. J
non-adenoma in clinical routine. Thorac Oncol 9:442–446. https://doi.org/10.1097/
JTO.0000000000000133
Boland GW, Lee MJ, Gazelle GS, Halpern EF, McNicholas
MM, Mueller PR (1998) Characterization of adrenal
masses using unenhanced CT: an analysis of the CT
5 Summary literature. AJR Am J Roentgenol 171:201–204. https://
doi.org/10.2214/ajr.171.1.9648789
Spectral (contrast enhanced) CT offers additional Botsikas D, Triponez F, Boudabbous S, Hansen C,
Becker CD, Montet X (2014) Incidental adrenal
diagnostic information for the evaluation of adrenal
lesions detected on enhanced abdominal dual-energy
lesions. VNC-CT images can be used as a substi- CT: can the diagnostic workup be shortened by the
tute for unenhanced images to confidently diag- implementation of virtual unenhanced images? Eur
nose lipid-rich adenoma with a threshold of ≤10 J Radiol 83:1746–1751. https://doi.org/10.1016/j.
ejrad.2014.06.017
HU, making additional unenhanced scans or work-
Bovio S, Cataldi A, Reimondo G, Sperone P, Novello S,
up unnecessary in these cases. However, sensitivity Berruti A, Borasio P, Fava C, Dogliotti L, Scagliotti
for lipid-rich adenoma is lower compared to unen- GV, Angeli A, Terzolo M (2006) Prevalence of adre-
hanced scans, since VNC-CT images tend to sys- nal incidentaloma in a contemporary computerized
tomography series. J Endocrinol Investig 29:298–302.
tematically overestimate HU of lipid-rich
https://doi.org/10.1007/BF03344099
adenomas. The American College of Radiology’s Connolly MJ, McInnes MDF, El-Khodary M, McGrath
white paper on the management of incidental adre- TA, Schieda N (2017) Diagnostic accuracy of vir-
nal lesions explicitly mentions dual-energy CT and tual non-contrast enhanced dual-energy CT for diag-
nosis of adrenal adenoma: a systematic review and
the potential usage of VNC-CT with a threshold of
meta-analysis. Eur Radiol 27:4324–4335. https://doi.
≤10 HU (Mayo-­Smith et al. 2017) for diagnosis. org/10.1007/s00330-­017-­4785-­0
Replacement of the true unenhanced scan for Durieux P, Gevenois PA, Muylem AV, Howarth N, Keyzer
washout calculation with VNC-CT and iodine-den- C (2018) Abdominal attenuation values on virtual
and true unenhanced images obtained with third-­
sity is not explicitly mentioned.
generation dual-source dual-energy CT.  AJR Am J
Roentgenol 210:1042–1058. https://doi.org/10.2214/
Compliance with Ethical Standards AJR.17.18248
Foti G, Faccioli N, Manfredi R, Mantovani W, Mucelli
RP (2010) Evaluation of relative wash-in ratio of adre-
Ethical Approval:  For this chapter no studies on humans
nal lesions at early biphasic CT.  Am J Roentgenol
or animals were performed.
194:1484–1491. https://doi.org/10.2214/AJR.09.3636
Gnannt R, Fischer M, Goetti R, Karlo C, Leschka S,
Alkadhi H (2012) Dual-energy CT for characteriza-
References tion of the incidental adrenal mass: preliminary obser-
vations. AJR Am J Roentgenol 198:138–144. https://
doi.org/10.2214/AJR.11.6957
Adam SZ, Nikolaidis P, Horowitz JM, Gabriel H,
Haider MA, Ghai S, Jhaveri K, Lockwood G (2004)
Hammond NA, Patel T, Yaghmai V, Miller FH
Chemical shift MR imaging of hyperattenuating
(2016) Chemical shift MR imaging of the adre-
(>10 HU) adrenal masses: does it still have a role?
284 M. Benndorf et al.

Radiology 231:711–716. https://doi.org/10.1148/ Committee. J Am Coll Radiol 14:1038–1044. https://

radiol.2313030676 doi.org/10.1016/j.jacr.2017.05.001
Helck A, Hummel N, Meinel FG, Johnson T, Nikolaou Mileto A, Nelson RC, Marin D, Roy Choudhury K, Ho
K, Graser A (2014) Can single-phase dual-energy LM (2014) Dual-energy multidetector CT for the
CT reliably identify adrenal adenomas? Eur characterization of incidental adrenal nodules: diag-
Radiol 24:1636–1642. https://doi.org/10.1007/ nostic performance of contrast-enhanced material
s00330-­014-­3192-­z density analysis. Radiology 274:445–454. https://doi.
Ho LM, Marin D, Neville AM, Barnhart HX, Gupta RT, org/10.1148/radiol.14140876
Paulson EK, Boll DT (2012) Characterization of adre- Nagayama Y, Inoue T, Oda S, Tanoue S, Nakaura T,
nal nodules with dual-energy CT: can virtual unen- Ikeda O, Yamashita Y (2020) Adrenal adenomas
hanced attenuation values replace true unenhanced versus metastases: diagnostic performance of dual-­
attenuation values? AJR Am J Roentgenol 198:840– energy spectral CT virtual noncontrast imaging and
845. https://doi.org/10.2214/AJR.11.7316 iodine maps. Radiology. https://doi.org/10.1148/
Johnson PT, Horton KM, Fishman EK (2009a) Adrenal radiol.2020192227
mass imaging with multidetector CT: pathologic con- Patino M, Prochowski A, Agrawal MD, Simeone FJ,
ditions, pearls, and pitfalls. RadioGraphics 29:1333– Gupta R, Hahn PF, Sahani DV (2016) Material sepa-
1351. https://doi.org/10.1148/rg.295095027 ration using dual-energy CT: current and emerging
Johnson PT, Horton KM, Fishman EK (2009b) Adrenal applications. RadioGraphics 36:1087–1105. https://
imaging with multidetector CT: evidence-based doi.org/10.1148/rg.2016150220
protocol optimization and interpretative prac- Peña CS, Boland GW, Hahn PF, Lee MJ, Mueller PR
tice. RadioGraphics 29:1319–1331. https://doi. (2000) Characterization of indeterminate (lipid-
org/10.1148/rg.295095026 poor) adrenal masses: use of washout characteristics
Kim YK, Park BK, Kim CK, Park SY (2013) Adenoma at contrast-­enhanced CT.  Radiology 217:798–802.
characterization: adrenal protocol with dual-energy https://doi.org/10.1148/radiology.217.3.r00dc29798
CT. Radiology 267:155–163. https://doi.org/10.1148/ Platzek I, Sieron D, Plodeck V, Borkowetz A, Laniado
radiol.12112735 M, Hoffmann R-T (2019) Chemical shift imaging for
Korobkin M, Giordano TJ, Brodeur FJ, Francis IR, evaluation of adrenal masses: a systematic review and
Siegelman ES, Quint LE, Dunnick NR, Heiken JP, meta-analysis. Eur Radiol 29:806–817. https://doi.
Wang HH (1996) Adrenal adenomas: relationship org/10.1007/s00330-­018-­5626-­5
between histologic lipid and CT and MR findings. Ream JM, Gaing B, Mussi TC, Rosenkrantz AB (2015)
Radiology 200:743–747. https://doi.org/10.1148/ Characterization of adrenal lesions at chemical-shift
radiology.200.3.8756925 MRI: a direct intraindividual comparison of in- and
Mansmann G, Lau J, Balk E, Rothberg M, Miyachi opposed-phase imaging at 1.5 T and 3 T. AJR Am J
Y, Bornstein SR (2004) The clinically inapparent Roentgenol 204:536–541. https://doi.org/10.2214/
adrenal mass: update in diagnosis and management. AJR.14.12941
Endocr Rev 25:309–340. https://doi.org/10.1210/ Seo JM, Park BK, Park SY, Kim CK (2014)
er.2002-­0031 Characterization of lipid-poor adrenal adenoma:
Martin SS, Weidinger S, Czwikla R, Kaltenbach B, chemical-shift MRI and washout CT.  AJR Am J
Albrecht MH, Lenga L, Vogl TJ, Wichmann JL Roentgenol 202:1043–1050. https://doi.org/10.2214/
(2018) Iodine and fat quantification for differentia- AJR.13.11389
tion of adrenal gland adenomas from metastases using Szolar DH, Kammerhuber FH (1998) Adrenal adenomas
third-generation dual-source dual-energy computed and nonadenomas: assessment of washout at delayed
tomography. Investig Radiol 53:173–178. https://doi. contrast-enhanced CT.  Radiology 207:369–375.
org/10.1097/RLI.0000000000000425 https://doi.org/10.1148/radiology.207.2.9577483
Mayo-Smith WW, Song JH, Boland GL, Francis IR, Warda MHA, Shehata SM, Zaiton F (2016) Chemical-­
Israel GM, Mazzaglia PJ, Berland LL, Pandharipande shift MRI versus washout CT for characterizing adre-
PV (2017) Management of incidental adrenal masses: nal incidentalomas. Clin Imaging 40:780–787. https://
a white paper of the ACR Incidental Findings doi.org/10.1016/j.clinimag.2016.02.006
 Urogenital Imaging: Kidneys
(Lesion Characterization)

Ali Pourvaziri, Anushri Parakh,
Avinash Kambadakone, and Dushyant Sahani

Contents
1 Virtual Monoenergetic/Monochromatic Imaging (VMI)   286
2 Material Specific Imaging   287
3 Virtual Unenhanced Imaging (VUE)   287
4 Iodine Imaging   288
5 Radiation Dose Consideration   290
6 Conclusion   291
References   291

Renal lesion detection and characterization is an the past decade (Chow et al. 1999). Continuous
evolving conundrum with significant economic increase in renal lesion incidence can be
and healthcare ramifications. Up to 40% of explained, at least partially, by growth of cross-­
patients have at least one incidental renal lesion imaging utilization and higher accuracy of
discovered on imaging done for another reason modalities (Brenner 2010, 2009; Brenner and
(Carrim and Murchison 2003). More than 60% of Hall 2007). Technological improvement, such as
renal cell carcinomas (RCC) are now incidentally multidetector imaging and dose modulation, has
discovered in asymptomatic individuals (Carrim steadily increased the clinical application of CT
and Murchison 2003; Jayson and Sanders 1998; imaging (Brenner and Hall 2007; Frush and
Pandharipande et  al. 2010). The incidence of Applegate 2004).
renal lesions has been rising continuously over In most centers, a multiphasic contrast-­
enhanced CT scan is the preferred next step in
characterizing incidentally found renal lesions.
A. Pourvaziri · A. Parakh · A. Kambadakone
Department of Radiology, Massachusetts General Conventional renal lesion assessment requires at
Hospital, Harvard Medical School, least two phases (non-contrast and contrast scan)
Boston, MA, USA for evaluation of any underlying enhancement,
D. Sahani (*) which is fundamental for classification methods
Department of Radiology, University of Washington such as the Bosniak system. However, conven-
School of Medicine, Seattle, WA, USA tional CT imaging is not without limitations; (a)
e-mail: [email protected]

© The Author(s), under exclusive license to Springer Nature Switzerland AG 2022 285
H. Alkadhi et al. (eds.), Spectral Imaging, Medical Radiology Diagnostic Imaging,
https://doi.org/10.1007/978-3-030-96285-2_19
286 A. Pourvaziri et al.

The Hounsfield value measurements are subjected platforms (rsDECT), and image domain is imple-
to variation based on kV utilization (Afifi et  al. mented in dual-layer (dlDECT) and dual-source
2020). (b) The inherit motion artifact that stems DECT (dsDECT) platforms. VMI reconstruc-
from breathing may lead to difficulty in assessing tions, as the name implies, are gray-scale images
the pixel by pixel-wise comparison of unenhanced that resemble the appearance of what would be
and contrast-enhanced images. This is signifi- otherwise acquired using single energy photons.
cantly more cumbersome in small foci of enhance- VMI acquisition is more thoroughly discussed in
ment or lesions. (c) Additionally, other chapters. However, it is worth noting that
pseudoenhancement is an artifact described when by definition, monoenergetic imaging is unaf-
measuring the density of small renal lesions— fected by beam hardening artifacts (Albrecht
usually less than 1–2  cm. Pseudoenhancement et  al. 2019). Predictably, VMI utilization has
leads to a spuriously increased estimation of den- been shown to decrease pseudoenhancement arti-
sity. This artifact is more encountered in the set- fact in small renal cysts without the need for gen-
ting of prominent renal parenchymal enhancement eration of a true unenhanced acquisition (Jung
and is proposed to be due to beam hardening and et al. 2012; Mileto et al. 2014a).
partial volume averaging artifacts (Birnbaum There is a high reproducibility of attenuation
et  al. 2002; Coulam et  al. 2000; Tappouni et  al. measurements (Hounsfield Unit, HU) with only
2012). These shortcomings have led radiologists minor inconsistencies across different DECT
to explore solutions that characterize renal lesions technologies of VMI (Sellerer et  al. 2018;
in more accurate and efficient ways. In this chap- Jacobsen et al. 2017; Silva et al. 2011). The supe-
ter, we discuss how dual-energy CT scan imple- riority of image quality of VMIs in comparison to
mentation in clinical practice can address some of conventional 120-kVp images have been demon-
these shortcomings and expand on the utility of strated in several studies (Yu et  al. 2011, 2012;
various images generated from dual-energy scans. Matsumoto et  al. 2011). The energy range of
Since the introduction of dual-energy CT scan 60–70 keV corresponds to a midpoint between
(DECT) into routine clinical practice more than a the mean energies of 80 kVp and 140 kVp. Data
decade ago (Acharya et  al. 2015), significant suggest that as a general rule, range of 60–70 keV
hardware and image processing developments provides an optimal contrast-to-noise ratio
have enabled a myriad of clinically useful appli- (CNR) with the lowest amount of noise (Yu et al.
cations in genitourinary imaging. DECT provides 2011, 2012; Matsumoto et al. 2011). However, in
virtual monoenergetic imaging (VMI) and evaluating renal cell carcinoma in genitourinary
material-­specific information that is unobtainable imaging, 60 keV may provide better image qual-
using conventional single energy CT imaging. ity for nephrographic phase and 50 keV for corti-
Additionally, in DECT, two imaging datasets comedullary phase (Martin et al. 2017).
from two different energy is generated near-­ Lesion specific attenuation curve can be gen-
simultaneously, which considerably improves the erated using ranges of VMIs. The keV range is
prospect of a pixel by pixel-wise comparison from 40 to 140 keV in rapid switching (rsDECT),
with virtual unenhanced images (VUE) (Heye 40–190 keV in dual-source (dsDECT), and
et al. 2012; Heye 2012). 40–200 keV in dual-layer platforms (dlDECT).
Enhancing lesions can be differentiated from
cysts on a single-phase nephrographic images.
1 Virtual Monoenergetic/ Iodine attenuation significantly increases at lower
Monochromatic Imaging keV range, whereas cyst levels remain relatively
(VMI) similar and demonstrate a relatively flat curve
(Fig. 1). Additionally, there is potential for distin-
Depending on the DECT system utilized, VMI is guishing malignant from benign lesions using
acquired in the data or image domain. Data quantitative spectral analysis of these curves
domain is used in rapid kVp switching DECT (Table 1) (Patel et al. 2017).
Urogenital Imaging: Kidneys (Lesion Characterization) 287

a b

Fig. 1 (a) Axial post-contrast portal phase images in a attenuation in lower keVs level, reflecting a non-enhanced
dual-layer DECT (dlDECT) platform demonstrates a cyst. (b) Adjacent normal parenchyma enhances brightly
hypodense lesion in the left kidney measuring 32.4 measuring 214 HU. Note the up-pick of the HU attenua-
HU.  The spectral curve indicates a minimal increase in tion plot in lower keVs (arrows) in comparison to the cyst

Table 1  Virtual monoenergetic imaging utility in renal 2011; Hartman et  al. 2012; Kaza et  al. 2017a,
lesion characterization 2012). A similar technique is used in two mate-
VM imaging in renal lesion characterization rial decomposition techniques. However, two
High reproducibility of HU measurement material decomposition techniques with two
Higher image quality markedly different atomic number or mass-­
Decrease pseudoenhancement artifact attenuation coefficient is used to generate two
Optimal CNR and lower noise
sets of images; material density images such as
Generation and analysis of attenuation curves
• Differentiation of enhancing lesions from cysts in
iodine or water images (virtual unenhanced
one phase images) (Fornaro et  al. 2011; Hartman et  al.
• Potential to distinguish malignant from a benign 2012; Kaza et al. 2017a, 2012, 2011).
lesion
CNR contrast to noise ratio, VM virtual monoenergetic
3  irtual Unenhanced Imaging
V
(VUE)
2 Material Specific Imaging
Virtual unenhanced or water-density images can
DECT generated data can be used to determine provide information similar to true non-contrast
the voxel compositions. Each voxel is analyzed images and may be used as a surrogate for true
on the basis of two-material decomposition prin- non-contrast images. Data generated from virtual
ciple for data domain technique and three-­ unenhanced images has been shown to provide a
material decomposition principle for image reliable assessment of non-enhanced characteris-
domain techniques. In three-material decomposi- tics of renal lesions such as the presence of fat,
tion approach, absorption characteristics of three hyperdense material or calcification (Fig. 3b, e)
materials, such as fat, iodine, and soft tissue at (Ascenti et al. 2012). Attenuation measurement is
two energy levels are idealized and used to create readily available in the latest generation of rapid
specific material image series. The iodine can be kVp switching DECT (rsDECT) technology in
extracted to create virtual unenhanced images. addition to dual-source (dsDECT) and dual-layer
On the other hand, iodine can be overlaid in dif- systems (dlDECT). Most studies point to excel-
ferent ratios to create iodine overlay images or lent correlation between HU measurement of vir-
purely used to create iodine maps (Fornaro et al. tual non-enhanced images and true non-contrast
288 A. Pourvaziri et al.

Table 2  Virtual unenhanced imaging in renal lesion readily detect lesions enhancement in one
Virtual unenhanced imaging in renal lesion nephrographic phase (Fig.  2). In fact, studies
characterization have shown qualitative assessment of iodine
Provide information similar to true non-contrast allows more confident and faster characteriza-
Reliable assessment for the presence of fat,
hyperdense material or calcification
tions of smaller lesions (Pourvaziri et al. 2019). It
Excellent correlation of HU measurement with true enables a decrease in readers’ interpretation time
non-contrast exam and simplifies the workflow of renal CT proto-
Limitations cols (Kaza et  al. 2011; Ascenti et  al. 2012;
• More than 10 HU variability in measurement of Pourvaziri et al. 2019; Mileto et al. 2014b, 2015;
some lesions.
Marin et al. 2014; Graser et al. 2010). This par-
• Possible suppression of signal of punctate renal
stones (less than 3 mm). ticularly is useful in evaluating multiple lesions,
• Lack of HU measurement capability in an older such as in patients with polycystic kidney dis-
generation of rapid switching. ease, where it increases readers’ confidence and
decreases readers’ time (Glomski et  al. 2018;
Arndt et al. 2012).
images (Slebocki et al. 2017; Meyer et al. 2019; Quantitative assessment of iodine concentra-
Kaza et al. 2017b; Borhani et al. 2017). However, tion (IC) is measured based on calculation of the
HU measurement on virtual unenhanced imaging iodine concentration in milligrams per millime-
cannot reliably be used for follow-up CT, if there ter (mg/mL) (Coursey et al. 2010). Quantitative
are two different dual-energy CT platforms used assessment of iodine provides accurate and
(Obmann et al. 2019). rapid determination of enhancement without the
There are limitations to the usage of virtual need for HU measurement on both unenhanced
non-enhanced images. First, more than 10 HU of and contrast-enhanced measurement. In evaluat-
variation in attenuation compared to a true non-­ ing renal lesions with homogenous attenuation,
contrast exam has been reported (Kaza et  al. placement of only one ROI in each phase is suf-
2017b; Borhani et  al. 2017; Çamlıdağ 2020). ficient for adequate HU measurement. However,
Further development of post-processing algo- in lesions with cystic, necrotic, or complex
rithms might be needed for the complete replace- structures, multiple matching ROIs placed in
ment of true non-enhanced images with VUE both nonenhanced and contrast phase are
images. Second, due to excessive iodine suppres- required. This approach could potentially be
sion, signal of small stones with usually less than biased by inclusion of small areas of necrosis or
3 mm, could be suppressed in virtual non-­ cystic change. Whereas, measurement of iodine
enhanced images, and these stones could be concentration is the sum of all the enhancing
missed on these images (Table  2) (Takahashi areas only, and is less likely affected by the
et al. 2010; Vernuccio et al. 2018). inclusion of cystic or necrotic areas within the
ROIs placement (Ascenti et  al. 2012, 2013;
Mileto et  al. 2014c). Iodine concentration of
4 Iodine Imaging more than 0.5 mg/ml on dual-source DECT
(dsDECT) and 1 mg/ml on rapid switching plat-
Dual-energy CT data allows for qualitative and forms (rsDET) has been show to optimal thresh-
quantitative assessment of iodine uptake of old for depiction of internal enhancement
tissues. (Pourvaziri et al. 2019; Leng et al. 2015; Mileto
In the qualitative assessment of iodine imag- et  al. 2014d; Chandarana et  al. 2011). Iodine
ing, renal cysts and avascular lesions are promptly quantification not only provides means to quan-
diagnosed with lack of intralesional signal, tify enhancement without the need for true non-
whereas iodine signal within lesions could poten- contrast exam, it also provides higher accuracy
tially be interpreted as an intralesional enhance- than HU measurement (Pourvaziri et  al. 2019;
ment. Thus, iodine qualitative assessment can Ascenti et al. 2013).
Urogenital Imaging: Kidneys (Lesion Characterization) 289

a b

c d

Fig. 2 59-year-old male with incidental enhancing post-contrast nephrographic phase on dual-layer platform
lesions 5 months apart. (a, b) Axial post-contrast nephro- 5 months later shows an interval increase in the size of
graphic phase on dual-source platform (dsDECT) demon- previously noted enhancing lesion. Note that enhancing
strates an enhancing lesion in the left kidney. (c, d) Axial lesion is better demonstrated on Iodine imaging (a, c)

Another exciting feature of iodine quantifica- 2004; Vandenbroucke et al. 2015). Iodine content
tion is the potential of differentiating clear cell in ablations zones can be assessed both qualita-
carcinoma from other enhancing lesions tively and quantitatively on iodine imaging.
(Pourvaziri et  al. 2019; Mileto et  al. 2014d). In Regions of intermediate enhancement surround-
dual-source (dsDECT) approach, an iodine con- ing the ablation zones could be noted and are bet-
centration threshold of 0.9 mg/ml has been shown ter appreciated on iodine imaging (Vandenbroucke
to distinguish papillary from clear cell renal cell et al. 2015). However, ring enhancement, such as
carcinoma with sensitivity and specificity of 98% seen in hepatic lesions, is not common in post-­
and 86%, respectively (Mileto et  al. 2014d). In ablation of renal lesions (Vandenbroucke et  al.
rapid switching platforms (rsDECT), a threshold 2015). Iodine quantification has been able to dif-
of 2.1 mg/ml distinguished clear cell from other ferentiate reactive tissue from residual tumor
enhancing lesions with a sensitivity of 92% after post-radiofrequency ablative treatment
(Pourvaziri et al. 2019). (Fig. 3) (Li et al. 2013). Dual-energy iodine and
Another utility of iodine images is assessing virtual non-enhanced images have been able to
tumor response and predicting tumor progression predict renal tumors’ progression after ablation
after ablative therapy. Ablation zones are wedge with 100% sensitivity and 91.5% specificity
shape and lack any iodine content (Tan et  al. (Park et al. 2014).
290 A. Pourvaziri et al.

a b c

d e f

Fig. 3  69-year-old male with chronic kidney disease on wave ablative therapy. The lower set of images is 1-month
hemodialysis was found to have an enhancing 2.6  cm post-ablation and 5 months after images in the upper row.
lesion (white arrow in all images) in the lower pole of the Renal CT protocol with rapid switching DECT demon-
right kidney. Renal CT protocol with rapid switching strates. (d) Axial post-contrast 65 keV monochromatic
DECT; (a) Axial post-contrast 65 keV monochromatic images shows a mixed-density 2.5  cm with surrounding
images demonstrated an enhancing lesion measuring 86 postprocedural stranding (triangle). (e) Axial material-­
HU. (b) Axial virtual unenhanced images revealed an iso-­ density water image shows hyperdense material in the
dense lesion measuring 30 HU in density. Note the hyper- region of the previously described tumor, consistent with
dense structures IVC filter, calcified plaque in the adjacent hemorrhage. (f) Material-density iodine image demon-
aorta, and small punctate stone in the right kidney (trian- strates a lack of internal iodine signal. Iodine concentra-
gles) (c) Axial material-density iodine images; iodine tion measured 0.08 mgr/ml, consistent with lack of
concentration measured 1.9 mgr/ml consistent with enhancement and no residual tumor
enhancing lesion. The patient underwent interval micro-

Quantitative assessment of iodine concentra- centration measurement shows significantly

tion has been used in assessing response to treat- more relative reduction than measuring the
ment in metastatic renal cell carcinomas. Iodine Hounsfield unit alone (49.8% vs. 29.5%)
concentration is significantly more sensitive and (Table 3) (Hellbach et al. 2017).
reproducible in determining response to anti-­
angiogenic therapy (Hellbach et al. 2017). Anti-­
angiogenic treatments decrease tumor angiogenic 5 Radiation Dose
activity and perfusion without considerable volu- Consideration
metric changes. In assessing the response to
treatment in metastatic RCC, criteria that use Historically, one of the hurdles of dual-energy
changes in attenuation and morphology rather CT scan utilization was the concern for a hypo-
than commonly used response evaluation criteria thetical increase in radiation dose burden to
in solid tumors or Choi criteria have been shown patients. This perpetual increase stems from gen-
to be more accurate and reproducible (Smith erating two datasets, instead of one, from two
et al. 2010). After a successful treatment of meta- separate tubes with different voltages. First and
static RCC, there is a significant reduction in the foremost, in comparing two CT protocols, it is
enhancement of lesions. However, iodine con- imperative to take into account a potential vari-
Urogenital Imaging: Kidneys (Lesion Characterization) 291

Table 3  Iodine imaging utility in renal lesion formed on single energy CT scan. Although
characterization
Iterative reconstruction techniques have been
Iodine imaging in renal lesion characterization implemented successfully to reduce dose reduc-
Qualitative assessment tion in dual-energy CT scans (Ohana et al. 2015;
• Internal enhancement evaluation in single phase
Zhao et al. 2017), more studies are needed to fur-
•  Faster interpretation time
ther explore iterative reconstruction techniques in
•  Higher readers confidence
Quantitative assessment this regard. The latest generation of reconstruc-
• Fast measurement of internal enhancement in a tion techniques implements deep learning tech-
single phase nology (Jensen et  al. 2020; Kim et  al. 2021).
• More accurate than HU measurement Implementation of deep learning-based image
• Possible differentiation of clear cell RCCs from reconstruction has been shown to provide higher
other enhancing lesions
CNR and lower noise in abdominal CT scans
Possible better prediction of tumor response and
progression after ablative therapy
(Jensen et  al. 2020). Although more studies are
More sensitive and reproducible assessment of needed, deep learning reconstructions technol-
response to anti-angiogenic therapy ogy offers new horizons for further dose reduc-
tion (Greffier et al. 2020; Racine et al. 2020).

ability of radiation dose measurement, which can

even vary in the same patient. It has been demon- 6 Conclusion
strated that the mean coefficient for variance for
size-specific dose estimates for CTs of the abdo- In this chapter, we discussed how dual-energy CT
men and pelvis may vary with 10.26-­ mGy scan for genitourinary imaging addresses many
(Mileto et al. 2017). Second, mounted bodies of shortcomings of conventional CT imaging. We dis-
evidence have disapproved of such concerns cussed the utility of various dual-energy recon-
(Takeuchi et  al. 2012; Grajo and Sahani 2018; structions in a vendor-neutral fashion, how they
Shuman et  al. 2014; Lin et  al. 2012; Dubourg compared with conventional imaging. In summary,
et  al. 2014). Optimization of DECT protocols dual-energy imaging in the assessment of renal
and advancement in radiation dose Modulation lesions provides a faster, more confident, more
have decreased radiation dosage to the level that accurate, and more reproducible evaluation of renal
is comparable or even lower than that of SECT lesions without an increase in radiation exposure.
protocols (Takeuchi et al. 2012; Grajo and Sahani
2018; Shuman et  al. 2014; Lin et  al. 2012; Compliance with Ethical Standards
Dubourg et  al. 2014). Currently, many institu-
tions perform renal protocol dual-energy CT scan Disclosure of Interests None.
with radiation dose range at the level or below of
16 mGy, which is the achievable dose recom- Ethical Approval  This article does not contain any stud-
mended by American College of Radiology ies with human participants performed by any of the
authors. This article does not contain any studies with ani-
(ACR) (Ascenti et al. 2012; Takeuchi et al. 2012; mals performed by any of the authors.
Grajo and Sahani 2018; Kanal et  al. 2017;
Jepperson et al. 2015). Dose neutrality offers the
opportunity to perform all genitourinary imaging References
with dual-energy CT (Vernuccio et al. 2018).
Dual-energy CT scan provides several pros- Acharya S, Goyal A, Bhalla AS, Sharma R, Seth
pects of reducing radiation dosage. Virtual unen- A, Gupta AK (2015) In vivo characterization
hanced imaging might obviate the need for a true of urinary calculi on dual-energy CT: going a
step ahead with sub-­ differentiation of calcium
non-enhanced phase and decrease in radiation stones. Acta Radiol 56(7):881–889. https://doi.
dose up to 50% (Graser et  al. 2010, 2009). org/10.1177/0284185114538251
Iterative reconstruction technique has cut radia- Afifi MB, Abdelrazek A, Deiab NA, El-Hafez AIA,
tion dose up to 30% in abdominal CT scan per- El-Farrash AH (2020) The effects of CT X-ray tube
292 A. Pourvaziri et al.

voltage and current variations on the relative electron Coulam CH, Sheafor DH, Leder RA, Paulson EK, DeLong
density (RED) and CT number conversion curves. J DM, Nelson RC (2000) Evaluation of pseudoenhance-
Radiat Res Appl Sci 13(1):1–11. https://doi.org/10.10 ment of renal cysts during contrast-enhanced CT. Am J
80/16878507.2019.1693176 Roentgenol 174(2):493–498. https://doi.org/10.2214/
Albrecht MH, Vogl TJ, Martin SS et al (2019) Review of ajr.174.2.1740493
clinical applications for virtual monoenergetic dual-­ Coursey CA, Nelson RC, Boll DT et  al (2010) Dual-­
energy CT.  Radiology 293(2):260–271. https://doi. energy multidetector CT: how does it work, what can
org/10.1148/radiol.2019182297 it tell us, and when can we use it in abdominopelvic
Arndt N, Staehler M, Siegert S, Reiser MF, Graser A imaging? RadioGraphics 30(4):1037–1055. https://
(2012) Dual energy CT in patients with polycystic doi.org/10.1148/rg.304095175
kidney disease. Eur Radiol 22(10):2125–2129. https:// Dubourg B, Caudron J, Lestrat J-P et  al (2014) Single-­
doi.org/10.1007/s00330-­012-­2481-­7 source dual-energy CT angiography with reduced
Ascenti G, Mazziotti S, Mileto A et al (2012) Dual-source iodine load in patients referred for aortoiliofemoral
dual-energy CT evaluation of complex cystic renal evaluation before transcatheter aortic valve implanta-
masses. Am J Roentgenol 199(5):1026–1034. https:// tion: impact on image quality and radiation dose. Eur
doi.org/10.2214/AJR.11.7711 Radiol 24(11):2659–2668. https://doi.org/10.1007/
Ascenti G, Mileto A, Krauss B et al (2013) Distinguishing s00330-­014-­3263-­1
enhancing from nonenhancing renal masses with Fornaro J, Leschka S, Hibbeln D et al (2011) Dual- and
dual-source dual-energy CT: iodine quantification multi-energy CT: approach to functional imag-
versus standard enhancement measurements. Eur ing. Insights Imaging 2(2):149–159. https://doi.
Radiol 23(8):2288–2295. https://doi.org/10.1007/ org/10.1007/s13244-­010-­0057-­0
s00330-­013-­2811-­4 Frush DP, Applegate K (2004) Computed tomogra-
Birnbaum BA, Maki DD, Chakraborty DP, Jacobs JE, phy and radiation: understanding the issues. J Am
Babb JS (2002) Renal cyst pseudoenhancement: eval- Coll Radiol 1(2):113–119. https://doi.org/10.1016/j.
uation with an anthropomorphic body CT phantom. jacr.2003.11.012
Radiology 225(1):83–90. https://doi.org/10.1148/ Glomski SA, Wortman JR, Uyeda JW, Sodickson AD
radiol.2251010930 (2018) Dual energy CT for evaluation of polycystic
Borhani AA, Kulzer M, Iranpour N et  al (2017) kidneys: a multi reader study of interpretation time and
Comparison of true unenhanced and virtual unen- diagnostic confidence. Abdom Radiol 43(12):3418–
hanced (VUE) attenuation values in abdominopelvic 3424. https://doi.org/10.1007/s00261-­018-­1674-­1
single-source rapid kilovoltage-switching spec- Grajo JR, Sahani DV (2018) Dual-energy CT of the
tral CT.  Abdom Radiol 42(3):710–717. https://doi. abdomen and pelvis: radiation dose considerations.
org/10.1007/s00261-­016-­0991-­5 J Am Coll Radiol 15(8):1128–1132. https://doi.
Brenner DJ (2009) Slowing the increase in the collective org/10.1016/j.jacr.2017.08.012
dose resulting from CT scans. In: Dössel O, Schlegel Graser A, Johnson TRC, Hecht EM et  al (2009) Dual-­
WC (eds) World congress on medical physics and bio- energy CT in patients suspected of having renal
medical engineering. Springer, Berlin, pp  533–535. masses: can virtual nonenhanced images replace true
https://doi.org/10.1007/978-­3-­642-­03902-­7_152 nonenhanced images? Radiology 252(2):433–440.
Brenner DJ (2010) Should we be concerned about the rapid https://doi.org/10.1148/radiol.2522080557
increase in CT usage? Rev Environ Health 25(1):63. Graser A, Becker CR, Staehler M et al (2010) Single-­phase
https://doi.org/10.1515/REVEH.2010.25.1.63 dual-energy CT allows for characterization of renal
Brenner DJ, Hall EJ (2007) Computed tomography – an masses as benign or malignant. Investig Radiol 45(7):399.
increasing source of radiation exposure. N Engl J https://doi.org/10.1097/RLI.0b013e3181e33189
Med 357(22):2277–2284. https://doi.org/10.1056/ Greffier J, Hamard A, Pereira F et al (2020) Image quality
NEJMra072149 and dose reduction opportunity of deep learning image
Çamlıdağ İ (2020) Compatibility of true and virtual unen- reconstruction algorithm for CT: a phantom study. Eur
hanced attenuation in rapid kV-switching dual energy Radiol 30(7):3951–3959. https://doi.org/10.1007/
CT.  Diagn Interv Radiol 26(2):95–100. https://doi. s00330-­020-­06724-­w
org/10.5152/dir.2019.19345 Hartman R, Kawashima A, Takahashi N et  al (2012)
Carrim ZI, Murchison JT (2003) The prevalence of simple Applications of dual-energy CT in urologic imaging:
renal and hepatic cysts detected by spiral computed an update. Radiol Clin N Am 50(2):191–205. https://
tomography. Clin Radiol 58(8):626–629 doi.org/10.1016/j.rcl.2012.02.007
Chandarana H, Megibow AJ, Cohen BA et  al (2011) Hellbach K, Sterzik A, Sommer W et  al (2017) Dual
Iodine quantification with dual-energy CT: phantom energy CT allows for improved characterization of
study and preliminary experience with renal masses. response to antiangiogenic treatment in patients with
Am J Roentgenol 196(6):693–700. https://doi. metastatic renal cell cancer. Eur Radiol 27(6):2532–
org/10.2214/AJR.10.5541 2537. https://doi.org/10.1007/s00330-­016-­4597-­7
Chow W-H, Devesa SS, Warren JL, Joseph F, Fraumeni Heye T (2012) Dual-energy CT applications in the abdo-
J (1999) Rising incidence of renal cell cancer in the men. Am J Roentgenol 199(5):9196
United States. JAMA 281(17):1628–1631. https://doi. Heye T, Nelson RC, Ho LM, Marin D, Boll DT (2012)
org/10.1001/jama.281.17.1628 Dual-energy CT applications in the abdomen. Am J
Urogenital Imaging: Kidneys (Lesion Characterization) 293

Roentgenol 199(5):64–70. https://doi.org/10.2214/ energy CT.  Radiology 276(2):562–570. https://doi.

AJR.12.9196 org/10.1148/radiol.2015140857
Jacobsen MC, Schellingerhout D, Wood CA et al (2017) Li Y, Shi G, Wang S, Wang S, Wu R (2013) Iodine
Intermanufacturer comparison of dual-energy CT quantification with dual-energy CT: phantom study
iodine quantification and monochromatic attenuation: and preliminary experience with VX2 residual
a phantom study. Radiology 287(1):224–234. https:// tumour in rabbits after radiofrequency ablation.
doi.org/10.1148/radiol.2017170896 BJR 86(1029):20130143. https://doi.org/10.1259/
Jayson M, Sanders H (1998) Increased incidence of bjr.20130143
serendipitously discovered renal cell carcinoma. Lin XZ, Wu ZY, Tao R et  al (2012) Dual energy spec-
Urology 51(2):203–205. https://doi.org/10.1016/ tral CT imaging of insulinoma—value in preoperative
S0090-­4295(97)00506-­2 diagnosis compared with conventional multi-detector
Jensen CT, Liu X, Tamm EP et  al (2020) Image qual- CT.  Eur J Radiol 81(10):2487–2494. https://doi.
ity assessment of abdominal CT by use of new deep org/10.1016/j.ejrad.2011.10.028
learning image reconstruction: initial experience. Am Marin D, Boll DT, Mileto A, Nelson RC (2014)
J Roentgenol 215(1):50–57. https://doi.org/10.2214/ State of the art: dual-energy CT of the abdomen.
AJR.19.22332 Radiology 271(2):327–342. https://doi.org/10.1148/
Jepperson MA, Cernigliaro JG, Ibrahim E-SH, Morin RL, radiol.14131480
Haley WE, Thiel DD (2015) In vivo comparison of Martin SS, Wichmann JL, Pfeifer S et  al (2017) Impact
radiation exposure of dual-energy CT versus low-dose of noise-optimized virtual monoenergetic dual-energy
CT versus standard CT for imaging urinary calculi. computed tomography on image quality in patients
J Endourol 29(2):141–146. https://doi.org/10.1089/ with renal cell carcinoma. Eur J Radiol 97:1–7. https://
end.2014.0026 doi.org/10.1016/j.ejrad.2017.10.008
Jung DC, Oh YT, Kim MD, Park M (2012) Usefulness of Matsumoto K, Jinzaki M, Tanami Y, Ueno A, Yamada
the virtual monochromatic image in dual-energy spec- M, Kuribayashi S (2011) Virtual monochromatic
tral CT for decreasing renal cyst pseudoenhancement: spectral imaging with fast kilovoltage switch-
a phantom study. Am J Roentgenol 199(6):1316–1319. ing: improved image quality as compared with
https://doi.org/10.2214/AJR.12.8660 that obtained with conventional 120-kVp CT.
Kanal KM, Butler PF, Sengupta D, Bhargavan-Chatfield Radiology 259(1):257–262. https://doi.org/10.1148/
M, Coombs LP, Morin RL (2017) U.S. diagnostic radiol.11100978
reference levels and achievable doses for 10 adult CT Meyer M, Nelson RC, Vernuccio F et  al (2019) Virtual
examinations. Radiology 284(1):120–133. https://doi. unenhanced images at dual-energy CT: Influence on
org/10.1148/radiol.2017161911 renal lesion characterization. Radiology 291(2):381–
Kaza RK, Caoili EM, Cohan RH, Platt JF (2011) 390. https://doi.org/10.1148/radiol.2019181100
Distinguishing enhancing from nonenhancing renal Mileto A, Nelson RC, Samei E et  al (2014a) Impact of
lesions with fast kilovoltage-switching dual-energy dual-energy multi–detector row CT with virtual mono-
CT. Am J Roentgenol 197(6):1375–1381. https://doi. chromatic imaging on renal cyst pseudoenhancement:
org/10.2214/AJR.11.6812 in vitro and in vivo study. Radiology 272(3):767–776.
Kaza RK, Platt JF, Cohan RH, Caoili EM, Al-Hawary https://doi.org/10.1148/radiol.14132856
MM, Wasnik A (2012) Dual-energy CT with single- Mileto A, Marin D, Ramirez-Giraldo JC et  al (2014b)
and dual-source scanners: current applications in Accuracy of contrast-enhanced dual-energy MDCT
evaluating the genitourinary tract. RadioGraphics for the assessment of iodine uptake in renal lesions.
32(2):353–369. https://doi.org/10.1148/rg.322115065 Am J Roentgenol 202(5):466–474. https://doi.
Kaza RK, Ananthakrishnan L, Kambadakone A, Platt JF org/10.2214/AJR.13.11450
(2017a) Update of dual-energy CT applications in the Mileto A, Marin D, Nelson RC, Ascenti G, Boll DT
genitourinary tract. Am J Roentgenol 208(6):1185– (2014c) Dual energy MDCT assessment of renal
1192. https://doi.org/10.2214/AJR.16.17742 lesions: an overview. Eur Radiol 24(2):353–362.
Kaza RK, Raff EA, Davenport MS, Khalatbari S (2017b) https://doi.org/10.1007/s00330-­0 13-­3 030-­8
Variability of CT attenuation measurements in virtual Mileto A, Marin D, Alfaro-Cordoba M et  al (2014d)
unenhanced images generated using multimaterial Iodine quantification to distinguish clear cell from
decomposition from fast kilovoltage-switching dual-­ papillary renal cell carcinoma at dual-energy multide-
energy CT.  Acad Radiol 24(3):365–372. https://doi. tector CT: a multireader diagnostic performance study.
org/10.1016/j.acra.2016.09.002 Radiology 273(3):813–820. https://doi.org/10.1148/
Kim JH, Yoon HJ, Lee E, Kim I, Cha YK, Bak SH (2021) radiol.14140171
Validation of deep-learning image reconstruction for Mileto A, Nelson RC, Paulson EK, Marin D (2015) Dual-­
low-dose chest computed tomography scan: empha- energy MDCT for imaging the renal mass. Am J
sis on image quality and noise. Korean J Radiol Roentgenol 204(6):640–647. https://doi.org/10.2214/
22(1):131–138. https://doi.org/10.3348/kjr.2020.0116 AJR.14.14094
Leng S, Yu L, Fletcher JG, McCollough CH (2015) Mileto A, Nelson RC, Larson DG et al (2017) Variability
Maximizing iodine contrast-to-noise ratios in abdomi- in radiation dose from repeat identical CT examina-
nal CT imaging through use of energy domain tions: longitudinal analysis of 2851 patients undergo-
noise reduction and virtual monoenergetic dual-­ ing 12,635 thoracoabdominal CT scans in an academic
294 A. Pourvaziri et al.

health system. Am J Roentgenol 208(6):1285–1296. nal dual-energy computed tomography: correlation

https://doi.org/10.2214/AJR.16.17070 between true noncontrast and virtual noncontrast
Obmann MM, Kelsch V, Cosentino A, Hofmann V, Boll images considering renal and liver cysts and adrenal
DT, Benz MR (2019) Interscanner and intrascan- masses. J Comput Assist Tomogr 41(2):294–297.
ner comparison of virtual unenhanced attenuation https://doi.org/10.1097/RCT.0000000000000503
values derived from twin beam dual-energy and Smith AD, Shah SN, Rini BI, Lieber ML, Remer EM
dual-source, dual-energy computed tomography. (2010) Morphology, attenuation, size, and structure
Investig Radiol 54(1):1. https://doi.org/10.1097/ (MASS) criteria: assessing response and predicting
RLI.0000000000000501 clinical outcome in metastatic renal cell carcinoma
Ohana M, Labani A, Jeung MY, Ghannudi SE, Gaertner S, on antiangiogenic targeted therapy. Am J Roentgenol
Roy C (2015) Iterative reconstruction in single source 194(6):1470–1478. https://doi.org/10.2214/
dual-energy CT pulmonary angiography: is it sufficient AJR.09.3456
to achieve a radiation dose as low as state-of-the-art Takahashi N, Vrtiska TJ, Kawashima A et  al (2010)
single-energy CTPA? Eur J Radiol 84(11):2314–2320. Detectability of urinary stones on virtual nonen-
https://doi.org/10.1016/j.ejrad.2015.07.010 hanced images generated at pyelographic-phase dual-­
Pandharipande PV, Gervais DA, Hartman RI et al (2010) energy CT.  Radiology 256(1):184–190. https://doi.
Renal mass biopsy to guide treatment decisions for org/10.1148/radiol.10091411
small incidental renal tumors: a cost-effectiveness Takeuchi M, Kawai T, Ito M et  al (2012) Split-bolus
analysis. Radiology 256(3):836–846. https://doi. CT-urography using dual-energy CT: feasibility,
org/10.1148/radiol.10092013 image quality and dose reduction. Eur J Radiol
Park SY, Kim CK, Park BK (2014) Dual-energy CT in 81(11):3160–3165. https://doi.org/10.1016/j.
assessing therapeutic response to radiofrequency abla- ejrad.2012.05.005
tion of renal cell carcinomas. Eur J Radiol 83(2):73– Tan BJ, El-Hakim A, Morgenstern N, Semerdzhiev
79. https://doi.org/10.1016/j.ejrad.2013.11.022 Y, Smith A, Lee BR (2004) Comparison of laparo-
Patel BN, Bibbey A, Choudhury KR, Leder RA, Nelson scopic saline infused to dry radio frequency ablation
RC, Marin D (2017) Characterization of small (< 4 of renal tissue: evolution of histological infarct in the
cm) focal renal lesions: diagnostic accuracy of spec- porcine model. J Urol 172(5):2007–2012. https://doi.
tral analysis using single-phase contrast-enhanced org/10.1097/01.ju.0000138083.84066.98
dual-energy CT.  Am J Roentgenol 209(4):815–825. Tappouni R, Kissane J, Sarwani N, Lehman EB (2012)
https://doi.org/10.2214/AJR.17.17824 Pseudoenhancement of renal cysts: influence of lesion
Pourvaziri A, Parakh A, Mojtahed A, Kambadakone A, size, lesion location, slice thickness, and number of
Sahani DV (2019) Diagnostic performance of dual-­ MDCT detectors. Am J Roentgenol 198(1):133–137.
energy CT and subtraction CT for renal lesion detec- https://doi.org/10.2214/AJR.10.6057
tion and characterization. Eur Radiol. https://doi. Vandenbroucke F, Van Hedent S, Van Gompel G et  al
org/10.1007/s00330-­019-­06224-­6 (2015) Dual-energy CT after radiofrequency abla-
Racine D, Becce F, Viry A et al (2020) Task-based charac- tion of liver, kidney, and lung lesions: a review of
terization of a deep learning image reconstruction and features. Insights Imaging 6(3):363–379. https://doi.
comparison with filtered back-projection and a partial org/10.1007/s13244-­015-­0408-­y
model-based iterative reconstruction in abdominal Vernuccio F, Meyer M, Mileto A, Marin D (2018) Use of
CT: a phantom study. Phys Med 76:28–37. https://doi. dual-energy computed tomography for evaluation of
org/10.1016/j.ejmp.2020.06.004 genitourinary diseases. Urol Clin North Am 45(3):297–
Sellerer T, Noël PB, Patino M et  al (2018) Dual-energy 310. https://doi.org/10.1016/j.ucl.2018.03.012
CT: a phantom comparison of different platforms for Yu L, Christner JA, Leng S, Wang J, Fletcher JG,
abdominal imaging. Eur Radiol 28(7):2745–2755. McCollough CH (2011) Virtual monochromatic
https://doi.org/10.1007/s00330-­017-­5238-­5 imaging in dual-source dual-energy CT: radiation
­
Shuman WP, Green DE, Busey JM et  al (2014) Dual-­ dose and image quality. Med Phys 38(12):6371–6379.
energy liver CT: effect of monochromatic imaging on https://doi.org/10.1118/1.3658568
lesion detection, conspicuity, and contrast-to-noise Yu L, Leng S, McCollough CH (2012) Dual-energy
ratio of hypervascular lesions on late arterial phase. CT–based monochromatic imaging. Am J
Am J Roentgenol 203(3):601–606. https://doi. Roentgenol 199(5):9–15. https://doi.org/10.2214/
org/10.2214/AJR.13.11337 AJR.12.9121
Silva AC, Morse BG, Hara AK, Paden RG, Hongo Zhao Y, Wu Y, Zuo Z, Suo H, Zhao S, Zhang H (2017) CT
N, Pavlicek W (2011) Dual-energy (spectral) CT: pulmonary angiography using different noise index
applications in abdominal imaging. Radiographics values with an iterative reconstruction algorithm and
31(4):1031–1046. https://doi.org/10.1148/ dual energy CT imaging using different body mass
rg.314105159 indices: image quality and radiation dose. J Xray
Slebocki K, Kraus B, Chang D-H, Hellmich M, Maintz Sci Technol 25(1):79–91. https://doi.org/10.3233/
D, Bangard C (2017) Incidental findings in abdomi- XST-­160608
 Urogenital Imaging: Kidneys:
Urinary Stones

Nils Große Hokamp

Contents
1 Introduction: Clinical Workup in Suspected Urolithiasis   295
2  ual-Energy CT in Suspected Urolithiasis: Protocol Decisions 
D  296
2.1  Protocol Recommendations   296
3  ECT-Concepts to Determine Stone Composition 
D  297
3.1  In Vitro Capabilities for Stone Composition   297
3.2  In Vivo Capabilities for Stone Composition   297
4 Advanced Concepts   299
4.1   tone Composition 
S  299
4.2  Other Applications   299
4.3  Future Developments   299
5 Conclusion   300
References   300

Abstract identified using modern DECT scanners and

software, evidence for characterization beyond
Urolithiasis is a common disease in the devel- uric acid varies. Depending on the scanner
oped world with an incidence as high as 15%; type used to acquire images different protocol
furthermore, the risk for a recurrent episode of and imaging strategies have been suggested. In
kidney stone disease is as high as 50%. CT has future, the advent of >3-material decomposi-
become the standard diagnostic tool to detect tion enabled by photon-counting CT might
and locate kidney stones. Dual-energy CT further improve in vivo stone characterization.
might increase the role in diagnostic work up
of suspected kidney stone disease as it can pro-
vide guidance with regards to the type of stone
detected. While uric acid stones can be reliably 1 Introduction: Clinical
Workup in Suspected
Urolithiasis
N. Große Hokamp (*)
Institute for Diagnostic and Interventional Radiology,
University Hospital Cologne, Cologne, Germany Flank pain and hematuria are the most common
e-mail: [email protected] symptoms in patients with kidney stone disease.

© The Author(s), under exclusive license to Springer Nature Switzerland AG 2022 295
H. Alkadhi et al. (eds.), Spectral Imaging, Medical Radiology Diagnostic Imaging,
https://doi.org/10.1007/978-3-030-96285-2_20
296 N. Große Hokamp

In the emergency setting, besides history and obtained by fusion of low and high energy tube
physical examination, ultrasonography is fre- information in the image domain, some noise is
quently performed as first diagnostic testing in introduced. A study by Franken et  al. demon-
suspected urolithiasis. Ultrasonography is par- strated that in this setting some (particularly
ticular helpful in identifying obstructive stones as small) stones (12%) were missed (Franken et al.
indicated by pyeloureteral dilatation. However, a 2018). In spectral detector DECT no prospective
computed tomography scan of the abdomen and decision toward acquisition of DECT data has to
pelvis represents the diagnostic modality of be made as information is gathered in every scan;
choice in adults with suspected kidney stones dis- however, dual-energy information is available in
ease (Brisbane et al. 2016; Curhan et al. 2014). scans conducted with 120 kVp tube voltage, only,
Computed tomography has an excellent sensi- possibly inappropriate for small patients.
tivity and specificity in for detection of ureteral To overcome dose and/or noise restrictions in
calculi (>0.95 for both) even when using low-­ emission-base DECT, different strategies are
dose imaging protocols (Brisbane et  al. 2016). available: Advances in image reconstruction led
Such low dose protocols do not impair diagnostic to a relevant noise reduction in blended images
accuracy in non-obese patients, while in patients obtained from dual source DECT. Yet, a proof of
with a body-mass-index <30  kg/m2, smaller concept to which extent this allows for improved
stones might be missed (Franken et  al. 2018). detection of calculi has not been conducted. In
However, the limit of detection has been chal- this context, it needs to be restated, that the rele-
lenged by modern image reconstruction algo- vance of small stones can be clinically questioned
rithms and furthermore, the clinical relevance of according to many guidelines; however, in symp-
very small stones may be questioned (Chang tomatic patients, the diagnosis of small calculi
et al. 2019). Therefore, in clinical routine a low might be of diagnostic value to end the search for
dose computed tomography scan is conducted differential diagnoses. These observations
even in (moderately) obese patients. account for blended images (representing con-
ventional images), only, and can certainly be
overcome by not-enabling dual-energy mode.
2  ual-Energy CT in Suspected
D To combine availability of low dose protocols
Urolithiasis: Protocol and improved stone characterization by means of
Decisions dual-energy CT, a hybrid approach has been sug-
gested: Here, a standard CT scan without
As elucidated above, scans in suspected urolithia- DE-mode enabled is conducted first and inter-
sis are preferably conducted using low dose pro- preted for presence of stone. In case, that a kid-
tocols. According to common understanding in ney stone is present, a second DECT scan (with
this context, this refers to protocols with a com- higher dose) is conducted in position of the cal-
puted tomography dose index (CTDI) up to culi, only. Using this strategy, a reasonable radia-
5 mGy (Brisbane et al. 2016; Scott Kriegshauser tion dose can be well achieved (Manglaviti et al.
et al. 2015). This reference needs to be considered 2011; Nestler et al. 2019). It needs to be consid-
when considering DECT to improve stone charac- ered, that this approach might be disruptive to
terization depending on the system used: In standardized workflows and possibly prone to
sequential scanning, the radiation dose is signifi- errors if the initial reading is faulty.
cantly higher as compared to standard acquisition.
In kVp-switching DECT, low dose examinations
can be carried out while dose modulation may or 2.1 Protocol Recommendations
may not be used depending on generation of sys-
tems in use; this possibly limits the extent of dose Emission-based DECT: Low dose standard CT
reduction. In dual source DECT, low dose scan- from upper pole of the kidneys to pelvis (axial, thin
ning is possible; however, as standard images are slices, and coronal 3 mm reconstructions). If stone
Urogenital Imaging: Kidneys: Urinary Stones 297

present additional higher dose DECT acquisition 3.1 I n Vitro Capabilities for Stone
over stone (axial, thin slices, further processing Composition
pending on the vendor provided software solution).
Detector-based DECT: Low dose from upper pole Several groups explored the benefits from DECT
of the kidneys to pelvis (axial, thin slices, and coro-
in assessing kidney stone composition in  vitro.
nal 3  mm reconstructions and further processing Again, most groups used the DECT ratio or -
pending on the vendor provided software solution). index concept to assess stone composition. The
majority of studies used kidney stones harvested
from patients undergoing infrared spectroscopy
3 DECT-Concepts to Determine or radiograph diffraction analysis to determine
Stone Composition the standard of reference (Nestler et al. 2019).
One of the earliest studies by Boll et al. inves-
The vast majority of stones encountered in vivo tigated differentiation between a wider variety of
consist of either calcium-oxalate mono- or dihy- stones (i.e. calcific, uric acid, cystine, struvite,
drate (sometimes referred to as whewellite and and mixed) using their DECT ratio. They report
weddelitte, 70–85%). Followed by uric acid (10– an excellent differentiation between these types
15%) and struvite (sometimes referred to as using a dual source DECT; however, images were
infect-stones, 5–10%). Cystine, calcium-­acquired with a CTDI of 22.4 mGy which appears
phosphate (brushit), and xanthine are considered unrealistic for a clinical scan (Boll and Patil
rare (each <5%) (Große Hokamp et al. 2018). 2009). Similarly, our group more recently inves-
Standard CT allow for differentiation between tigated the capability for stone differentiation
calcific and non-calcific stones. The vast majority using dual layer DECT comparing low and nor-
of the latter are uric acid calculi which (as rule of mal dose protocols. Here, we reported that using
thumb) exhibit attenuation between 200 and a normal dose protocol (CTDI 10 mGy), calcific,
400  HU, while calcific stones usually exhibit cystine, struvite, uric acid, and xanthine stones
attenuation >600 HU. However, there is a signifi- can be reliably identified based on their DECT
cant overlap between the HU of different stone ratio. Using, low dose this discrimination was
types which impairs the ability to securely diag- rendered unfeasible for calcific and cystine
nose stone composition based on attenuation stones, while the other stone types remained
characteristics in conventional CT. identifiable (Große Hokamp et al. 2018).
Stone composition analysis can be improved by
means of DECT, where attenuation of low and
high energy photons is separately registered. The 3.2 In Vivo Capabilities for Stone
slope between attenuation of low and high energy Composition
photons can be understood as a simplified approxi-
mation of the material-specific attenuation coeffi- Most in  vivo validation in this regard has been
cient. Hence, this DECT-slope or its inverse carried out for uric acid as its DECT ratio is close
(DECT ratio) can be used to characterize a kidney to 1 and therefore clearly different from other
stone’s main component. Other mathematical materials encountered in kidney stone imaging.
operations have been suggested including the so- Several groups investigated uric acid versus other
called dual-energy index or the spectral coeffi- stones and report excellent results for both, in and
cient. Despite being calculated differently, they ex vivo (Franken et al. 2018; Nestler et al. 2018;
relate to the very same concept (Graser et al. 2008; Große Hokamp et al. 2020). For example, Graser
Große Hokamp et  al. 2018). In this context, the et al. reported that the dual-energy index can be
attenuation characteristics of uric acid need to be applied both in and ex  vivo to differentiate
highlighted as they are quite unique: Here, attenu- between uric acid and non-uric acid stones; how-
ation of low and high energy photons occurs to a ever, they found an overlap between mixed UA
similar extent rendering this material well identifi- and calcified stones (Graser et al. 2008). In line,
able with all available DECT techniques (Fig. 1). Eiber et al. reported excellent performance of the
298 N. Große Hokamp

a-1 b-1

a-2

b-2

Fig. 1  Two coronal images of two patients with a uric monoenergetic images of 40–200 keV. Note that uric acid
acid stone (a-1) and a calcific, likely calcium-oxalate shows nearly equivalent attenuation in both low and high
stone (b-1). Attenuation characteristics of low and high keV (a-2). Images acquired on a dual layer DECT. Image
energy attenuation are illustrated for both stones in the courtesy of Robert P. Reimer, University Hospital Cologne
lower part (a-2 and b-2) as approximated by virtual

dual-energy index for differentiating between Beyond this binary differentiation fewer reports
uric acid and non-uric acid stones using dual are available. Mangliaviti et al. reported that using
energy index-based assessment on a dual source a dual source DECT, they were able to differenti-
scanner while minor misclassifications occurred ate Calcium-oxalate, cystine, and uric acid stones
in mixed UA stones (Eiber et al. 2012). As a mat- with 100% accuracy; however, only stones >5 mm
ter of fact, most vendors now provide the so-­ in diameter were included in their analysis.
called uric acid maps as a standard reconstruction Differences between in vitro and in vivo accu-
from DECT data which illustrate structures con- racy for stone composition assessment likely result
taining uric acid. Commonly, these maps are from increase in cross scatter and a relative dose
illustrated color-coded and superimposed to the reduction due to inhomogeneity of the human
conventional gray-scale CT image (Fig. 2). body compared to homogeneous phantoms.
Urogenital Imaging: Kidneys: Urinary Stones 299

Fig. 2  Small calculi in the right kidney including plotting be likely made of brushit. Image courtesy of Matthias
of the DECT ratio as provided by the vendor analysis tool Frank Frölich, University Hospital Mannheim
(Dual Source DECT). The software deemed this stone to

4 Advanced Concepts media is feasible in up to 600 HU contrast-­

associated attenuation and that the resulting vir-
4.1 Stone Composition tual non-contrast images allow for detection of
renal calculi with a sensitivity >99% (Lazar et al.
As in any field related to imaging, the advent of 2020). This might hold benefit in the setting of
artificial intelligence holds promise to improve unclear hematuria. In this setting, conducting a
stone composition analysis as well. In a recent multiphasic CT examination represents the stan-
study by De Perrot et  al. it was reported that dard of care. To rule out kidney stone disease as a
radiomics and machine learning can help to dif- common reason for hematuria, an unenhanced
ferentiate phleboliths from small kidney stones acquisition is generally included in the diagnostic
which represents a frequent challenge in image workup; however, as suggested by their results,
interpretation. The authors used radiomics fea- this might be overcome due to the high sensitivity
tures to train a machine learning-based classifier of virtual non-contrast images if the examination
which allows for accurate differentiation in 85% is conducted on a DECT resulting in a significant
of cases (De Perrot et  al. 2019). Our group radiation dose reduction.
attempted to use a shallow neural network to
allow for reliable prediction of the main compo-
nent even in mixed stones. We found that this is 4.3 Future Developments
feasible in an ex  vivo setting; however, transla-
tion to in vivo studies has not been demonstrated Photon-counting CT has just become clinically
to date (Große Hokamp et al. 2020). available. This technology will likely impact kid-
ney stone imaging from two angles: (1) Higher
spatial resolution might allow for a more precise
4.2 Other Applications assessment of stone size. This is of importance as
treatment decision as of now is solely based on
Composition analysis is not the only possible stone size and location. Composition, on the other
application of DECT in assessment of urolithia- hand does not play a leading role in treatment
sis. Lazar et al., for example, investigated if cal- decision as of today, despite specific treatment
culi can be assessed in contrast enhanced options are available for certain stone types (e.g.
examinations by means of virtual non-contrast uric acid or struvite). (2) Photon-counting CT will
images. They report that the removal of contrast likely improve material decomposition and there-
300 N. Große Hokamp

fore possibly improve stone characterization. Chang D, Slebocki K, Khristenko E et  al (2019) Low-­
With a reliable means for in  vivo composition dose computed tomography of urolithiasis in obese
patients: a feasibility study to evaluate image recon-
analysis becoming available, targeted therapies struction algorithms. Diabetes Metab Syndr Obes
may gain importance in kidney stone disease. Targets Ther 12:439–445
Curhan G, Denu-ciocca CJ, Matlaga BR et  al (2014)
American Urological Association (AUA) guideline
medical management of kidney stones: American
5 Conclusion urological association medical management of kidney
stones. AUA Clin Guidel:1–26
De Perrot T, Hofmeister J, Burgermeister S et  al (2019)
Dual-Energy CT is helpful in assessment of kidney Differentiating kidney stones from phleboliths in unen-
stone disease. It allows for reliable differentiation hanced low-dose computed tomography using radiomics
of uric acid and non-uric acid stones based on ven- and machine learning. Eur Radiol 29(9):4776–4782
dor provided reconstructions. Furthermore, a dif- Eiber M, Holzapfel K, Frimberger M et al (2012) Targeted
dual-energy single-source CT for characterisation of
ferentiation between different types of non-­ uric urinary calculi: experimental and clinical experience.
acid stones (e.g. struvite, xanthine) appears feasible Eur Radiol 22(1):251–258
according to recent literature. When conducting a Franken A, Gevenois PA, Van Muylem A et al (2018) In
CT scan with question of kidney stones on an emis- vivo differentiation of uric acid versus non-uric acid
urinary calculi with third-generation dual-source
sion-based DECT, carrying out a low dose conven- dual-energy CT at reduced radiation dose. AJR Am J
tional scan (DECT-mode disabled) followed by a Roentgenol 210(2):358–363
targeted DECT scan on the kidney stone in ques- Graser A, Johnson TRC, Bader M et  al (2008) Dual
tion appears appropriate to reduce radiation dose. energy CT characterization of urinary calculi: ini-
tial in  vitro and clinical experience. Investig Radiol
Future applications include a more detailed stone 43(2):112–119
composition analyses using artificial intelligence Große Hokamp N, Salem J, Hesse A et  al (2018) Low-­
and photon-counting CT as well as reliable stone dose characterization of kidney stones using spectral
identification on virtual non-contrast images. detector computed tomography: an ex  vivo study.
Investig Radiol 53(8):457–462
Große Hokamp N, Lennartz S, Salem J et  al (2020)
Compliance with Ethical Standards Dose independent characterization of renal stones
by means of dual energy computed tomography
Funding None. and machine learning: an ex-vivo study. Eur Radiol
30(3):1397–1404
Disclosure of Interests Nils Große Hokamp receives Lazar M, Ringl H, Baltzer P et al (2020) Protocol analy-
speaker’s fees and research support from Philips sis of dual-energy CT for optimization of kidney stone
Healthcare. Nils Große Hokamp is consultant for Bristol- detection in virtual non-contrast reconstructions. Eur
Myers Squibb. Nils Große Hokamp is on the editorial Radiol 30(8):4295–4305
board of European Radiology. Manglaviti G, Tresoldi S, Guerrer CS et  al (2011) In
vivo evaluation of the chemical composition of uri-
nary stones using dual-energy CT. Am J Roentgenol
Studies involving human 197(1):76–83
Nestler T, Nestler K, Neisius A et  al (2018) Diagnostic
Ethical Approval This chapter does not contain any accuracy of third-generation dual-source dual-energy
studies with human participants performed by any of the CT: a prospective trial and protocol for clinical imple-
authors. mentation. World J Urol
Nestler T, Nestler K, Neisius A, Isbarn H, Netsch C,
Waldeck S, Schmelz HU, Ruf C. Diagnostic accu-
racy of third-generation dual-source dual-energy
References CT: a prospective trial and protocol for clinical
implementation. World J Urol. 2019;37(4):735–741.
Boll DT, Patil NA, Paulson EK et al (2009) Renal stone https://doi.org/10.1007/s00345-018-2430-4. Epub
assessment with dual-energy multidetector CT and 2018 Aug 3. PMID: 30076456.
advanced postprocessing techniques: improved char- Scott Kriegshauser J, Naidu SG, Paden RG et  al (2015)
acterization of renal stone composition--pilot study. Feasibility of ultra-low radiation dose reduction for
Radiology 250(3):813–820 renal stone CT using model-based iterative reconstruc-
Brisbane W, Bailey MR, Sorensen MD (2016) An over- tion: prospective pilot study. Clin Imaging 39(1):99–103
view of kidney stone imaging techniques. Nat Rev
Urol 13(11):654–662
 Skeletal Imaging: Bones

Christian Booz, Julian L. Wichmann,
and Tommaso D’Angelo

Contents
1 Metal Artifact Reduction Techniques   302
1.1  I ntroduction   302
1.2  Technical Background   302
1.3  Literature Overview   303
2 Bone Marrow Assessment   304
2.1  I ntroduction   304
2.2  Technical Background   304
2.3  Applications in Clinical Routine   305
3 Bone Mineral Density Analysis   307
3.1  I ntroduction   307
3.2  Technical Background   308
3.3  Initial Experience in Literature   308
4 Future Innovations   310
5 Conclusion   310
References   311

Abstract responding virtual monoenergetic images,

significant metal artifact reduction can be
Since the advent of dual-energy CT, improve- achieved compared to conventional CT, allow-
ments have been achieved in CT imaging of ing for a more detailed assessment of metal
the skeleton. Through reconstruction of a vir- implants and surrounding tissue. Furthermore,
tual monochromatic energy spectrum and cor- visualization of bone marrow pathologies is
achieved by creation of virtual non-calcium
images which enable subtraction of calcium
C. Booz (*) · J. L. Wichmann from cancellous bone based on three-material
Department of Diagnostic and Interventional
Radiology, University Hospital Frankfurt, decomposition. In addition, phantomless bone
Frankfurt, Hassia, Germany mineral density measurements of trabecular
T. D’Angelo bone can be performed with dual-energy CT,
Department of Diagnostic and Interventional allowing for opportunistic osteoporosis
Radiology, University Hospital Messina, screening in context of routine CT scans,
Messina, Sicily, Italy

© The Author(s), under exclusive license to Springer Nature Switzerland AG 2022 301
H. Alkadhi et al. (eds.), Spectral Imaging, Medical Radiology Diagnostic Imaging,
https://doi.org/10.1007/978-3-030-96285-2_21
302 C. Booz et al.

potentially resulting in avoidance of further 1998). In this context, invasive joint aspiration is
examinations and—particularly for patients the definitive test in the setting of suspected
undergoing regular follow-up CT examina- infection potentially causing complications such
tions—significant radiation exposure reduc- as joint inoculation in the presence of overlying
tion. With introduction of photon-counting cellulitis (White et al. 2000).
CT, advances regarding all mentioned applica- CT has suffered from image degradation
tions are expected due to its technical advances because of excessive beam attenuation of metal
compared to dual-energy CT.  In addition, implants, that leads to photon starvation, radia-
photon-­counting CT may also further facili- tion scatter, beam hardening artifact, excessive
tate visualization of complex and thin skeletal quantum noise, and scatter edge effects (Nicolaou
structures by providing highest spatial resolu- et al. 2012; Mallinson et al. 2016). There are spe-
tion imaging, as well as bone tumor identifica- cific methods that can reduce the degree of arti-
tion and characterization due to improved fact found on conventional polychromatic CT
material decomposition compared to dual-­ images including optimizing patient position,
energy CT. extending attenuation scales, iterative reconstruc-
tion utilization, use of soft tissue filters, and
increasing the tube voltage or current. These
Abbreviations approaches can significantly decrease photon
starvation and beam hardening artifact but can
BMD Bone mineral density come at the expense of increased dose, decreased
BME Bone marrow edema soft tissue definition, and diminished spatial res-
DEratio Dual-energy ratio olution (Lee et al. 2007). Hardware and software
DXA Dual y-ray absorptiometry requirements also limit availability and can hin-
HU Hounsfield unit der widespread utility.
MARS Metal artifact reduction system Dual-energy CT provides new possibilities for
ROI Region of interest metal artifact reduction through implementation of
VMI Virtual monoenergetic image energy-specific postprocessing by allowing for
VNCa Virtual non-calcium reconstruction of a virtual monochromatic energy
VOI Volume of interest spectrum and corresponding virtual monoener-
getic image (VMI) series. In this context, the
higher energy beam (typically set to 140  kV)
1  etal Artifact Reduction
M undergoes less attenuation and therefore less beam
Techniques hardening, while the lower energy beam (80–
100 kV) provides superior soft tissue contrast.
1.1 Introduction

Radiologists are frequently faced with the task of 1.2 Technical Background
analyzing metallic prostheses for periprosthetic
fractures, metallic failure and fracture, liner wear, Images created from a dual-energy CT dataset
aseptic loosening, infection, particle disease, and can simulate those acquired with a monochro-
recurrent tumors after surgery in clinical routine. matic beam set between 40 and 190  keV.  This
Radiographs are commonly obtained after such exceeds the physical range of the beams but is
surgeries, although low sensitivity and specificity achieved by adjusting the weighting value
have been reported (Viano et  al. 2000; Lüdeke between the two datasets when blending the data
et al. 1985). MRI is advocated for its superior soft creating the image. The basic principle for metal
tissue detail and sensitivity for bone marrow artifact reduction is that monochromatic images
edema (BME), however, distortion due to metal are created from projection space data, which
artifact remains a significant problem (Suh et al. demonstrate a lower susceptibility to beam hard-
Skeletal Imaging: Bones 303

a b

Fig. 1  Metal artifact reduction in dual-energy CT. Case image (VMI) series (b, 130 keV in this case) significant
of a 67-year-old woman with dorsal spondylodesis in the metal artifact reduction can be achieved compared to stan-
lumbar spine undergoing regular dual-energy CT imag- dard linearly-blended images (a), allowing for better
ing. Through creation of high keV virtual monoenergetic assessment of the foreign material and surrounding tissue

ening artifacts (Nicolaou et al. 2012; Mallinson 1.3 Literature Overview

et al. 2016; Yu et al. 2009). In this context, VMI
allows for reconstruction of dual-energy CT data- There have been several studies evaluating the
sets at a chosen hypothetical energy level that potential of dual-energy CT to reduce metal arti-
would result from an acquisition with a true facts by application of VMI in the last decade.
monoenergetic X-ray beam (Fig. 1). Guggenberger et al. analyzed spine fusion devices
In clinical routine, these images are readily to in phantoms by using single-energy 120-­kV CT
reconstruct once the data have been processed. It imaging and 140/100-kV dual-energy CT imag-
is performed by the application of a simple slid- ing and demonstrated that dual-energy CT pro-
ing tool on the dual-energy CT workstation that vided both improved image quality and reduced
allows for the image energy level to be custom- metal artifacts (Guggenberger et al. 2012). In this
ized continuously, comparable to window/width context, the optimal dual-energy CT energy range
level adjustment. This helps the reader to opti- was found to be between 124 and 146 keV.
mize the balance between soft tissue detail and Bamberg et  al. compared reconstructed high-­
artifact reduction. energy VMI series derived from dual-source dual-
In case of metal implants, VMI keV levels energy CT images with conventional CT spectra in
between 110 and 146  keV have been demon- a group of 31 patients with various metallic skele-
strated to provide best image quality for assessing tal implants (Bamberg et  al. 2011). Subjectively,
foreign material and surrounding tissue, depend- superior image quality was found in 29 of 31
ing on implant size and type (Nicolaou et al. 2012; patients and superior diagnostic quality was dem-
Mallinson et al. 2016; Meinel et al. 2012; Bamberg onstrated in 27. Objectively, artifact attenuation
et al. 2011; Zhou et al. 2011). Finally, the mono- decreased from 2882 to 2341 Hounsfield units
chromatic spectrum can also be used to minimize (HU). Importantly, several lesions were discern-
noise at the optimal tube voltage energy. able only on high-energy VMI series.
304 C. Booz et al.

Meinel et  al. investigated the optimal settings onstrated such as improved material decomposi-
for dual-source dual-energy CT evaluation of hip tion. In this context, virtual non-calcium (VNCa)
prostheses both in phantoms and in 22 patients imaging has become an increasingly used post-
(Meinel et al. 2012). They concluded that the opti- processing application that allows for subtraction
mum tube voltage settings were 140/100-kV with calcium from anatomical structures, enabling
extrapolated energies between 105 and 120  keV, color-coded visualization of bone marrow pathol-
providing high image quality in all types of metal ogies such as BME (Fig.  2) (Pache et  al. 2010;
implants. Moreover, the optimized reconstructions Booz et al. 2019a, 2020a, b; Frellesen et al. 2018;
demonstrated additional findings unseen on con- Koch et al. 2021).
ventional CT images, including disk protraction, Bone marrow pathologies are usually associ-
and osteonecrosis and hardware malposition. ated with a reduction of fat component in the tra-
Zhou et al. investigated 47 patients with ortho- becular bone, replaced by water, hemorrhage or
pedic devices in the context of fractures by using cancer tissue depending on the underlying pathol-
a dual-source CT system and found that an ogy. While bone marrow assessment on conven-
extrapolated tube voltage of 130  keV provided tional CT is impeded by the presence of calcium,
optimal and superior images with significantly MRI represents the current gold standard tech-
reduced metal artifact compared with conven- nique for assessing bone marrow disorders
tional 120 kV datasets (Zhou et al. 2011). including traumatic BME but also oncologic and
Lee et  al. analyzed both phantoms and 26 inflammatory bone marrow disorders. However,
patients demonstrating that rapid kilovoltage due to its limitations in clinical routine such as
switching dual-energy CT in combination with contraindications or limited availability, dual-­
metal artifact reduction software (MARS) energy CT may be considered a potentially
reduced metal artifacts and improved prosthetic/ cheaper, faster, and more available imaging alter-
periprosthetic assessment/visualization (Lee native for bone marrow assessment through cre-
et al. 2007). However, this study advised caution ation of VNCa reconstructions. In this context, a
in case of titanium implants, as the image quality large body of evidence has demonstrated the
was found to be slightly poorer compared with potential of VNCa imaging to serve as a viable
other prosthetic compositions. imaging alternative to MRI for bone marrow
In conclusion, there have been several both assessment in case of MRI contraindications or
in  vitro and in  vivo studies demonstrating that limited availability, particularly in emergency
dual-energy CT can significantly reduce metal setting for traumatized patients.
artifacts, improve image quality, and increase
diagnostic yield when compared with conven-
tional CT imaging. The extrapolated recom- 2.2 Technical Background
mended keV levels for VMI series vary between
110 and 146  keV.  In this context, the optimal Based on three-material decomposition, the
VMI keV level is likely to depend on the specific amount of calcium on dual-energy CT datasets is
dual-energy CT system, hardware size and com- estimated and subtracted from images to highlight
position, the pathologic condition being evalu- the anatomical structures that can be covered with
ated, and preferences of the radiologist. bone mineral or gross calcifications in VNCa
reconstructions. In this context, a baseline is made
connecting CT values of yellow and red marrow
2 Bone Marrow Assessment (Johnson 2012). Target voxels are projected to the
baseline using the characteristic slope of the
2.1 Introduction DEratio of calcium. As the baseline passes close to
the CT value of water (0 HU for both 100 kV and
With the advent of spectral imaging based on Sn140 kV), the differences among voxels on the
dual-energy CT, numerous and noteworthy baseline reflect mainly the water content in the
advantages over conventional CT have been dem- bone, with calcium removed. These differences
Skeletal Imaging: Bones 305

a b c d

Fig. 2  Bone marrow assessment in dual-energy CT. Case marrow edema (BME) (arrow, green colored area) in L1,
of a 62-year-old man with known impression fracture of indicating acute fracture components. Additionally, per-
L1 but recently increasing lumbar pain undergoing dual-­ formed sagittal MRI sequences (c STIR, d T1w) con-
energy CT of the spine. On sagittal conventional grayscale firmed the finding of acute fracture components within the
CT series (a), the L1 fracture was assessed as being com- fracture resulting in associated traumatic vertebral BME
pletely old. Creation of colored sagittal virtual non-­ (arrow in c)
calcium (VNCa) reconstructions (b) showed signs of bone

can be visually interpreted, using color-coded Technical limitations of VNCa imaging should
maps, or quantitatively assessed by means of also be taken into account. It has been demon-
region of interest (ROI) measurements (Booz strated the inability to accurately visualize minor
et al. 2020b; Johnson 2012). bone marrow alterations directly adjacent to cor-
Image quality on VNCa datasets is influenced tical bone due to incomplete masking of the cor-
by dual-energy CT scanning parameters. Best tex and to spatial averaging. Incomplete
results have been obtained with a DEratio of subtraction of cortical or cancellous bone might
70/150  kV.  However, when wide DEratio are not also occur in case of arthrosis, and in the pres-
recommended because of increase of image ence of gas or severe osteosclerosis, which can
noise, such as abdomen and pelvis, higher radia- cause beam hardening artifacts that may limit
tion doses help providing optimal image quality bone marrow assessment. For this reason, any
(Wang et al. 2013). Pitch and rotation time do not user of VNCa imaging should be aware of its
considerably affect image quality, although spiral potential pitfalls (Booz et  al. 2019a, b; Kaup
artifacts can appear when pitch is too low. Color-­ et al. 2016).
coded VNCa datasets are usually automatically
processed from raw-data of most of modern dual-­
energy CT platforms, with processing time last- 2.3 Applications in Clinical
ing few minutes, showing the potential to be Routine
time-efficiently used in routine clinical practice
(Kelcz et  al. 1979; Primak et  al. 2009). Slice 2.3.1 Spine
thickness of 1–2  mm and smoother reconstruc- MRI and CT are currently considered the diag-
tion kernels are recommended and datasets nostic imaging modalities of choice to evaluate
should be reformatted along two anatomical spine disorders. While CT imaging is indicated in
planes for optimal qualitative evaluation (Booz trauma setting to detect fracture lines due to its
et al. 2020a; Müller et al. 2019). excellent spatial resolution, MRI represents the
306 C. Booz et al.

gold standard technique for evaluation of disks, The spine represents the most common site of
nerves, musculotendinous structures, and bone bone metastases. Only breast, prostate, and lung
marrow disorders. Moreover, MRI is particularly cancers are together responsible for more than
useful to diagnose BME secondary to trauma, 80% of cases of metastatic bone disease.
which allows to assess the chronicity of a fracture Contrast-enhanced CT scan is regularly per-
by the presence of interstitial fluid or potential formed in oncologic patients. However, the
instability. However, MRI access can be limited assessment of bone marrow lesions on standard
in routinely trauma setting due to its high costs CT remains challenging. Therefore, patients with
and long acquisition times, which require pro- high suspicion for bone metastasis frequently
longed and potentially painful patient positioning need to undergo additional imaging, such as
as previously stated. MRI, scintigraphy, or positron emission tomog-
Several studies have been carried out to evalu- raphy. The efficacy of VNCa reconstructions to
ate the diagnostic performance of VNCa recon- detect metastatic spine lesions has been recently
structions to detect acute vertebral fractures assessed in several studies using different cal-
(Wang et  al. 2013; Kaup et  al. 2016; Petritsch cium suppression indices. In particular, the use of
et al. 2017). Traumatic BME detection has been low- and medium-suppression indices resulted in
qualitatively assessed by using color-coded an increase of about 85% concerning the sensitiv-
images, and quantitatively based on ROI mea- ity compared to conventional CT, associated with
surements of bone marrow attenuation. When a good inter-reader agreement at subjective image
vertebral microfractures are present within can- analysis (Abdullayev et al. 2019). In a study from
cellous bone, bone marrow attenuation increases Abdullayev et al. quantitative analysis using low-
since its fatty content is replaced by edema and and medium-suppression indices showed promis-
microhemorrhage. Color-coded VNCa recon- ing results to discriminate between normal and
structions show good to excellent results for metastatic bone, using thresholds of -143HU
qualitative assessment of vertebral BME, either and -31 HU, respectively (Abdullayev et al. 2019).
in terms of sensitivity (range: 72–96%), specific- High diagnostic accuracy of VNCa recon-
ity (range: 70–100%), and accuracy (range: structions has also been demonstrated for assess-
90–99%) (Wang et  al. 2013; Kaup et  al. 2016; ing infiltrative oncologic disease of vertebral
Petritsch et  al. 2017; Diekhoff et  al. 2017). In bone marrow such as multiple myeloma. In this
addition, quantitative analysis of vertebral BME context, studies have shown high accuracy of
on VNCa datasets has also demonstrated excel- VNCa imaging (ranging between 93% and 99%)
lent sensitivity, specificity, and accuracy, ranging, in depicting bone marrow alterations based on
respectively, between 85% and 96%, 82% and threshold values ranging between -45HU
90%, and 85% and 91%, with a threshold ranging and − 36 HU in objective analyses compared to
between −80 and 0 HU in these studies. MRI (Kosmala et al. 2017; Wang et al. 2017).
Promising results have also been carried out
for diagnosis of sacral insufficiency fracture-­ 2.3.2 Appendicular Skeleton
associated BME, showing high sensitivity and VNCa imaging has been shown to be particularly
specificity (93% and 95%, respectively) for qual- helpful to detect subtle, non-displaced hip frac-
itative assessment, and values of 85% and 95% tures that might be missed on conventional radio-
for quantitative assessment using a cut-off value graphs or conventional CT, especially in patients
of -43HU (Booz et  al. 2020b). This may allow affected by diffuse skeletal disorders such as
DECT to act as a promising technique to avoid osteoporosis or Paget’s disease. Different authors
misinterpretation of sacral insufficiency fractures focused on diagnostic performance of VNCa
and their related complications, particularly in reconstructions to detect pelvic fractures, using
patients suffering from osteoporosis or diffuse clinical follow-up as reference standard. In these
bone disease. studies, dual-energy CT performed superior
Skeletal Imaging: Bones 307

compared to standard CT, showing an improve- arthritis, either in large and small joints, showing
ment of sensitivity (>5%) when color-coded good qualitative assessment and excellent agree-
VNCa images were evaluated (Suh et  al. 2018; ment with MRI (Jans et al. 2018).
Burke et al. 2019). Moreover, quantitative analy-
sis with a threshold of −55.3 HU yielded a sensi-
tivity and specificity of 100% and 94%, 3  one Mineral Density
B
respectively (Jang et al. 2019). Analysis
Different rheumatological disorders such as
axial spondylarthritis and sacroiliitis usually 3.1 Introduction
require patients to undergo spine and pelvic MRI
for assessing bone marrow pathologies caused by Osteoporosis is a common bone disease affecting
inflammation. In this context, studies have shown older patient populations, and as such, a great
high diagnostic accuracy of VNCa reconstruc- deal of work has been done to find noninvasive,
tions for assessing inflammatory changes of pel- cost-efficient, safe, and expedient methods for
vic bone marrow (Wu et  al. 2019; Chen et  al. the diagnosis. As per current World Health
2020; Foti et  al. 2020). Authors concluded that Organization guidelines, the reference standard
dual-energy CT may serve as an appropriate for osteoporosis assessment and diagnosis is the
imaging alternative in case of MRI contraindica- utilization of dual X-ray absorptiometry (DXA)
tion or limited availability. to clinically assess bone mineral density (BMD)
Dual-energy CT can complement the informa- in conjunction with an evaluation of relevant
tion provided by conventional CT imaging and patient risk factors to determine a 10-year frac-
enhance the diagnostic capabilities of VNCa for ture risk (Kanis et  al. 2008). The accessibility,
evaluation of acute knee fractures. In a study by ease of use and interpretation, short acquisition
Booz et al., qualitative assessment of knee frac- time, high image resolution, stable instrument
tures by color-coded VNCa images yielded sensi- calibration, reliability, and very low radiation
tivity and specificity of 95%, while at quantitative doses associated with DXA have led to its wide-
analysis these values were 96% and 97%, respec- spread implementation within this clinical
tively, using a threshold of −51 HU (Booz et al. setting.
2020a). Similar results have been shown by Wang Despite the extensive use of DXA among cli-
et al., who proposed a cut-off of −67 HU, yield- nicians globally, there have been several studies
ing a sensitivity and specificity of 81% and 99%, describing the various shortcomings of this diag-
respectively (Wang et al. 2019). Compared with nostic tool (Bolotin and Sievänen 2001; Bolotin
conventional CT, dual-energy CT has demon- 2007). DXA is particularly susceptible to image
strated an increase of up to 20% regarding sensi- distortions secondary to overlying structures, and
tivity to detect fractures, especially for less more importantly, distortions secondary to osteo-
experienced radiologists (Yang et al. 2020). degenerative changes of the spine (Bolotin and
Several authors have investigated the perfor- Sievänen 2001; Bolotin 2007; Antonacci et  al.
mance of VNCa reconstructions to detect trau- 1996). Given that the prevalence of osteoporosis
matic BME in small bones of distal joints such as is highest among older patients in whom degen-
scaphoid and calcaneus (Booz et al. 2019a; Koch erative change is more common, DXA image dis-
et al. 2021; Müller et al. 2019). In these studies, tortions could limit its sensitivity and specificity.
dual-energy CT was able to highlight traumatic Moreover, DXA is a two-dimensional imaging
BME with high sensitivity and specificity com- modality that averages densities throughout the
pared to MRI, both for qualitative and quantita- entire vertebral body. Given the varying meta-
tive analysis. Additionally, it has been bolic activity among trabecular and cortical bone
demonstrated VNCa imaging also allows for high within the vertebral body, a 3D imaging modality
diagnostic accuracy in depicting inflammatory confined to areas of high metabolic activity may
bone marrow alterations related to rheumatoid provide more accurate assessment. It has also
308 C. Booz et al.

been proposed that fat content within trabecular equations are derived by using this model (for
bone can significantly modify attenuation values, illustrative purposes we used commonly applied
creating a source of error when utilizing DXA 80 and 140 keV levels):
(Genant and Boyd 1977).
Despite these concerns, quantitative CT imag- χ 80 HU = ( µ 80 – γ 80 g ) ⋅ VTB
ing was initially proposed as a viable modality to

( )
+ β 80 t – γ 80 g ⋅ VF + γ 80 g + δ

(1)
provide a true volumetric density calculation;
however, due to reliance on individual phantom χ 140 HU = ( µ 140 – γ 140 g ) ⋅ VTB
calibrations preventing opportunistic BMD
assessment derived from routine CT scans and a
( )
+ β 140 t – γ 140 g ⋅ VF + γ 140 g + δ (2)

high effective radiation dose (1–3  mSv), there These equations link the intensities Χ HU and 80

have been concerns which have significantly lim- Χ140HU in the two CT series obtained at tube ener-
ited its widespread utilization (Engelke et  al. gies of 80 and 140 kV to the fraction of the vol-
2008). ume occupied by the matrix material (bone
In this context, dual-energy CT has been pro- mineral + collagen) VTB and the volume of adi-
posed as a viable phantomless alternative for 3D pose tissue VF. The values for t and g are 0.92 and
volumetric opportunistic BMD assessment based 1.02, respectively, whereas the other variables are
on its improved material decomposition com- energy related constants. By calculating the mean
pared to conventional CT.  Additionally, dual-­ intensity for the trabecular bone in both CT data
energy CT can image patients with decreased sets, values for VTB and VF can be attained.
radiation doses when compared with conven- Finally, from VTB the BMD value ρBM (given in
tional CT, potentially leading to reduced radia- g/cm3) can be calculated by application of the
tion exposure for patients in clinical routine. material constants l = 3.06 g/cm3 and λ = 2.11:
1 ⋅ VTB
ρ BM = (3)
3.2 Technical Background 1+ λ
For assessment of spatial BMD distribution, a
Phantomless dual-energy CT-based volumetric specific BMD value for each voxel is finally
BMD assessment requires prior delineation of obtained.
the trabecular volume of interest (VOI) for each Despite dual-source CT-based BMD assess-
vertebra, which is either manually or automatic ment, similar approaches that allow for phantom-
determined using specific software depending on less volumetric BMD assessment based on other
the approach. For phantomless volumetric BMD dual-energy CT techniques such as dual-layer CT
assessment based on dual-source CT—which and fast kVp switching have been also described
currently represents the most frequently applied recently (Li et  al. 2020; van Hamersvelt et  al.
dual-energy CT BMD approach in clinical rou- 2017).
tine—the data obtained from VOI and the two
dual-energy CT kV series are used for volumetric
software-based BMD assessment on the basis of 3.3 Initial Experience in Literature
material decomposition for each voxel, as ini-
tially described by Nickoloff et al. and applied by Although earliest literature describing the possi-
Wesarg and Wichmann et al. (Fig. 3) (Nickoloff bility of dual-energy CT for BMD assessment
et al. 1988; Wichmann et al. 2014; Wesarg et al. was published in the late 1970s and throughout
2012). This algorithm for material decomposi- the 1980s, only few studies have evaluated the
tion is based on a biophysical model accounting potential of phantomless dual-energy CT-derived
for the five major substances of trabecular bone BMD assessment (Mallinson et  al. 2016;
(bone minerals, collagen matrix, water, red mar- Nickoloff et al. 1988). In 2012, Wesarg et al. ini-
row, and adipose tissue). The following two tially reported promising results using third-­
Skeletal Imaging: Bones 309

a b

c d

Fig. 3 Phantomless volumetric bone mineral density material decomposition for each voxel, as initially
(BMD) assessment using dual-energy CT.  Phantomless described by Nickoloff et al. and applied by Wesarg and
dual-energy CT-based volumetric BMD assessment Wichmann et al. (Nickoloff et al. 1988; Wichmann et al.
requires prior delineation of the trabecular volume of 2014; Wesarg et  al. 2012). After software-based BMD
interest (VOI) for each vertebra, which is commonly man- assessment, the results can be color-codedly visualized
ually determined using specific software depending on the (osteoporotic BMD, red; normal BMD, blue) for 12 zones
approach (a, b). For phantomless volumetric BMD assess- in each vertebra (c) with corresponding volumetric BMD
ment based on dual-source CT the data obtained from VOI values. Additionally, freely rotatable colored 3D visual-
and the two dual-energy CT kV series are used for volu- ization of the trabecular BMD distribution is feasible fur-
metric software-based BMD assessment on the basis of ther facilitating the analysis (d)

generation dual-source CT for phantomless between BMD values and pull-out forces
volumetric BMD assessment, however, this study (Wichmann et al. 2015). In addition, Booz et al.
was limited to exclusively in vitro vertebral body have shown that phantomless volumetric BMD
analysis (Wesarg et al. 2012). In 2014, Wichmann assessment based on dual-source CT yields supe-
et al. conducted an in vivo analysis of 160 lumbar rior diagnostic accuracy for the detection of osteo-
vertebrae and confirmed that 3D visualization of porosis compared to simple HU measurements
trabecular bone could be obtained from routinely (sensitivity 96% vs. 65%) (Booz et al. 2020c).
performed third-generation dual-source CT scans Zhou et  al. showed strong correlations
(Wichmann et  al. 2014). In another study, between dual-energy CT- and quantitative
Wichmann et  al. further evaluated cancellous CT-derived BMD at both the participant level and
BMD assessment of thoracic and lumbar pedicles the vertebral level (adjusted R2  =  0.983–0.987)
in a cadaver study and showed high correlation using rapid kVp switching (Zhou et al. 2021). Li
310 C. Booz et al.

et al. demonstrated in a phantom study that BMD spatial resolution photon-counting CT compared
can be accurately measured either by using rapid to conventional CT systems.
kVp switching dual-energy CT or quantitative Moreover, it has been shown that photon-­
CT with even smaller bias using dual-energy CT counting CT enables improved material decom-
(Li et al. 2020). Hamersvelt et al. evaluated the position even compared with dual-energy CT
possibility of accurate BMD quantification using (Willemink et al. 2018). In this context, photon-­
dual-layer spectral CT and showed strong linear counting CT may be used to reconstruct sharper
correlations (R2 ≥ 0.970, P < 0.001) to DXA (van and more detailed VNCa images, potentially
Hamersvelt et al. 2017). Additionally, Roski et al. allowing for more accurate evaluation of BME
found high correlations between BMD values without the need for MRI. In addition, improved
derived from dual-layer spectral CT and those material decomposition may also further improve
from quantitative CT by analyzing 174 vertebrae volumetric BMD assessment as well as provide
in 33 patients (Roski et al. 2019). new opportunities for bone tumor identification,
However, no data have been published that visualization and characterization.
directly evaluate the potential of this technique Finally, photon-counting CT potentially
for fracture prediction and identification of allows for improved evaluation of metal implants
patients at risk. In addition, cost-effectiveness and associated prosthesis loosening due to
studies are missing to date. Nevertheless, dual-­ reduced beam hardening artifacts, lack of elec-
energy CT and its phantom BMD application tronic noise, and higher spatial resolution. In this
represent a promising advance in the assessment context, metal artifact reduction algorithms that
of osteoporosis. rely on the multiple energy bins of photon-­
counting CT have been already proposed show-
ing promising initial results (Nasirudin et  al.
4 Future Innovations 2015).

With the advent of photon-counting CT, it has

been demonstrated that this new technique allows 5 Conclusion
for improved spatial resolution, higher contrast-­
to-­
noise ratio and substantial radiation dose In the last 10  years, dual-energy CT has been
reductions compared to conventional and dual-­ shown to provide additional clinically relevant
energy CT (Willemink et al. 2018). information compared to conventional CT in sev-
High-spatial resolution is essential for imag- eral musculoskeletal applications. Greatest expe-
ing of thin or complex skeletal structures such as rience exists in bone marrow assessment using
the temporal bone. Small structures such as the VNCa imaging to date, particularly for trauma,
auditory ossicles are inadequately depicted on but also for inflammatory and oncologic bone
low resolution images. Leng et  al. applied a marrow pathologies. In addition, dual-energy CT
research full-body photon-counting CT system to can significantly improve assessment of metal
scan the cadaveric temporal bone of a swine at a implants through reduction of artifacts using
high-spatial resolution (Leng et  al. 2016). VMI at high keV levels. Furthermore, phantom-
Photon-counting CT allowed for clear visualiza- less opportunistic volumetric BMD assessment is
tion of crucial anatomic structures, such as the provided, potentially resulting in avoidance of
stapes superstructure, while reducing the radia- further examinations in context of routine CT
tion dose compared with high-spatial resolution scans and significant radiation exposure reduc-
conventional CT, which uses a removable comb tion, particularly for young patients undergoing
that blocks up to three-quarters of the incident regular follow-up CT examinations. With the
photons. Additionally, visualization of complex advent of photon-counting CT, advances regard-
fractures of small bones (e.g. of the wrist) can ing all mentioned applications are expected due
potentially be further improved by using high-­ to its technical advances compared to dual-energy
Skeletal Imaging: Bones 311

CT.  In addition, photon-counting CT may also ficiency fracture in comparison to MRI. Eur J Radiol
further facilitate visualization of complex and 129:109046
Booz C, Noeske J, Albrecht MH, Lenga L, Martin SS,
thin skeletal structures by providing high-spatial Yel I et al (2020c) Diagnostic accuracy of quantitative
resolution imaging as well as bone tumor identi- dual-energy CT-based bone mineral density assess-
fication and characterization due to improved ment in comparison to Hounsfield unit measurements
material decomposition. using dual x-ray absorptiometry as standard of refer-
ence. Eur J Radiol 132:109321
Burke MC, Garg A, Youngner JM, Deshmukh SD, Omar
Compliance with Ethical Standards IM (2019) Initial experience with dual-energy com-
puted tomography-guided bone biopsies of bone
Disclosure of Interests  Authors declare that there are no lesions that are occult on monoenergetic CT.  Skelet
conflicts of interest present regarding this book chapter. Radiol 48(4):605–613
Chen M, Herregods N, Jaremko JL, Carron P, Elewaut D,
Van den Bosch F et al (2020) Bone marrow edema in
sacroiliitis: detection with dual-energy CT. Eur Radiol
30:3393–3400
References Diekhoff T, Hermann K, Pumberger M, Hamm B,
Putzier M, Fuchs M (2017) Dual-energy CT virtual
Abdullayev N, Hokamp NG, Lennartz S, Holz J, Romman non-­calcium technique for detection of bone marrow
Z, Pahn G et  al (2019) Improvements of diagnostic edema in patients with vertebral fractures: a prospec-
accuracy and visualization of vertebral metastasis tive feasibility study on a single-source volume CT
using multi-level virtual non-calcium reconstructions scanner. Eur J Radiol 87:59–65
from dual-layer spectral detector computed tomogra- Engelke K, Adams JE, Armbrecht G, Augat P, Bogado
phy. Eur Radiol 29(11):5941–5949 CE, Bouxsein ML et al (2008) Clinical use of quantita-
Antonacci MD, Hanson DS, Heggeness MH (1996) tive computed tomography and peripheral quantitative
Pitfalls in the measurement of bone mineral density by computed tomography in the management of osteo-
dual energy x-ray absorptiometry. Spine 21(1):87–90 porosis in adults: the 2007 ISCD Official Positions. J
Bamberg F, Dierks A, Nikolaou K, Reiser MF, Becker Clin Densitom 11(1):123–162
CR, Johnson TR (2011) Metal artifact reduction by Foti G, Faccioli N, Silva R, Oliboni E, Zorzi C, Carbognin
dual energy computed tomography using monoener- G (2020) Bone marrow edema around the hip in non-­
getic extrapolation. Eur Radiol 21(7):1424–1429 traumatic pain: dual-energy CT vs MRI.  Eur Radiol
Bolotin H (2007) DXA in  vivo BMD methodology: an 30(7):4098–4106
erroneous and misleading research and clinical gauge Frellesen C, Azadegan M, Martin SS, Otani K, DʼAngelo
of bone mineral status, bone fragility, and bone remod- T, Booz C et  al (2018) Dual-energy computed
elling. Bone 41(1):138–154 tomography-­based display of bone marrow edema in
Bolotin H, Sievänen H (2001) Inaccuracies inherent in incidental vertebral compression fractures: diagnostic
dual-energy X-ray absorptiometry in  vivo bone min- accuracy and characterization in oncological patients
eral density can seriously mislead diagnostic/prognos- undergoing routine staging computed tomography.
tic interpretations of patient-specific bone fragility. J Investig Radiol 53(7):409–416
Bone Miner Res 16(5):799–805 Genant HK, Boyd D (1977) Quantitative bone mineral
Booz C, Nöske J, Albrecht MH, Lenga L, Martin SS, analysis using dual energy computed tomography.
Wichmann JL et  al (2019a) Traumatic bone marrow Investig Radiol 12(6):545–551
edema of the calcaneus: evaluation of color-coded Guggenberger R, Winklhofer S, Osterhoff G, Wanner
virtual non-calcium dual-energy CT in a multi-reader G, Fortunati M, Andreisek G et  al (2012) Metallic
diagnostic accuracy study. Eur J Radiol 118:207–214 artefact reduction with monoenergetic dual-energy
Booz C, Nöske J, Martin SS, Albrecht MH, Yel I, Lenga CT: systematic ex vivo evaluation of posterior spinal
L et  al (2019b) Virtual noncalcium dual-energy CT: fusion implants from various vendors and different
detection of lumbar disk herniation in comparison with spine levels. Eur Radiol 22(11):2357–2364
standard grayscale CT. Radiology 290(2):446–455 Jang SW, Chung BM, Kim WT, Gil JR (2019)
Booz C, Nöske J, Lenga L, Martin SS, Yel I, Eichler Nondisplaced fractures on hip CT: added value of
K et  al (2020a) Color-coded virtual non-calcium dual-energy CT virtual non-calcium imaging for
dual-­energy CT for the depiction of bone mar- detection of bone marrow edema using visual and
row edema in patients with acute knee trauma: a quantitative analyses. Acta Radiol 60(11):1465–1473
multireader diagnostic accuracy study. Eur Radiol Jans L, De Kock I, Herregods N, Verstraete K, Van den
30(1):141–150 Bosch F, Carron P et al (2018) Dual-energy CT: a new
Booz C, Nöske J, Albrecht MH, Lenga L, Martin SS, imaging modality for bone marrow oedema in rheu-
Bucher AM et  al (2020b) Diagnostic accuracy of matoid arthritis. Ann Rheum Dis 77(6):958–960
color-coded virtual noncalcium dual-energy CT for Johnson TR (2012) Dual-energy CT: general principles.
the assessment of bone marrow edema in sacral insuf- Am J Roentgenol 199(5_suppl):S3–S8
312 C. Booz et al.

Kanis J, Johnell O, Odén A, Johansson H, McCloskey E Nickoloff E, Feldman F, Atherton J (1988) Bone mineral
(2008) FRAX™ and the assessment of fracture prob- assessment: new dual-energy CT approach. Radiology
ability in men and women from the UK. Osteoporos 168(1):223–228
Int 19(4):385–397 Nicolaou S, Liang T, Murphy DT, Korzan JR, Ouellette
Kaup M, Wichmann JL, Scholtz J-E, Beeres M, Kromen H, Munk P (2012) Dual-energy CT: a promising new
W, Albrecht MH et al (2016) Dual-energy CT–based technique for assessment of the musculoskeletal sys-
display of bone marrow edema in osteoporotic ver- tem. Am J Roentgenol 199(5_Suppl):S78–S86
tebral compression fractures: impact on diagnos- Pache G, Krauss B, Strohm P, Saueressig U, Blanke P,
tic accuracy of radiologists with varying levels of Bulla S et  al (2010) Dual-energy CT virtual noncal-
experience in correlation to MR imaging. Radiology cium technique: detecting posttraumatic bone marrow
280(2):510–519 lesions—feasibility study. Radiology 256(2):617–624
Kelcz F, Joseph PM, Hilal SK (1979) Noise considerations Petritsch B, Kosmala A, Weng AM, Krauss B, Heidemeier
in dual energy CT scanning. Med Phys 6(5):418–425 A, Wagner R et al (2017) Vertebral compression frac-
Koch V, Müller FC, Gosvig K, Albrecht MH, Yel I, Lenga tures: third-generation dual-energy CT for detection of
L et al (2021) Incremental diagnostic value of color-­ bone marrow edema at visual and quantitative analy-
coded virtual non-calcium dual-energy CT for the ses. Radiology 284(1):161–168
assessment of traumatic bone marrow edema of the Primak A, Ramirez Giraldo J, Liu X, Yu L, McCollough
scaphoid. Eur Radiol 31(7):4428–4437 CH (2009) Improved dual-energy material discrimina-
Kosmala A, Weng AM, Heidemeier A, Krauss B, Knop tion for dual-source CT by means of additional spec-
S, Bley TA et al (2017) Multiple myeloma and dual-­ tral filtration. Med Phys 36(4):1359–1369
energy CT: diagnostic accuracy of virtual noncalcium Roski F, Hammel J, Mei K, Baum T, Kirschke JS,
technique for detection of bone marrow infiltration of Laugerette A et al (2019) Bone mineral density mea-
the spine and pelvis. Radiology 286(1):205–213 surements derived from dual-layer spectral CT enable
Lee M-J, Kim S, Lee S-A, Song H-T, Huh Y-M, Kim D-H opportunistic screening for osteoporosis. Eur Radiol
et al (2007) Overcoming artifacts from metallic ortho- 29(11):6355–6363
pedic implants at high-field-strength MR imaging and Suh J-S, Jeong E-K, Shin K-H, Cho JH, Na JB, Kim DH
multi-detector CT. Radiographics 27(3):791–803 et  al (1998) Minimizing artifacts caused by metallic
Leng S, Yu Z, Halaweish A, Kappler S, Hahn K, Henning implants at MR imaging: experimental and clinical
A et  al (2016) A high-resolution imaging technique studies. AJR Am J Roentgenol 171(5):1207–1213
using a whole-body, research photon counting detec- Suh CH, Yun SJ, Jin W, Lee SH, Park SY, Ryu C-W (2018)
tor CT system. In: Medical imaging 2016: physics of Diagnostic performance of dual-energy CT for the
medical imaging: international society for optics and detection of bone marrow oedema: a systematic review
photonics, p 97831I and meta-analysis. Eur Radiol 28(10):4182–4194
Li X, Li J, Jiao X, Jia X, Zhang X, Fan G et  al (2020) van Hamersvelt RW, Schilham AM, Engelke K, den
The accuracy of bone mineral density measurement Harder AM, de Keizer B, Verhaar HJ et  al (2017)
using dual-energy spectral CT and quantitative CT: Accuracy of bone mineral density quantification using
a comparative phantom study. Clin Radiol 75(4):320. dual-layer spectral detector CT: a phantom study. Eur
e9–320.e15 Radiol 27(10):4351–4359
Lüdeke K, Röschmann P, Tischler R (1985) Susceptibility Viano AM, Gronemeyer SA, Haliloglu M, Hoffer FA
artefacts in NMR imaging. Magn Reson Imaging (2000) Improved MR imaging for patients with metal-
3(4):329–343 lic implants☆. Magn Reson Imaging 18(3):287–295
Mallinson PI, Coupal TM, McLaughlin PD, Nicolaou S, Wang C-K, Tsai J-M, Chuang M-T, Wang M-T, Huang
Munk PL, Ouellette HA (2016) Dual-energy CT for the K-Y, Lin R-M (2013) Bone marrow edema in verte-
musculoskeletal system. Radiology 281(3):690–707 bral compression fractures: detection with dual-energy
Meinel FG, Bischoff B, Zhang Q, Bamberg F, Reiser MF, CT. Radiology 269(2):525–533
Johnson TR (2012) Metal artifact reduction by dual-­ Wang Q, Sun Z, Li S, Zhang H, Li J, Zhang L et  al
energy computed tomography using energetic extrap- (2017) Bone marrow imaging by third-generation
olation: a systematically optimized protocol. Investig dual-source dual-energy CT using virtual noncalcium
Radiol 47(7):406–414 technique for assessment of diffuse infiltrative lesions
Müller FC, Børgesen H, Gosvig K, Rodell A, Booz C, of multiple myeloma. Acta Acade Med Sin 39(1):
Schmidt B et  al (2019) Optimising dual-energy CT 114–119
scan parameters for virtual non-calcium imaging of Wang M-Y, Zhang X-Y, Xu L, Feng Y, Xu Y-C, Qi L et al
the bone marrow: a phantom study. Eur Radiol Exp (2019) Detection of bone marrow oedema in knee
3(1):1–8 joints using a dual-energy CT virtual non-calcium
Nasirudin RA, Mei K, Panchev P, Fehringer A, Pfeiffer technique. Clin Radiol 74(10):815.e1–815.e7
F, Rummeny EJ et al (2015) Reduction of metal arti- Wesarg S, Kirschner M, Becker M, Erdt M, Kafchitsas
fact in single photon-counting computed tomography K, Khan M (2012) Dual-energy CT-based assessment
by spectral-driven iterative reconstruction technique. of the trabecular bone in vertebrae. Methods Inf Med
PLoS One 10(5):e0124831 51(5):398
Skeletal Imaging: Bones 313

White LM, Kim JK, Mehta M, Merchant N, Schweitzer noncalcium CT in the detection of bone marrow edema
ME, Morrison WB et al (2000) Complications of total in the sacroiliac joints. Radiology 290(1):157–164
hip arthroplasty: MR imaging—initial experience. Yang SJ, Jeon JY, Lee S-W, Jeong YM (2020) Added
Radiology 215(1):254–262 value of color-coded virtual non-calcium dual-energy
Wichmann JL, Booz C, Wesarg S, Kafchitsas K, Bauer CT in the detection of acute knee fractures in non-­
RW, Kerl JM et  al (2014) Dual-energy cT–based radiology inexpert readers. Eur J Radiol 129:109112
phantomless in  vivo three-dimensional bone mineral Yu L, Li H, Mueller J, Kofler JM, Liu X, Primak AN
density assessment of the lumbar spine. Radiology et  al (2009) Metal artifact reduction from reformat-
271(3):778–784 ted projections for hip prostheses in multislice helical
Wichmann JL, Booz C, Wesarg S, Bauer RW, Kerl JM, computed tomography: techniques and initial clinical
Fischer S et  al (2015) Quantitative dual-energy CT results. Investig Radiol 44(11):691
for phantomless evaluation of cancellous bone min- Zhou C, Zhao YE, Luo S, Shi H, Zheng L, Zhang LJ
eral density of the vertebral pedicle: correlation et  al (2011) Monoenergetic imaging of dual-energy
with pedicle screw pull-out strength. Eur Radiol CT reduces artifacts from implanted metal ortho-
25(6):1714–1720 pedic devices in patients with factures. Acad Radiol
Willemink MJ, Persson M, Pourmorteza A, Pelc NJ, 18(10):1252–1257
Fleischmann D (2018) Photon-counting CT: tech- Zhou S, Zhu L, You T, Li P, Shen H, He Y et al (2021) In
nical principles and clinical prospects. Radiology vivo quantification of bone mineral density of lumbar
289(2):293–312 vertebrae using fast kVp switching dual-energy CT:
Wu H, Zhang G, Shi L, Li X, Chen M, Huang X et  al correlation with quantitative computed tomography.
(2019) Axial spondyloarthritis: dual-energy virtual Quant Imaging Med Surg 11(1):341
 Gout

Torsten Diekhoff

Contents
1 Pathogenesis and Clinical Presentation   316
2 Diagnostic Options   316
3 Dual-Energy CT   317
3.1   hysics: Two-Material Decomposition 
P  317
3.2  Clinical Indications for DECT   318
3.3  Scan Protocols   319
3.4  Contrast Media   320
3.5  Image Reconstruction   321
3.6  Image Interpretation   322
3.7  Report: What Information to Include   322
3.8  Issues Regarding Tophus Density and Radiation Exposure   323
3.9  DECT for Follow-Up   323
4 Case Discussions   324
4.1   ase 1: Patient with Acute Wrist Arthritis 
C  324
4.2  Case 2: Chronic Pain of the Feet   325
4.3  Case 3: Pain of the First MTP Joint   325
4.4  Case 4: Patient with Knee Pain   326
4.5  Case 5: Patient with Rheumatoid Arthritis   328
5 Summary   329
References   329

Abstract development of soft tissue tophi of uric acid,

which can specifically be detected by
Imaging of gouty arthritis was one of the first DECT. The highly innovative imaging modal-
clinical indications for dual-energy computed ity is in continuous competition with estab-
tomography (DECT). The course of this auto- lished tests like arthrocentesis or
inflammatory disease is characterized by the ultrasonography for advanced, atypical cases
of gout. Here, DECT benefits from its highly
T. Diekhoff (*) standardized scanning and evaluation process
Department of Radiology, Charité - and at the same time provides additional infor-
Universitätsmedizin Berlin, Berlin, Germany mation from conventional CT and virtual non-­
e-mail: [email protected]

© The Author(s), under exclusive license to Springer Nature Switzerland AG 2022 315
H. Alkadhi et al. (eds.), Spectral Imaging, Medical Radiology Diagnostic Imaging,
https://doi.org/10.1007/978-3-030-96285-2_22
316 T. Diekhoff

calcium images. While it is less sensitive for Unlike autoimmune diseases such as rheuma-
very early gouty arthritis, it has gained an toid arthritis, gout is a so-called autoinflamma-
important role in the diagnostic process and tion with a typical immune response (Ziegeler
also has a unique ability to assess the disease et al. 2020b). Patients experience sudden attacks
burden and therapy response. This chapter of pain with redness, swelling, and stiffness that
summarizes the pathogenesis of gouty arthri- decline within 7–10 days. Changing attack loca-
tis and the clinical indications, protocols, and tions with asymptomatic intervals are common,
image interpretation of DECT in these patients while later stages are characterized by polyartic-
and provides clinical examples to illustrate its ular manifestations without complete symptom
different applications. relief between acute episodes (Parathithasan
et al. 2016). Gouty tophi can manifest in several
locations in the joint capsule, surrounding soft
tissues and regions with increased mechanical
1 Pathogenesis and Clinical stress such as entheses, ligaments, bursae or ear
Presentation helix (Wright and Pinto 2003). Most commonly
affected are the joints of the lower extremities,
- or: why birds suffer from gout but not especially the first metatarsophalangeal joint.
dolphins.
Uric acid is the end product of purine meta-
bolic pathways in the human body and usually 2 Diagnostic Options
eliminated by the kidneys. However, its solubility
limit is far lower than that of its precursors, hypo- In the vast majority of cases, no advanced diag-
xanthine and xanthine (Schett et  al. 2015). nostic testing is necessary. When a patient pres-
Species like birds or humans that cannot process ents with typical clinical signs and symptoms
uric acid to the more soluble allantoin are at risk (podagra), at most a laboratory test for determi-
of uric acid crystal precipitation within joints and nation of uric acid in serum is indicated
soft tissues. Interestingly, also early ancestors of (Underwood 2006). However, as gouty arthritis is
birds shared this condition, and some fossilized so common, many patients have uncharacteristic
Tyrannosaurus bones show typical gouty ero- clinical presentations. In such cases of suspected
sions (Rothschild et al. 1997). but unproven gout, arthrocentesis is the standard
Gout is the most common inflammatory joint of reference for diagnosis. Here, polarization
disease in the western world with a prevalence of microscopy depicts the birefringent crystals in
1–4% (Kuo et al. 2015) and has been increasing the joint fluid (see Fig. 1) (Phelps and McCarty
over recent years (Zhu et  al. 2011). Men are Jr. 1966). However, in clinical practice, the sensi-
more often affected than women. However, post- tivity of this test is reduced in symptom-free
menopausal women may sometimes develop intervals, and rheumatologists use aspiration cau-
gout and then suffer even more severely. A tiously, fearing its complications (Neogi et  al.
genetic predisposition, cultural habits, and kid- 2015; Taylor et al. 2016). For this reason, imag-
ney function play a role in the development of ing plays an essential role in the diagnosis of
the disease as well as purine-rich food, alcohol, atypical forms.
and diet sodas (Choi et  al. 2004a, b; Choi and Imaging in gouty arthritis has three aims: to
Curhan 2008). All of these aspects should be establish the diagnosis by demonstrating typical
taken into account when treating gouty arthritis. bone changes or uric acid depositions and to
Gout is a systemic disease with several meta- exclude differential diagnoses, to assess the joint
bolic and cardiovascular comorbidities (Zhu for subsequent arthrocentesis, and to monitor the
et  al. 2012). Therefore, early detection of its response to treatment. To ascertain the diagnosis,
musculoskeletal manifestations is crucial for a usually, a high-resolution cross-sectional imag-
swift treatment initiation. ing method, namely arthrosonography or dual-­
Gout 317

Fig. 1  Microscopic views of uric acid crystals. The needle-shaped crystals in conventional microscopy (left) are nega-
tively birefringent in polarization microscopy (right). The material was obtained by surgical tophus excision

energy computed tomography (DECT), is 3 Dual-Energy CT

needed. Only advanced stages show typical
“punched-out” erosions, sometimes more distant CT vendors have developed several methods to
from the joint, and periosteal reactions and faint perform dual-energy or spectral imaging that can
gouty tophi in radiography (Buckley 1996). be used to image gouty arthritis patients. The
However, radiography can depict differential direct demonstration of uric acid tophi with high
diagnoses such as osteoarthritis or calcium pyro- sensitivity and specificity and the excellent repro-
phosphate deposition disease. It is also a readily ducibility with standardized protocols favor
available tool for follow-up examinations. DECT over other imaging methods in gout and
Arthrosonography is well-suited to detect ear- earned it the first indication in clinical guidelines
lier stages of gouty arthritis by directly demonstrat- (Neogi et al. 2015). While the detection of tophi
ing uric acid crystals in the joint fluid (so-called with a two-material decomposition algorithm is
double-contour sign) or tophi (Thiele and at the core of DECT imaging in gout, it provides
Schlesinger 2007). It can also be used for planning further information about bone destruction and
a joint aspiration or to guide the needle during this even active inflammation, when applied
intervention (Grassi et  al. 2001). Finally, arthro- correctly.
sonography can also demonstrate acute arthritic
inflammation, which, however, is not specific to
gouty arthritis (Chowalloor and Keen 2013). 3.1 Physics: Two-Material
Magnetic resonance imaging (MRI) can also Decomposition
demonstrate active inflammation in the course of
gouty arthritis. Nevertheless, these changes do DECT allows the differentiation of two objects
not help in distinguishing gout from other inflam- with similar X-ray attenuation (e.g., calcium
matory joint diseases (Cimmino et al. 2011). As phosphate and uric acid) when their effective
conventional MR sequences cannot depict the atomic number Zeff is sufficiently different. To
bony surface directly, it is less well suited to accomplish this, information about the attenua-
assess erosions and their morphology than ultra- tion of high and low energy X-rays is needed.
sound and CT (Ulas, Diekhoff et al. 2019). It also Thus, most CT machines apply two different
fails to demonstrate uric acid depositions and can X-ray spectra (tube voltages) and compare the
only detect large tophi. Therefore, it is inferior to attenuation. While materials with high Zeff (e.g.,
DECT when it comes to establishing the diagno- calcium or iodine) exhibit greater attenuation at
sis (Schumacher et al. 2006). low energy compared to high energy scans, this
318 T. Diekhoff

High energy attenuation

Cortical bone

Uric acid

Th
re
sh Trabecular bone
ol
d

Soft tissues
Low energy attenuation

Fig. 2  Principles of two-material decomposition. The above the line is characterized as uric acid tophus. At the
software distinguishes two materials, e.g., bone and uric same time, the software tool takes other parameters such
acid, according to their dual-energy index by applying a as proximity to bone into account
detection threshold. Every voxel above the threshold and

effect becomes less pronounced with decreasing but is controversial as some patients with ele-
Zeff and is reversed for molecules with very small vated levels will never develop gout and it can be
atoms, such as hydrogen and carbon. This behav- normal even during an acute attack (Campion
ior is measured by the dual-energy index or gra- et  al. 1987; Rymal and Rizzolo 2014). While
dient, which is relatively specific for a certain arthrocentesis is the diagnostic gold standard, it
material and, thus, allows its distinction from is often not performed in clinical practice.
others, even if they have the same attenuation in DECT is reported to show high sensitivity and
conventional CT (Diekhoff et al. 2015). specificity in the search for gouty arthritis. It also
DECT in gout utilizes the relatively low Zeff of provides all information of a conventional CT
uric acid together with its relatively high CT den- scan such as erosions or other soft tissue calcifi-
sity when present in sufficiently high concentra- cations and is, therefore, well-suited for differen-
tion to specifically detect tophi. Figure  2 tial diagnoses. However, DECT has limited
illustrates the postprocessing algorithm graphi- sensitivity in early stages of the disease as it
cally. However, as soft tissues such as muscles relies on the presence of tophi (Bongartz et  al.
and ligaments show similar Zeff but lower density, 2015). On the other hand, tophi have been
a threshold must be established to eliminate reported in patients with asymptomatic hyperuri-
false-positive detection. Nonetheless, this thresh- cemia (Wang et al. 2018), pointing out that ear-
old decreases the sensitivity for early tophi with lier stages of gouty arthritis (beginning with
low uric acid concentrations (see Fig. 3). hyperuricemia and acute arthritis in contrast to
established tophaceous disease) might not reflect
the clinical reality, and a new system is warranted
3.2 Clinical Indications for DECT (Dalbeth and Stamp 2014).
Studies show that joint aspiration can prove
Whenever atypical gouty arthritis is suspected in gout in DECT-negative patients but might also be
a patient, choosing the proper diagnostic test is negative, for several reasons, when DECT dem-
key to establish the diagnosis. The measurement onstrates tophi (Notzel et  al. 2018). Therefore,
of serum uric acid can point in the right direction when available, the less invasive test should be
Gout 319

Fig. 3  Threshold adjustment. A patient with tophaceous more and larger tophi are depicted (arrows), there are also
gouty arthritis underwent DECT of the right foot. The more artifacts, leading to false-positive detection in toe
same scan was reconstructed applying a conventional nails and tendons (arrowheads)
threshold (left) and a lowered threshold (right). While

performed first and, if negative, followed by joint 3.3 Scan Protocols

aspiration if the diagnosis is still suspected.
While its diagnostic accuracy is similar to that of Details of the DECT protocol depend on the ven-
arthrosonography, DECT might be less sensitive dor, the patient and the joint. Modern machines
in early disease (Shang et al. 2020). Ultrasound, support automated tube current settings and
on the other hand, can capture some findings ­modulation during the scan to optimize image
(e.g., double-contour sign or synovitis) that are quality and radiation exposure. For systems that
hidden from DECT but is hard to standardize allow individual adjustment of tube currents in
and, therefore, less suited for follow-up examina- high and low energy scans, a similar image qual-
tions. To date, ultrasonography and DECT can be ity of both datasets is desirable—usually the ratio
regarded as equivalent for the diagnosis and of mAs between high and low energy spectra is
should be used according to their availability and best between 1:2 and 1:4 for gout imaging. The
the experience of the interpreter (Ramon et  al. tin filter, which is available for some systems,
2018). While arthrosonography and DECT are optimizes the X-ray spectrum and should be
still regarded as second-line imaging after radi- applied whenever possible.
ography, both are gaining in importance and can It is well known that in patients with gouty
establish the diagnosis also in earlier disease. arthritis joints can be painful and swollen but
320 T. Diekhoff

exhibit no proof of tophi in imaging whereas the arm elevated and sideways flexion of the
asymptomatic joints show definite positive find- head might be an alternative. Under no circum-
ings. Therefore, the regions to be scanned need stances should hands or elbows be scanned rest-
careful consideration. In clinical practice, a pro- ing on the abdomen or besides the patient, as
tocol has become established that covers the cur- this will result in unnecessarily high radiation
rently painful joints and both feet and ankles, as exposure and a severe loss of image quality.
they are most commonly affected (Huppertz et al. Furthermore, it is crucial never to place the
2014). This approach offers the highest chance of joint in a way that the whole forearm is in the
finding tophi. beam path.
The lower extremities can be scanned in There are only a few reports on gout imaging
supine position. The knees can easily be imaged of the axial skeleton (Gibney and Murray 2020).
in extension or light (10–20°) flexion. Both Here, the protocol should use a sufficiently large
ankles and forefeet are usually imaged in one amount of radiation to ensure optimal image
scan in firm plantar flexion to minimize artifacts. quality. Especially the low energy scan is critical
To accomplish this, flexion of the knee, supported as increased image noise might lead to more
by a bolster, can be helpful so that the patient can false-positive detections. As DECT cannot use
rest his or her soles on the table. Positioning is conventional methods to cope with obese patients
less complex with the patient prone and the (e.g., use of higher tube voltage), an optimal
arches of the feet resting on the table. For all image quality can only be achieved with an
scans, it is useful to place the legs as close as pos- increase in tube current and rotation time.
sible together and in the middle of the gantry to
ensure high resolution through a small field of
view and to reduce possible artifacts. In some 3.4 Contrast Media
cases, it can be appropriate to image only one
foot. Then, the other leg should be placed outside In most patients with gouty arthritis, imaging can
the beam path, e.g., by bending the knee as much be performed without administration of contrast
as possible. medium. However, recent studies suggest that
For the upper extremities, the patient is usu- contrast-enhanced CT (Diekhoff, Ulas et  al.
ally in prone position with the arms elevated 2019) or DECT (Fukuda et  al. 2017) depicts
over the head. Both hands and wrists can be acute soft tissue inflammation and, therefore,
scanned simultaneously or separately. Note that enhances the diagnostic capability of a DECT
a simultaneous scan in a larger field of view scan (see Fig.  4). Nonetheless, it is unclear
will decrease the spatial resolution, which is whether tophus enhancement might distort the
not of advantage in the small structures of the effective Z of a voxel sufficiently to interfere with
frist and finger joints. With some machines, it is two-material decomposition. Therefore, contrast
possible to image one hand/wrist comfortably medium should be used with caution.
with the patient standing beside the table or In the search for active soft tissue inflamma-
behind the gantry. If this maneuver includes a tion like synovitis or enthesitis, the iodinated
table movement, a secure rest of the arm on the contrast medium should be administered at a
table is desirable and can be achieved by using ­volume of 1 ml/kg of body weight followed by a
a sandbag or straps. The elbow is sometimes sufficient amount of saline solution. Three min-
harder to image. Machines with enough z-axis utes after contrast medium injection, inflamma-
coverage allow the scanning without table tory soft tissue already shows enhancement while
movement and a patient standing behind the contrast medium diffusion into joint effusion is
gantry and reaching into the scanner. For other still minimal. Therefore, this is the best timepoint
systems, prone positioning of the patient with for imaging.
Gout 321

Fig. 4  Synovitis in DECT iodine maps after contrast inflammation. Synovitis is confirmed by arthrosonogra-
medium injection. The iodine map (left) demonstrates phy (right), which shows hypoechoic soft tissue swelling
contrast enhancement at the wrist and the second metacar- and increased perfusion in the power Doppler mode. The
pophalangeal (MCP) joint, indicating acute soft tissue patient was later diagnosed with rheumatoid arthritis

3.5 Image Reconstruction Depending on the particular settings of the CT

machine, secondary reconstruction of the gout
It is recommended to reconstruct standard, con- images using two-material decomposition and
ventional CT images in a sharp bone kernel and virtual non-calcium (VNCa) images using three-­
medium soft tissue kernel and to generate appro- material decomposition is done automatically, on
priate multiplanar reformations. Either the CT console, a stand-alone workstation or a
70–75 keV monochromatic or 120 kV-equivalent cloud-based software. Examples are given in
weighted average images come closest to a stan- Fig. 5. For gout images, some software applica-
dard CT.  When no automated reconstruction of tions have an option to eliminate certain artifacts,
these images is possible, reconstruction of the e.g., from beam hardening or toe nails, directly
base scans is also fine. Using the high energy during the reconstruction process. This is espe-
scan for soft tissue reconstructions results in bet- cially important when measuring and reporting
ter visualization of gouty tophi (as uric acid the total tophus volume. Whenever possible, iter-
attenuates slightly more in high energy scans) ative reconstruction or artificial intelligence
and low image noise. Bone kernel reconstruc- reconstruction that reduces the image noise
tions can be done from either the low energy scan should be applied (Diekhoff et al. 2018).
(which will result in better contrast at the expense When DECT is performed with contrast
of image noise) or the high energy scan. The lat- medium administration additional iodine maps
ter might be preferable for the axial skeleton. will depict regions of soft tissue enhancement.
322 T. Diekhoff

Fig. 5  Reconstructions from DECT. Bone kernel recon- well depicted in bone kernel reconstructions (arrow) com-
struction depicts erosions and other bone pathologies. pared to the other images. Tophi are not subtracted in
Soft tissue kernel reconstruction shows gouty tophi. VNCa images, occasionally resulting in their superior
VNCa reconstructions are used to search for bone marrow visualization compared to soft tissue or uric acid images
lesions and two-material decomposition uric acid images (arrowheads)
are used to identify gouty tophi. Note that tophi are less

3.6 Image Interpretation tration or secondary calcification (see Fig.  6)

(Lee et  al. 2019). Furthermore, morphology of
The interpretation of color-coded images gener- the tophi is important and careful consideration
ated with the two-material decomposition algo- can help to prevent false-positive findings
rithm, highlighting gouty tophi in green or red, is (Ziegeler et  al. 2020a). Especially in gout-­
fairly straightforward (Huppertz et al. 2014). The negative cases, CT can confirm or rule out dif-
unequivocal presence of tophi in a typical loca- ferential diagnoses such as osteoarthritis, calcium
tion confirms the diagnosis with high specificity pyrophosphate deposition disease or septic
(Choi et al. 2009). Nevertheless, especially small arthritis.
tophi should be confirmed in standard CT images VNCa images provide useful additional infor-
and typical artifacts should be considered. mation. In case of severe inflammation, they
Artifacts include physiological soft tissues with delineate the underlying osteitis (Jans et  al.
high density, e.g., toe nails or calluses at the heel, 2018). When gout cannot be proven with DECT,
atherosclerotic plaques, or tiny calcifications the presence of bone marrow lesions can identify
(Choi et al. 2012; Christiansen et al. 2020). Beam joints for further diagnostic examination such as
hardening on the bone surface can also lead to arthrocentesis or ultrasonography. Notably,
false-positive detection. However, those are easy VNCa reconstructions do not subtract the gouty
to identify and there are some robust software tophi so that they become even more conspicuous
solutions for this problem. Very tiny tophi of only compared to standard CT images (see Fig. 5).
a few voxels in size within ligaments or menisci
should not be overestimated.
Conventional CT reconstructions require care- 3.7 Report: What Information
ful attention. They contain the most information to Include
and are assessed for erosions and soft tissue
depositions. Depending on the concentration of It is not easy to make recommendations on what
uric acid, gouty tophi can be visible in conven- a radiology report should contain as findings and
tional CT images while they are not highlighted structure have to consider the needs of the radi-
in gout images due to their low uric acid concen- ologist and referring physicians. However, there
Gout 323

Fig. 6  False-negative tophus. A patient with suspected 135 kVp source data) depicts faint uric acid depositions
gouty arthritis shows no tophus after two-material decom- (arrows) that are confirmed by ultrasonography
position in DECT.  The conventional CT image (here

is a certain minimum of information that should detection of gouty tophi and their volume will
be provided. The report should state concisely improve with increasing radiation exposure and
and clearly whether the scan confirmed the sus- the use of noise-reducing software and recon-
pected diagnosis of gouty arthritis by unequivo- structions. The lower the threshold, the lower the
cal depiction of uric acid depositions, is detectable concentrations of uric acid (see Fig. 3).
compatible with gout (i.e., changes are sugges- With established DECT protocols currently in
tive but tophus is not definitely confirmed), is use, uric acid concentrations between 30% and
normal or a differential diagnosis is more likely. 40% can be detected (Diekhoff et al. 2018).
The tophus volume is an important marker for the
disease and the confidence of findings and should
be included. Also worth mentioning are erosions 3.9 DECT for Follow-Up
and other arthritis findings as well as active
inflammatory bone marrow lesions detected in DECT offers the unique option to quantify tophus
VNCa reconstructions. volume and, thus, the disease burden of a patient.
This is not only helpful in assessing the severity
of gout in an interindividual comparison but also
3.8 Issues Regarding Tophus allows intraindividual follow-up after initiation
Density and Radiation of dietary measures or treatment (see Fig. 7) (Sun
Exposure et  al. 2015). Interestingly, the development of
tophi and recurrence under therapy is not well
The analysis threshold selected for the two-­ understood. For example, it has been reported
material decomposition algorithm depends on the that, while treatment induces tophus resorption,
density of the soft tissues that might cause false-­ new tophi form on different joints or in different
positive detections (e.g., tendons) and image sites (Zhang et al. 2017). Also, measurement of
noise. Therefore, the sensitivity of DECT for the uric acid volume might not fully capture the dis-
324 T. Diekhoff

Fig. 7  Follow-up of a
patient with gouty
arthritis. A patient with
gouty tophi underwent
DECT for primary
diagnosis before starting
treatment. Three weeks
later, he presented with
worsening of pain and
treatment failure was
suspected. Whereas the
three-dimensional
reconstructions of the
two scans do not show
an unequivocal tophus
reduction, the volume
measurements and mass
calculations confirm a
treatment response.
Treatment was
continued and additional
analgesics administered

ease burden as tophi might vary in uric acid con- triangular fibrocartilage complex were seen.
centration (Kotlyarov et  al. 2020). Therefore, Clinically, septic arthritis, acute pseudogout,
further advances in software development and and gout were possible differential diagnoses.
clinical studies are needed before DECT can be DECT confirmed the presence of uric acid
recommended for regular follow-up in gout tophi, and a corresponding treatment was
patients. started. However, after 6 weeks, there was only
mild resolution of symptoms. A follow-up
DECT scan demonstrated an increase in uric
4 Case Discussions acid depositions and a switch in medical treat-
ment was indicated. The images are presented
4.1  ase 1: Patient with Acute
C in Fig. 8.
Wrist Arthritis
Conclusion  MRI often fails to demonstrate
A 70-year-old male patient presented to the gouty tophi and cannot differentiate between uric
emergency department with acute onset of acid and calcium pyrophosphate depositions.
severe wrist pain, redness, and swelling. MRI DECT is a noninvasive tool to establish the diag-
showed severe inflammation with synovitis, nosis and its quantitative measurements can be
peritendonitis, and tenosynovitis of the wrist. used for follow-up and monitoring the response
In addition, some hypointense structures at the to treatment.
Gout 325

Fig. 8 70-year-old patient with wrist arthritis. MRI material decomposition (arrows). Follow-up demonstrates
shows severe inflammation with synovitis of the wrist but progression of the depositions (arrows) that was quanti-
misses the uric acid tophi (arrow). CT confirms the soft fied by volumetric measurements. Also, other joints such
tissue depositions that are characterized as gout by two-­ as knee and ankles showed a progress of tophi

Conclusion  Standard CT images are important

4.2  ase 2: Chronic Pain
C
to establish the diagnosis in the absence of uric
of the Feet
acid depositions. They display structural lesions
and gouty tophi that do not reach the threshold in
An 82-year-old female patient suffered from
two-material decomposition.
occasional pain at the hands and feet that had per-
sisted for several years. Her family physician sus-
pected gouty arthritis and treated correspondingly.
As she still suffered several episodes of mild to 4.3  ase 3: Pain of the First MTP
C
moderate pain every year despite drug treatment Joint
and diet, she presented to a rheumatologist. The
DECT scan shows no uric acid depositions A 41-year-old male patient presents in the outpa-
besides a typical toe nail artifact. However, the tient rheumatology clinic with redness and swell-
standard CT reconstructions prove erosions on ing of the first metatarsophalangeal joint on the
the first metatarsal head with sclerotic rims and left side. He reports a one-year history with sev-
overhanging edges, typical of gouty arthritis. eral episodes of mild to moderate pain with exac-
Images are shown in Fig. 9. erbation a few days before presentation. He had
326 T. Diekhoff

Fig. 9  Gouty arthritis without tophi. In this chronic case head. However, two-material decomposition fails to dem-
of gouty arthritis, standard CT reconstructions in bone onstrate tophi. A false-positive toe nail artifact is present
kernel show typical erosions (arrow) at the first metatarsal (arrowhead)

no history of joint disease and was otherwise Conclusion  DECT may fail to demonstrate uric
healthy as well. acid depositions in very early gouty arthritis.
Radiography was already performed by the However, it provides crucial information on
referring physician and fairly unremarkable. active inflammation, even when performed with-
The rheumatologist suspected gouty arthritis; out contrast medium. Therefore, it can pave the
however, the diagnostic criteria were not ful- way to further diagnostic tests or suggest other
filled and further proof was needed to establish possible differential diagnoses.
the diagnosis. Therefore, a DECT scan was
ordered.
DECT did not show any tophus formation, 4.4 Case 4: Patient with Knee Pain
and the CT source images were unremarkable.
However, VNCa images showed acute bone mar- A 52-year-old construction worker was referred to
row lesions (osteitis) of the first metatarsophalan- undergo an MRI examination for severe knee pain.
geal joint, indicating severe inflammation. Hence, He had complaints for several years but was not
an additional arthrosonography was performed, able to schedule an appointment because he was
which confirmed synovitis and gouty arthritis by traveling to construction sites. MRI showed inflam-
showing a double-contour sign. The images are matory arthritis with soft tissue depositions, highly
presented in Fig. 10. suggestive of severe gouty arthritis. A DECT scan
Gout 327

Fig. 10  41-year-old patient with pain. DECT and CT do tralateral side. The diagnosis of gouty arthritis was later
not show any uric acid deposition. VNCa reconstructions confirmed by ultrasonography (US) with demonstration
from the DECT scan, however, reveal active inflammation of active synovitis (arrowheads) and a double-contour
on the first MTP joint (arrows) when compared to the con- sign (arrow)

was obtained and demonstrated gouty tophi at Conclusion  In severe cases, DECT is not needed
knees, feet, ankles, and elbows (see Fig. 11). The to establish the diagnosis. However, it confirms
DECT scan was completed in less than 15 minutes uric acid tophi with high specificity, gives an
with a total radiation exposure of less than 1 mSv excellent overview of the disease burden, and is
and gave an easy overview of the disease burden. faster than other imaging methods.
328 T. Diekhoff

Fig. 11  Severe gouty arthritis. MRI already shows soft in this patient, it provided an excellent overview of the
tissue depositions, consistent with advanced gouty arthri- total disease burden. Here, the software color-codes left
tis. While DECT was not needed to establish the diagnosis and right side differently

4.5  ase 5: Patient

C showed unequivocal tophi, the bursa olecrani was
with Rheumatoid Arthritis filled with calcified and uric-acid-positive struc-
tures (see Fig. 12). Surgery confirmed the pres-
A 52-year-old physician treated himself for sev- ence of severe gouty arthritis.
eral years for rheumatoid arthritis. For some
time, he sensed several rheumatic nodes on both Conclusion  Calcified tophi might be misclassi-
elbows that worsened over the last months. A fied by DECT. In those cases, DECT can be false-­
rheumatologist to whom he finally presented negative and underestimate the total amount of
ordered a DECT scan. While several joints uric acid.
Gout 329

Fig. 12  Calcified gouty tophi within a bursa. Only the faint characteristic tophi (arrows) are correctly characterized.
The more dense, calcified tophi are false-negative (arrowheads)

5 Summary tis, DECT has earned its prominent place in cur-

rent rheumatology guidelines.
DECT is a robust, sensitive, and specific tool to
assess gouty arthritis and its differential diagno- Compliance with Ethical Standards
ses. When used correctly and with consideration
of artifacts and possible false-positive and false-­ Funding  not applicable.
negative findings, it can establish the diagnosis,
elucidate differential diagnoses, and monitor Disclosure of Interests  The author has received a speaker
treatment. The field of DECT in gouty arthritis honorarium from Canon Medical Systems, Novarits and
MSD. The author’s institution received research grants
has undergone some changes during the last from Canon Medical Systems.
years. While previously only color-coded uric
acid tophi were considered to be relevant, more
recently, grayscale conventional CT images and
References
VNCa reconstructions have found their way into
clinical practice. Furthermore, the benefit of con- Bongartz T, Glazebrook KN, Kavros SJ, Murthy NS, Merry
trast medium administration is currently being SP, Franz WB 3rd, Michet CJ, Veetil BM, Davis JM 3rd,
investigated and may be useful in identifying soft Mason TG 2nd, Warrington KJ, Ytterberg SR, Matteson
EL, Crowson CS, Leng S, McCollough CH (2015) Dual-
tissue inflammation. Providing a multiparametric energy CT for the diagnosis of gout: an accuracy and
overview in patients with suspected gouty arthri- diagnostic yield study. Ann Rheum Dis 74(6):1072–1077
330 T. Diekhoff

Buckley TJ (1996) Radiologic features of gout. Am Fam delineating inflammation of inflammatory arthritis.
Physician 54(4):1232–1238 Eur Radiol 27(12):5034–5040
Campion EW, Glynn RJ, DeLabry LO (1987) Gibney B, Murray N (2020) Dual-energy CT of spinal
Asymptomatic hyperuricemia. Risks and conse- tophaceous gout. Radiology 296(2):276
quences in the Normative Aging Study. Am J Med Grassi W, Farina A, Filippucci E, Cervini C (2001)
82(3):421–426 Sonographically guided procedures in rheumatology.
Choi HK, Curhan G (2008) Soft drinks, fructose con- Semin Arthritis Rheum 30(5):347–353
sumption, and the risk of gout in men: prospective Huppertz A, Hermann KG, Diekhoff T, Wagner M,
cohort study. BMJ 336(7639):309–312 Hamm B, Schmidt WA (2014) Systemic staging for
Choi HK, Atkinson K, Karlson EW, Willett W, urate crystal deposits with dual-energy CT and ultra-
Curhan G (2004a) Alcohol intake and risk of inci- sound in patients with suspected gout. Rheumatol Int
dent gout in men: a prospective study. Lancet 34(6):763–771
363(9417):1277–1281 Jans L, De Kock I, Herregods N, Verstraete K, Van den
Choi HK, Atkinson K, Karlson EW, Willett W, Curhan Bosch F, Carron P, Oei EH, Elewaut D, Jacques P
G (2004b) Purine-rich foods, dairy and protein (2018) Dual-energy CT: a new imaging modality for
intake, and the risk of gout in men. N Engl J Med bone marrow oedema in rheumatoid arthritis. Ann
350(11):1093–1103 Rheum Dis 77(6):958–960
Choi HK, Al-Arfaj AM, Eftekhari A, Munk PL, Shojania Kotlyarov M, Hermann KGA, Mews J, Hamm B,
K, Reid G, Nicolaou S (2009) Dual energy computed Diekhoff T (2020) Development and validation of a
tomography in tophaceous gout. Ann Rheum Dis quantitative method for estimation of the urate burden
68(10):1609–1612 in patients with gouty arthritis using dual-energy com-
Choi HK, Burns LC, Shojania K, Koenig N, Reid G, puted tomography. Eur Radiol 30(1):404–412
Abufayyah M, Law G, Kydd AS, Ouellette H, Nicolaou Kuo CF, Grainge MJ, Zhang W, Doherty M (2015) Global
S (2012) Dual energy CT in gout: a prospective valida- epidemiology of gout: prevalence, incidence and risk
tion study. Ann Rheum Dis 71(9):1466–1471 factors. Nat Rev Rheumatol 11(11):649–662
Chowalloor PV, Keen HI (2013) A systematic review of Lee SK, Jung JY, Jee WH, Lee JJ, Park SH (2019)
ultrasonography in gout and asymptomatic hyperuri- Combining non-contrast and dual-energy CT
caemia. Ann Rheum Dis 72(5):638–645 improves diagnosis of early gout. Eur Radiol 29(3):
Christiansen SN, Müller FC, Østergaard M, Slot O, Møller 1267–1275
JM, Børgesen HF, Gosvig KK, Terslev L (2020) Neogi T, Jansen TL, Dalbeth N, Fransen J, Schumacher
Dual-energy CT in gout patients: do all c­ olour-­coded HR, Berendsen D, Brown M, Choi H, Edwards NL,
lesions actually represent monosodium urate crystals? Janssens HJ, Liote F, Naden RP, Nuki G, Ogdie A,
Arthritis Res Ther 22(1):212 Perez-Ruiz F, Saag K, Singh JA, Sundy JS, Tausche
Cimmino MA, Zampogna G, Parodi M, Andracco R, AK, Vaquez-Mellado J, Yarows SA, Taylor WJ
Barbieri F, Paparo F, Ferrero G, Garlaschi G (2011) (2015) 2015 Gout classification criteria: an American
MRI synovitis and bone lesions are common in acute College of Rheumatology/European League Against
gouty arthritis of the wrist even during the first attack. Rheumatism collaborative initiative. Ann Rheum Dis
Ann Rheum Dis 70(12):2238–2239 74(10):1789–1798
Dalbeth N, Stamp L (2014) Hyperuricaemia and gout: Notzel A, Hermann KG, Feist E, Kedor C, Ziegeler K,
time for a new staging system? Ann Rheum Dis Stroux A, Mews J, Schnorfeil M, Hamm B, Diekhoff T
73(9):1598–1600 (2018) Diagnostic accuracy of dual-energy computed
Diekhoff T, Kiefer T, Stroux A, Pilhofer I, Juran R, tomography and joint aspiration: a prospective study
Mews J, Blobel J, Tsuyuki M, Ackermann B, Hamm in patients with suspected gouty arthritis. Clin Exp
B, Hermann KG (2015) Detection and character- Rheumatol 36(6):1061–1067
ization of crystal suspensions using single-source Parathithasan N, Lee WK, Pianta M, Oon S, Perera
dual-energy computed tomography: a phantom W (2016) Gouty arthropathy: review of clinico-­
model of crystal arthropathies. Investig Radiol pathologic and imaging features. J Med Imaging
50(4):255–260 Radiat Oncol 60(1):9–20
Diekhoff T, Kotlyarov M, Mews J, Hamm B, Hermann Phelps P, McCarty DJ Jr (1966) Crystal-induced inflam-
KGA (2018) Iterative reconstruction may improve mation in canine joints. II. Importance of polymorpho-
diagnosis of gout: an ex  vivo (bio)phantom dual-­ nuclear leukocytes. J Exp Med 124(1):115–126
energy computed tomography study. Investig Radiol Ramon A, Bohm-Sigrand A, Pottecher P, Richette P,
53(1):6–12 Maillefert JF, Devilliers H, Ornetti P (2018) Role of
Diekhoff T, Ulas ST, Poddubnyy D, Schneider U, Hermann dual-energy CT in the diagnosis and follow-up of gout:
S, Biesen R, Burmester GR, Hamm B, Hermann KG systematic analysis of the literature. Clin Rheumatol
(2019) Ultra-low-dose CT detects synovitis in patients 37(3):587–595
with suspected rheumatoid arthritis. Ann Rheum Dis Rothschild BM, Tanke D, Carpenter K (1997) Tyrannosaurs
78(1):31–35 suffered from gout. Nature 387(6631):357–357
Fukuda T, Umezawa Y, Asahina A, Nakagawa H, Furuya Rymal E, Rizzolo D (2014) Gout: a comprehensive
K, Fukuda K (2017) Dual energy CT iodine map for review. Jaapa 27(9):26–31
Gout 331

Schett G, Schauer C, Hoffmann M, Herrmann M (2015) Underwood M (2006) Diagnosis and management of
Why does the gout attack stop? A roadmap for the gout. BMJ 332(7553):1315–1319
immune pathogenesis of gout. RMD Open 1(Suppl Wang P, Smith SE, Garg R, Lu F, Wohlfahrt A, Campos
1):e000046 A, Vanni K, Yu Z, Solomon DH, Kim SC (2018)
Schumacher HR Jr, Becker MA, Edwards NL, Palmer Identification of monosodium urate crystal deposits in
WE, MacDonald PA, Palo W, Joseph-Ridge N (2006) patients with asymptomatic hyperuricemia using dual-­
Magnetic resonance imaging in the quantitative assess- energy CT. RMD Open 4(1):e000593
ment of gouty tophi. Int J Clin Pract 60(4):408–414 Wright JD, Pinto AB (2003) Clinical manifestations
Shang J, Zhou LP, Wang H, Liu B (2020) Diagnostic per- and treatment of gout. Prim Care Update OB/GYNS
formance of dual-energy CT versus ultrasonography 10:19–23
in gout: a meta-analysis. Acad Radiol 29(1):56–68 Zhang Z, Zhang X, Sun Y, Chen H, Kong X, Zhou J,
Sun Y, Chen H, Zhang Z, Ma L, Zhou J, Zhou Y, Ding Y, Zhou Y, Ma L, Jiang L (2017) New urate deposi-
Jin X, Jiang L (2015) Dual-energy computed tomog- tions on dual-energy computed tomography in gouty
raphy for monitoring the effect of urate-lowering ther- arthritis during urate-lowering therapy. Rheumatol Int
apy in gouty arthritis. Int J Rheum Dis 18(8):880–885 37(8):1365–1372
Taylor WJ, Fransen J, Dalbeth N, Neogi T, Ralph Zhu Y, Pandya BJ, Choi HK (2011) Prevalence of gout
Schumacher H, Brown M, Louthrenoo W, Vazquez-­ and hyperuricemia in the US general population: the
Mellado J, Eliseev M, McCarthy G, Stamp LK, Perez-­ National Health and Nutrition Examination Survey
Ruiz F, Sivera F, Ea HK, Gerritsen M, Scire CA, 2007–2008. Arthritis Rheum 63(10):3136–3141
Cavagna L, Lin C, Chou YY, Tausche AK, da Rocha Zhu Y, Pandya BJ, Choi HK (2012) Comorbidities of
Castelar-Pinheiro G, Janssen M, Chen JH, Slot O, gout and hyperuricemia in the US general population:
Cimmino M, Uhlig T, Jansen TL (2016) Diagnostic NHANES 2007-2008. Am J Med 125(7):679–687.e1
arthrocentesis for suspicion of gout is safe and well Ziegeler K, Hermann S, Hermann KGA, Hamm B,
tolerated. J Rheumatol 43(1):150–153 Diekhoff T (2020a) Dual-energy CT in the differentia-
Thiele RG, Schlesinger N (2007) Diagnosis of gout by tion of crystal depositions of the wrist: does it have
ultrasound. Rheumatology (Oxford) 46(7):1116–1121 added value? Skelet Radiol 49(5):707–713
Ulas ST, Diekhoff T, Hermann KGA, Poddubnyy D, Ziegeler K, Eshed I, Diekhoff T, Hermann KG (2020b)
Hamm B, Makowski MR (2019) Susceptibility-­ Imaging of joints and bones in autoinflammation. J
weighted MR imaging to improve the specificity of Clin Med 9(12):4074
erosion detection: a prospective feasibility study in
hand arthritis. Skelet Radiol 48(5):721–728
 Dual-Energy CT in Radiation
Oncology

Christian Richter and Patrick Wohlfahrt

Contents
1 I ntroduction: Dual-Energy CT in Radiotherapy: From First
Idea to Clinical Application   334
2 Dual-Energy CT Techniques in Radiation Oncology   335
2.1  Spectral Separation   335
2.2  Temporal Coherence   336
2.3  Spatio-Temporal Resolution   336
2.4  Cross-Scattering   337
2.5  Imaging Dose   337
2.6  Field of View   337
2.7  Respiratory   337
3 Tumor and Organ Segmentation   337
4  reatment Planning and Dose Calculation 
T  338
4.1  Photon Therapy   341
4.2  Proton Therapy   341
5 Potential of Photon-Counting CT in Radiation Oncology   343
6 Conclusion   345
References   345

C. Richter (*)
OncoRay – National Center for Radiation Research
in Oncology, Faculty of Medicine and University P. Wohlfahrt
Hospital Carl Gustav Carus, Technische Universität OncoRay – National Center for Radiation Research
Dresden, Helmholtz-Zentrum Dresden - Rossendorf, in Oncology, Faculty of Medicine and University
Dresden, Germany Hospital Carl Gustav Carus, Technische Universität
Dresden, Helmholtz-Zentrum Dresden - Rossendorf,
Helmholtz-Zentrum Dresden - Rossendorf, Institute
Dresden, Germany
of Radiooncology - OncoRay, Dresden, Germany
Massachusetts General Hospital and Harvard Medical
Department of Radiotherapy and Radiation
School, Department of Radiation Oncology,
Oncology, Faculty of Medicine and University
Boston, MA, USA
Hospital Carl Gustav Carus, Technische Universität
Dresden, Dresden, Germany Now with Siemens Healthineers, Forchheim,
e-mail: [email protected] Germany

© The Author(s), under exclusive license to Springer Nature Switzerland AG 2022 333
H. Alkadhi et al. (eds.), Spectral Imaging, Medical Radiology Diagnostic Imaging,
https://doi.org/10.1007/978-3-030-96285-2_23
334 C. Richter and P. Wohlfahrt

Abstract of the dose deposition in the target and normal tis-

sue is crucial. For this purpose, X-ray computed
In radiation oncology, CT imaging is of cru- tomography (CT) is the common imaging tech-
cial importance for consistent and accurate nique in external beam radiotherapy. However, CT
delineation of target and organ-at-risk struc- is not only used for predicting the physical dose
tures as well as for highly precise treatment deposition but is needed for several tasks in the
planning and prediction of the deposited dose. radiotherapeutic treatment chain. The clinical
Even though the potential benefit of dual-­ workflow in radiation oncology (RO) relies on a
energy CT for those purposes was identified well-coordinated interaction of a multitude of pro-
early, its implementation in clinical practice is cesses guided by imaging: Therapy decision before
still in an early stage. Here, we want to give an and adaptation during the course of treatment,
overview of current and potential future appli- organ and tumor segmentation, treatment planning,
cations of dual-energy CT in the field of radia- and dose calculation as well as patient immobiliza-
tion oncology. Since the next generation of tion and motion handling. Dual- or multi-energy
X-ray computed tomography with a photon-­ CT offers potential benefits compared to classical
counting detector technology is rising at the single-energy CT (SECT) in general but also spe-
horizon, we also want to give an outlook on cifically for the individual tasks which have differ-
radiotherapeutic applications that will benefit ent requirements on CT imaging.
or even become possible for the first time with The idea of the application of dual-energy CT
this technological evolution. in radiotherapy is anything but new. In his epoch-­
breaking publication introducing the first clinical
X-ray CT scanner in 1973 (Hounsfield 1973),
Godfrey Hounsfield not only introduced the basic
1 Introduction: Dual-Energy idea of dual-energy CT (DECT) by using two CT
CT in Radiotherapy: scans with different tube voltages, but he also
From First Idea to Clinical mentioned its potential use for improved tissue
Application characterization, especially the effective atomic
number. Only 4  years later, Michael Goitein, a
Nowadays, radiotherapy is an important treat- Boston-based pioneer of the translation of proton
ment technique in oncology, being applied either therapy in clinical application saw its potential
alone or together with surgery or chemotherapy for improved treatment planning in proton ther-
in about every second cancer patient (Borras apy (Goitein 1977).
et al. 2015). In most cases, high-energy external Due to the different relative contributions of
photon beams of several Megavolt (MV) are used photoelectric effect and incoherent/coherent scat-
to locally maximize the dose deposition to the tering, the two CT scans contain partly comple-
target while at the same time minimizing the mentary attenuation information of tissues. The
damage of the surrounding normal tissue basic idea is that two different quantities can be
(Baumann et  al. 2016). Alternatively, beams of obtained from two CT scans with different ratio of
high-energetic particles, mostly protons but also the two main interaction processes of kilovoltage
carbon ions, are used. In general, particle beams (kV) X-rays. Such tissue parameters are typically
lead to a better dose sparing of normal tissue due the relative electron density (RED) and effective
their well-defined maximum of dose deposition atomic number (EAN), which characterize the
shortly before stopping in the patient. In contrast photoelectric effect and incoherent/coherent scat-
to photons, they possess no or negligible dose tering. Especially in proton or particle therapy,
deposition “behind” the tumor target. exact knowledge of tissue parameters is crucial to
In both photon and proton therapy, an accurate accurately predict the particles stopping behavior
prediction of interactions of the therapeutic beam and thereby the dose ­deposition. As the stopping
with patients’ tissue and thereby precisely planning behavior and dose deposition are extremely sensi-
Dual-Energy CT in Radiation Oncology 335

tive to the traversed tissue, the application of • consecutive (dual-spiral mode) (Chap. 3),
DECT in particle therapy seems utmost promis- • split-beam filter in scan direction (twin-beam
ing already since the very beginning of CT imag- mode) (Chap. 3),
ing. Despite these clear conceptual benefits, it • dual-layer detector (dual-layer mode) (Chap.
took until 2015 as DECT was first used for clini- 4),
cal treatment planning in proton therapy • fast tube-voltage switching (fast-kVp mode)
(Wohlfahrt et al. 2017) probably due to the lack of (Chap. 5),
practical hardware and software solutions in the • slow tube-voltage switching (slow-kVp mode)
decades before and challenges in data processing, (Chap. 6).
e.g. its increased noise level.
Although the main motivation for deploying and dual-source CT scanners (Chap. 3) open up
DECT in radiation oncology is currently the new possibilities to reduce uncertainties in the
improvement of proton treatment planning overall radiotherapeutic chain. However, they
(Paganetti et  al. 2020; Wohlfahrt and Richter also add a new layer of complexity in the deci-
2020), several other potential use cases are dis- sion process which CT equipment might be best
cussed and investigated including (van Elmpt suited for the type and range of clinical applica-
et al. 2016): tions used preferably in the respective radiother-
apy institution. Hence, an individual optimization
• Improvements in tumor and organ of CT scan and reconstruction settings together
segmentation; with a careful weighing of technical benefits and
• Direct calculation of relative electron density limitations of each method are essential. To reli-
for photon treatment planning; ably obtain accurate and precise quantitative
• Tissue classification, e.g. as input for Monte measures from DECT, the overall image quality
Carlo transport calculations; is an important prerequisite for material decom-
• Improvement of image quality including metal position, because even small differences between
artifact reduction and variable image CT numbers in the low- and high-energy CT
contrast; dataset can be amplified in the highly sensitive
• Virtual subtraction of contrast agent informa- post-processing algorithms. The following tech-
tion to generate a virtual non-contrast and nical criteria are distinctive features of the vari-
contrast-enhanced dataset from only one sin- ous DECT acquisition techniques, which
gle DECT scan; considerably influence the DECT image quality
• Functional imaging. and thus affecting the performance of clinical
tasks from delineation to treatment planning and
In the following, we want to give an overview adaptive dose delivery.
on these different applications in radiation oncol-
ogy and also evaluate the different DECT tech-
niques for RO-specific applications. Furthermore, 2.1 Spectral Separation
the potential benefit of RO applications from
photon-counting CT technology will be The overlap of the low- and high-energy X-ray
illuminated. spectrum characterizes the gain of independent
attenuation information for material
differentiation. No spectral overlap would be
­
2 Dual-Energy CT Techniques ideal, because fully disjunct material attenuation
in Radiation Oncology properties are gathered. In general, a high spec-
tral separation facilitates accurate and precise
The various clinically available DECT acquisi- material decomposition and can better tolerate
tion techniques enabled by single-source CT image noise. It also defines the capability to reli-
scanners: ably classify two materials with only a slightly
336 C. Richter and P. Wohlfahrt

different spectral attenuation behavior, e.g. low neous acquisition of DECT projections with
iodine concentrations in contrast-enhanced 90° offset (dual-­source), however, resulting in a
DECT scans. temporal shift of at least 66 ms (quarter of rota-
DECT acquisition techniques with indepen- tion time). The time offset in twin-beam and
dent adjustment of their X-ray tube configura- slow-kVp mode is at least one rotation time.
tions (dual-source, dual-spiral, slow-kVp) Due to an additional time delay for tube adjust-
achieve a high spectral separation by selecting ments in slow-kVp mode, the temporal coher-
the lowest and highest available tube voltage. The ence is further reduced. The lowest temporal
spectral overlap can be further decreased by add- coherence is present in dual-­ spiral mode,
ing an extra filter material for the high-energy CT because the consecutive high-energy CT scan is
scan (dual-source, dual-spiral). Since a perfect delayed by the scan time as well as additional
voltage modulation following a rectangular func- time lags for tube adjustment and potential
tion is technically challenging in fast-kVp mode, couch movements. Despite that deformable
the original X-ray spectra are partly mixed in the image registration can reduce the impact of
transition period resulting in a reduced spectral anatomical changes, the dual-spiral mode is
separation. The use of a single tube voltage with currently recommended for static tumors only.
two different filters in scan direction (twin-beam) Contrast-enhanced dual-spiral DECT scans are
or an energy discrimination at detector level restricted to late phase with an almost stable
(dual-layer) also show an inferior spectral separa- distribution of contrast agent.
tion (McCollough et al. 2020). A high temporal coherence ensures accurate
A quantitative determination of material DECT-based material differentiation in non-­
parameters required for dose calculation (elec- static imaging situations. A stable patient immo-
tron density and stopping power) or tumor delin- bilization or breath-hold techniques can further
eation in low-contrast regions (concentration help to mitigate body motion.
contrast agent) benefit from an increase in spec-
tral separation. The assessment whether a spec-
tral separation is sufficient depends on the 2.3 Spatio-Temporal Resolution
purpose of the respective clinical task as well as
the targeted accuracy and precision. The temporal resolution is characterized by half
of the rotation time and thus depends on the
respective CT scanner capabilities in general.
2.2 Temporal Coherence The spatial resolution is mainly influenced by
focal spot size, slice collimation, and number of
A short or ideally no time interval between acqui- X-ray projections. All DECT modes except for
sitions of low- and high-energy X-ray attenua- fast-kVp can exploit the full CT scanner capabili-
tions ensures a temporal coherence of the DECT ties. In fast-kVp mode, a large focal spot is used
data collection. It reduces the impact of motion-­ and the projections per rotation need to be divided
induced anatomical changes or varying concen- in a ratio of 65:35 to increase exposure times for
trations and tissue distributions of contrast agents low-energy projections, because the tube current
over time. This is a crucial prerequisite to enable cannot be changed so fast. Due to the reduced
projection-based algorithms for material decom- number of projections per energy, the spatial res-
position and advanced physical corrections, e.g. olution is deteriorated.
of beam hardening, based on spectral X-ray A high spatio-temporal resolution is intended
information. at tissue transitions with high density gradients to
DECT in dual-layer mode has an almost per- prevent edge blurring, which can lead to a poten-
fect temporal coherence. A high temporal tial loss of attenuation information and thus devi-
coherence is also guaranteed by fast voltage ations in quantitative results relevant for dose
alteration within 0.5 ms (fast-kVp) or simulta- calculation.
Dual-Energy CT in Radiation Oncology 337

2.4 Cross-Scattering 2.7 Respiratory

An impact of cross-scatter radiation on the detec- A time-resolved (4D) respiratory acquisition is

tor signal is present in dual-source and dual-layer not yet available in clinical routine for the recent
mode and can be partly mitigated in post-­ DECT techniques. Only stable respiratory phases
processing algorithms. An insufficient correction realized by a breath-hold approach can be con-
could result in a systematic bias in the assessment sidered. In first proof-of-concept studies, the
of material parameters. clinical feasibility of 4D DECT has been demon-
strated (Ohira et al. 2018; Wohlfahrt et al. 2018),
however, confirmation for a broad variety of
2.5 Imaging Dose motion amplitudes and frequencies is desired.

DECT can be performed with the same total dose The image quality of DECT acquisitions
as required for SECT while enabling more post-­ depends on a multitude of technical parameters
scan flexibility without compromising image and needs to be optimized for the various pro-
quality (Wohlfahrt et al. 2017). Standard CT dose cesses in the complex radiotherapeutic workflow.
reduction techniques, such as tube current modu- Clinical applications in radiation oncology desire
lation, can be also applied in all DECT modes
except for fast-kVp due to limitations in current • a highly accurate and precise calculation of
tube technology. In dual-layer mode, the noise quantitative material parameters for dose calcu-
ratio between low- and high-energy CT data lation enabled by a good spectral separation,
directly follows from the design of the two detec- • a large scan field of view to cover the entire
tor layers and is thus optimized for specific scan patient and all necessary immobilization
conditions. devices,
Imaging dose becomes even more important • a time-resolved respiratory CT acquisition for
in an adaptive radiotherapy workflow relying on moving targets to estimate tumor motion as
continuous quantitative CT imaging during the well as optimize the treatment plan and dose
course of treatment for an early detection of ana- delivery,
tomical changes and tumor response assessment. • lowering the dose for CT acquisitions during
With the use of iterative image reconstruction, the course of treatment to still ensure an image
the image noise can be clearly reduced to still quality for a reliable assessment of material
guarantee a good performance of material char- parameters.
acterization at lower CT doses.
However, since none of the recent techniques
can fully satisfy all requirements, more explor-
2.6 Field of View atory clinical studies are needed to generate evi-
dence under clinical conditions and balance pros
All DECT modes except for dual-source acquire and cons of the respective techniques.
DECT information in a scan field of view of at
least 500 mm. The DECT field of view in dual-­
source mode is restricted to maximal 350  mm 3 Tumor and Organ
due to limitations in detector space within the CT Segmentation
gantry. This currently hampers the use for treat-
ment planning in the thoracic, abdominal, or pel- The main aim of segmentation in radiation oncol-
vic region and a proper coverage of immobilization ogy is a reliable delineation of the clinical target
devices. volume, encompassing the tumor region with an
338 C. Richter and P. Wohlfahrt

additional clinical safety margin covering micro- oncology. Joint efforts in collaborative trials
scopic tumor spread and tumor motility, as well between diagnostic radiology and radiation
as surrounding organs at risk as input for the oncology could be worth aspiring to generate evi-
patient-specific treatment planning process. dence in patient cohorts monitored in a standard-
The benefits of DECT associated with an ized approach covering diagnosis, therapy
improved tissue differentiation are commonly preparation, course of treatment, and follow-up.
employed in diagnostic oncological imaging to This can contribute to encourage a more wide-
increase diagnostic efficiency and efficacy spread routine use of DECT for tumor and organ
(Agrawal et al. 2014). Typically, various datasets segmentation in radiation oncology.
with different image contrasts are generated from In addition to an anatomical representation of
DECT after acquisition and then jointly used to the patient, functional information of organs can
improve tumor visibility and thus detectability be considered as an important input for treatment
(Forghani and Mukherji 2018). Depending on the planning and foster its individualization by incor-
selected energy of DECT-derived virtual monoen- porating information on the patient’s condition.
ergetic image (VMI) datasets, the soft tissue con- In active and well-functioning regions within
trast and tissue enhancement of contrast agents organs, the dose could be proactively reduced.
(low energy from 40 to 60 keV) can be increased Such areas with the potential of dose sparing can-
or metal artifacts (high energy from 120 to not only be identified with magnetic resonance
200 keV) can be reduced (Fig. 1). The distribu- imaging (MRI) or positron-emission tomography
tion and tissue enrichment of contrast medium (PET), but also with the help of DECT.  For
obtained by a DECT-based material decomposi- example, the distribution of active bone marrow
tion is an additional indicator to differentiate can be derived from a native DECT scan
between tumor invasion and ossification. A reli- (McGuire et al. 2011) and the lung function can
able quantification of contrast medium in ana- be obtained from the iodine distribution of a
tomical structures and tumors could potentially contrast-­ enhanced DECT scan (Bahig et  al.
even serve as biomarker for an early assessment 2017).
of therapy response (Fukukura et al. 2020).
Based on the clinical experience and evidence
gathered in diagnostic radiology within the last 4 Treatment Planning
two decades and the gradual implementation of and Dose Calculation
DECT in radiation oncology, the above-­
mentioned DECT applications also seem to be With the input of tumor and normal tissue seg-
promising for radiotherapeutic purposes. Their mentation, as well as case-specific dose prescrip-
integration in the radiotherapeutic workflow tions to the target and organ-specific tolerance
enables an individual assessment of the optimal doses for healthy tissues according to clinical
usage and potential benefit for tumor and organ guidelines, the dose distribution is optimized in a
segmentation. The number of clinical studies in multi-step process to achieve a homogeneous dis-
radiation oncology is still scarce, which demon- tribution of the prescribed dose in the target vol-
strate that an increased tissue contrast or a combi- ume while minimizing the dose to critical healthy
nation of image representations of different tissue anatomical structures. This optimization process
characteristics can translate into an improved aims at reaching a high tumor control while at the
delineation accuracy and reduced intra- and inter-­ same time minimizing radiation-induced
observer variability (Wohlfahrt et  al. 2018; ­toxicities in healthy tissues. The calculation of
Wohlfahrt et  al. 2019). A multitude of explor- the physical dose deposition based on the indi-
atory studies are needed in future to comprehen- vidual anatomy and tissue properties as well as
sively evaluate the clinical usability for different the setup of the radiation beam is a core element
tumor entities and tumor stages as well as elabo- in this process. For this purpose, the voxelwise
rate on potential improvements for radiation image information from CT and in some cases
Dual-Energy CT in Radiation Oncology 339

Contrast-enhanced
Native

80 keV 80 keV 60 keV 50 keV 40 keV

140 keV 150 keV 160 keV

Low-energy VMI
can enhance tissue contrasts
for tumor & organ segmentation

180 keV 120 keV 180 keV

High-energy VMI
can facilitate reducing artifacts
caused by metallic implants
(low & medium impact)

Low & medium High

impact of metal impact of metal

70 keV 170 keV without metal artifact correction

140 keV 170 keV with metal artifact correction

Fig. 1 The reconstruction of virtual monoenergetic energies). For severe metal artifacts, the application of a
image (VMI) datasets with different energies from dual-­ dedicated metal artifact reduction technique can be bene-
energy CT can contribute to enhance the tissue contrast ficial to improve the visual image impression. Adapted
for tumor and organ segmentation (low energies) or partly from (Wohlfahrt 2018)
mitigate image artifacts caused by metallic implants (high
340 C. Richter and P. Wohlfahrt

a Photon therapy: Impact of uncertainty in electron-density prediction on dose

Reference -3.5% deviating Dose difference
electron density electron density 2 1 -
1 2

Relative dose difference / %

Relative isodose lines / %

Relative isodose lines / %
40
100

Relative dose / %
80 20

60 0

40 -20

20 -40

b Proton therapy: Impact of uncertainty in stopping-power prediction on dose

Reference -3.5% deviating Dose difference
stopping power stopping power 2 1 -
1 2

Relative dose difference / %

Relative isodose lines / %

Relative isodose lines / %

40
100
Relative dose / %

80 20

60 0

40 -20

20 -40

Fig. 2  Impact of a 3.5% deviation in electron-density distal to the target region are visible in proton therapy,
(photon therapy, a) and stopping-power (proton therapy, whereas only a small change in the proximal dose distri-
b) prediction on the clinical dose distribution of an exem- bution is noticeable for photon therapy
plary brain-tumor patient case. Severe dose differences

from MRI, have to be converted in physical quan- tion. A slight change in the predicted stopping
tities needed for dose calculation of the treatment behavior of the traversed tissue will result in a
beam, namely RED for photon therapy and pro- change of the proton’s stopping position causing
ton stopping-power ratio (SPR) for proton severe changes in the local dose deposition. In
therapy. contrast, in photon therapy, the variation of dose
An uncertainty in the CT-based prediction of deposition over the beam’s penetration depth is
the respective material parameter has a direct much smaller, resulting in a much smaller depen-
impact on the accuracy of the dose calculation. dency on the predicted tissue parameter.
This effect is particularly pronounced in proton The standard procedure for the voxelwise con-
therapy (Fig.  2), since protons stop at a certain version of CT information (CT number, CTN)
depth in the patient and deposit their maximal into a parameter required for treatment planning
energy shortly before within the so-called Bragg (RED, SPR) is the use of a heuristic stepwise lin-
peak, followed by a steep distal dose fall-off to ear correlation function, a so-called Hounsfield
zero dose deposition behind the stopping posi- look-up table (HLUT). The HLUT can be derived
Dual-Energy CT in Radiation Oncology 341

from CT scans of tissue-equivalent materials has been confirmed to be far below 1% (Möhler
with known RED/SPR properties (experimental et  al. 2017). Interestingly, some of those algo-
HLUT calibration) or from the elemental compo- rithms are even mathematically equivalent
sition of tabulated human tissues (Woodard and (Möhler et al. 2018). Ultimately, by implement-
White 1986) combined with a theoretical predic- ing such a DECT-based direct RED determina-
tion of the scanner-specific CT numbers of those tion in photon treatment planning, the HLUT
non-available tissue samples (stoichiometric approach would be completely discarded together
HLUT calibration). with the associated disadvantages. This leads to
Despite its broad use in photon and proton reduced dose calculation uncertainties and a bet-
therapy, the HLUT approach has several general ter consideration of non-tissue materials.
limitations (Wohlfahrt and Richter 2020). Tissues However, this has not yet become broad clinical
with similar CTN but different RED or SPR can- practice, probably due to a rather small sensitiv-
not be distinguished. Since non-tissue materials, ity of the photon dose calculation to RED uncer-
e.g. implants, are often not covered by the HLUT tainties and practical limitations of DECT
due to different elemental composition, severe techniques (limited field of view or the chal-
deviations can occur for them. Moreover, a lenges in the presence of motion).
HLUT is per definition not patient-specific and Another application of DECT with an even
thus cannot consider the intra- and inter-patient higher clinical impact is the ability to virtually
variability in the conversion from CTN to RED remove the influence of the injected contrast
or SPR, which can arise from variations in tissue agent in tissue from CT images. This allows for
composition (e.g. different calcium content in the generation of quasi-native CT datasets from
bones). contrast-enhanced DECT acquisitions, the so-­
In the following, the benefit of DECT for dose called virtual non-contrast (VNC) datasets.
calculation is discussed for photon and proton Hence, instead of acquiring a contrast-enhanced
therapy consecutively. CT scan to support tumor delineation and a native
CT scan required for dose calculation, only the
contrast-enhanced DECT scan is needed with the
4.1 Photon Therapy VNC approach. This reduces the imaging dose
by 50%, eliminates potential registration inaccu-
The physical dose deposition of high-energetic racies between native and contrast-enhanced CT
MeV photons in tissue is dominated by incoher- images, and also reduces the time of the patient
ent scattering, which depends only on the relative on the CT couch. First investigations showed that
electron density of tissue. The most obvious ben- VNC-based dose calculation does not lead to
efit of DECT is that the only quantity needed for clinically relevant dose differences compared to
MV photon dose calculation, the electron den- dose calculation on native CT scans (Noid et al.
sity, can be directly derived from the two scans 2021). However, a decrease of the CTN in bones
with very high accuracy. This is possible, because has been noticed, indicating the need for further
the influence of the photoelectric effect, that also optimization of the VNC approach towards better
depends on EAN, can be eliminated by a specific differentiation between dense bone and iodine
weighted superimposition of the DECT scans. In contrast.
this case, it is a weighted subtraction of the low-­
energy CT image, possessing a higher influence
of photoelectric effect, from the high-energy CT 4.2 Proton Therapy
image. This results in an image contrast charac-
terized by incoherent scattering, which only For proton therapy, the dose deposition is charac-
depends on RED.  Several DECT-based imple- terized by the so-called stopping-power ratio
mentations of RED determination have been (SPR) of the tissue, which mainly depends on
introduced and their methodological uncertainty RED but also on the mean excitation potential
342 C. Richter and P. Wohlfahrt

(MEP) of the tissue. The SPR, which is the stop- • The application of sophisticated noise reduc-
ping power of the material divided by the stop- tion approaches, as especially EAN datasets
ping power of water, can be described as product can introduce a high noise level in the RSN/
of RED and a term called relative stopping num- MEP and subsequent SPR images.
ber (RSN), which includes the MEP as exclusive
tissue parameter. Those characteristics of a DirectSPR imple-
As already described in the previous section, mentation ultimately define the accuracy of SPR
RED can be directly derived from DECT with prediction in a real-world clinical scenario. In
very high accuracy. Even though DECT imaging general, the accuracy of RED and EAN datasets
allows for determining two tissue parameters, generated in applications designed for diagnostic
MEP cannot be directly obtained from DECT, purposes do not necessarily satisfy the require-
because it does not directly affect the photon ments needed for SPR prediction in proton ther-
interactions in CT imaging. Hence, a heuristic apy. Hence, certain CT manufacturer start to
conversion from EAN, which is a common sec- provide dedicated PT implementations for SPR
ond tissue quantity determined from DECT, prediction with optimized image post-­processing.
towards MEP or RSN is needed. Even though This includes a sophisticated noise-suppression
this step is associated with all disadvantages of a algorithm over the full image post-processing
heuristic conversion as discussed for the HLUT chain, which reduces the noise in SPR datasets
approach, its influence on the overall SPR predic- by roughly a factor of 2 with respect to the classic
tion is very limited: The variability of RSN in convolution of the high- and low-energy CT scan.
human tissues only contributes to roughly 5% of The selection of a size-dependent calibration for
the overall SPR variability in those tissues each individual axial CT slice is another benefit
(Möhler et al. 2016), whereas the remaining 95% for an SPR calculation directly integrated in the
are caused by RED variations. Consequently, the CT software. Such an integration can be seen as a
impact of the remaining uncertainty in the EAN-­ paradigm shift, as so far the CTN-to-SPR conver-
to-­RSN or EAN-to-MEP conversion on the SPR sion has been calibrated by each proton center
uncertainty is highly suppressed. individually and then applied within the treat-
In the past decade, about two dozen of algo- ment planning system. In 2019, the first clinical
rithms have been proposed to realize a direct implementation of DirectSPR for proton treat-
DECT-based SPR prediction (DirectSPR) ment planning has been realized (Wohlfahrt and
(Wohlfahrt and Richter 2020). Most of them are Richter 2020) and medical products have recently
image-based and use the analytical SPR calcula- become available for the whole workflow from
tion as described above, but also machine learn- generating SPR datasets to TPS import and pro-
ing approaches or completely empirical cessing. Still, the method is far away from broad
parametrizations as well as projection-based SPR clinical application despite the great interest in
calculation have been proposed. Even though a the PT community (Taasti et  al. 2018).
correct and robust algorithmic parametrization is Nevertheless, it has been shown that the accuracy
important, other factors in the implementation of of range prediction in treatment planning can be
a DirectSPR approach are also crucial to fulfill improved from 3.5% of the absolute range to 2%
the high requirements on accuracy and precision or even less in specific geometries (1.7% for
and need to be considered as well: brain-tumor treatments) (Peters et al. 2022). This
results in a reduction of the safety margin by 35%
• The optimization of the CTN constancy for and thus a relevantly decreased integral dose as
different scanning conditions by appropriate well as dose to organs at risk, as shown in Fig. 3
beam hardening correction; for a representative patient case.
• The comprehensive calibration of the approach For both treatment modalities, Monte Carlo
also including different beam hardening dose calculation is becoming more and more
­conditions or even introducing an object-size popular as their calculation speed has improved.
dependent calibration; Since those algorithms require the elemental
Dual-Energy CT in Radiation Oncology 343

HLUT DirectSPR Dose difference

conventional patient-specific 2 1-
1 2

Relative dose difference / %

Relative isodose lines / %

Relative isodose lines / %
40
100

Relative dose / %
80 20

60 0

40 -20

20 -40

Clinical target Brainstem Optical nerves

Fig. 3  Exemplary differences in clinical proton dose dis- uncertainty of 1.7%. Critical anatomical structures such as
tribution using the conventional Hounsfield look-up table the brainstem can be spared more effectively while
(HLUT) with a relative range uncertainty of 3.5% of total increasing the tumor coverage close to the brainstem. The
proton range and a direct stopping-power prediction clinical patient case was kindly provided by Nils Peters
(DirectSPR) from dual-energy CT with a reduced range (OncoRay, Dresden)

material composition for each CT voxel, DECT achieved. In combination with an almost constant
can also be used for an improved material assign- detector sensitivity, even for low-energy X-rays,
ment. The implementation of direct RED and the soft tissue contrast as well as the contrast
SPR prediction is also fully compatible with enhancement after contrast agent administration
highly accurate particle transport calculations of can be further improved (Flohr et al. 2020). The
Monte Carlo algorithms (Permatasari et al. 2020). selection of more than two energy bins for CT
reconstruction even enables multi-material
decompositions and differentiation of multiple
5 Potential of Photon- contrast agents with different attenuation charac-
Counting CT in Radiation teristics (k-edge imaging) (Chap. 8). Since the
Oncology energy discrimination is realized on detector
level, the spectral information is perfectly
Within the last two decades, substantial progress aligned, not hampered by motion-induced ana-
in research and development of a new CT detec- tomical changes, and can thus be used for
tor technology has been made to potentially over- projection-­ based material decomposition and
come limitations in current CT imaging by physics-based artifact correction, e.g. beam
translating from common energy-integrating hardening.
solid-state scintillation detectors to photon-­ These advantages of a photon-counting CT
counting semiconductor detectors in future system also tackle current restrictions of the vari-
(Chap. 7). Photon-counting CT always allows for ous dual-energy CT techniques as described
spectrally resolved multi-energy CT acquisitions above (Table  1). With photon-counting CT, a
in the full scan field of view with a high spectral high spectral separation, good quantitative per-
separation, perfect temporal coherence, and high formance in low-dose scenarios, no limitation in
spatio-temporal resolution. Due to a threshold-­ field of view and time-resolved respiratory spec-
based signal read-out of photon-counting detec- tral CT acquisition are feasible. The technical
tors, electronic noise can be suppressed and thus advances might then also potentially translate
a lower image noise or a further dose reduction as into clinical benefits in the radiotherapeutic
well as a higher CT number stability can be chain:
 344

Table 1  Comparison of spectral CT imaging enabled by various dual-energy CT techniques and photon-counting CT concerning technical specifications and specific applica-
tions in radiation oncology (RO)
Dual-source CT Single-source CT with EID
with EID dual-­ slow-­ fast-kVp dual-layer
spiral kVp twin-beam Photon-counting CT
Technical Spectral separation Very high Very High Medium Low High
specification high
Temporal coherence High Low Medium High Perfect Medium Perfect
Spatio-temporal Full capabilities Full capabilities Limited capability Full capabilities Full capabilities
resolution impaired spatial superior spatial
resolution resolution
Cross scatter Yes No Yes No No
Imaging dose Individual tube Individual tube No tube current Tube current modulation Tube current
current modulation current modulation modulation modulation
less noise at same dose
Field of view Limited up to Full Full
350 mm
RO-relevant Time-resolved Feasible with phase Feasible with phase matching in No Feasible with phase No limitation
application respiratory imaging matching in post-processing limitation matching in
post-processing post-processing
Contrast-enhanced Multi-phase (arterial Limited to late or Multi-phase (arterial & venous) Limited to late or Multi-phase (arterial &
imaging & venous) delayed phase delayed phase venous)
multiple contrast agents
Tissue segmentation Improved compared Improved compared to SECT Superior spatial
to SECT resolution, contrast and
noise
Projection-based No No Yes Yes No Yes
material
decomposition
C. Richter and P. Wohlfahrt
Dual-Energy CT in Radiation Oncology 345

• Improved tumor and organ segmentation due i­nstitutional research agreement with Siemens
to a higher spatial resolution, less noise, and Healthineers in the field of dual-energy CT for particle
therapy as well as an institutional agreement as reference
increased image contrast, center for dual-energy CT in radiotherapy and a software
• Reduction of uncertainties in treatment plan- evaluation contract. The authors received no financial sup-
ning and dose calculation with a further port for the present contribution.
improved quantitative prediction of material
parameters by projection-based material
decomposition and physics-based corrections, References
• More frequent quantitative image guidance
during the course of treatment enabled by Agrawal MD, Pinho DF, Kulkarni NM, Hahn PF,
Guimaraes AR, Sahani DV (2014) Oncologic
dose reduction to better support the patient-­ applications of dual-energy CT in the abdomen.
specific adaptive workflow with continuous Radiographics 34:589–612. https://doi.org/10.1148/rg.
therapy response and toxicity assessment, 343135041
• Increased flexibility in contrast agent adminis- Bahig H, Campeau M-P, Lapointe A, Bedwani S, Roberge
D, de Guise J, Blais D, Vu T, Lambert L, Chartrand-­
tration and CT acquisition of different contrast Lefebvre C, Lord M, Filion E (2017) Phase 1-2 study
phases. of dual-energy computed tomography for assessment
of pulmonary function in radiation therapy planning.
In summary, photon-counting CT could facili- Int J Radiat Oncol Biol Phys 99:334–343. https://doi.
org/10.1016/j.ijrobp.2017.05.051
tate the widespread clinical implementation of Baumann M, Krause M, Overgaard J, Debus J, Bentzen
spectral CT in radiation oncology potentially SM, Daartz J, Richter C, Zips D, Bortfeld T (2016)
leading to relevant improvements in the field of Radiation oncology in the era of precision medicine.
radiation oncology imaging. Nat Rev Cancer 16:234–249. https://doi.org/10.1038/
nrc.2016.18
Borras JM, Lievens Y, Dunscombe P, Coffey M, Malicki
J, Corral J, Gasparotto C, Defourny N, Barton M,
Verhoeven R, van Eycken L, Primic-Zakelj M,
6 Conclusion Trojanowski M, Strojan P, Grau C (2015) The optimal
utilization proportion of external beam radiotherapy
For CT industry, radiation oncology is definitely in European countries: an ESTRO-HERO analysis.
Radiother Oncol 116:38–44. https://doi.org/10.1016/j.
only a niche market. As the vast majority of CT radonc.2015.04.018
scanners are used for diagnostic radiology, the Flohr T, Petersilka M, Henning A, Ulzheimer S, Ferda
development of CT innovations initially focuses J, Schmidt B (2020) Photon-counting CT review.
on use cases in radiology, just as it was the case for Phys Med 79:126–136. https://doi.org/10.1016/j.
ejmp.2020.10.030
dual-energy CT and is the case for photon-­counting Forghani R, Mukherji SK (2018) Advanced dual-energy
CT.  Still, radiation oncology can greatly benefit CT applications for the evaluation of the soft tis-
from those technologies, if explored and tailored sues of the neck. Clin Radiol 73:70–80. https://doi.
to the specific needs of RO applications. Improved org/10.1016/j.crad.2017.04.002
Fukukura Y, Kumagae Y, Higashi R, Hakamada H,
accuracy of proton therapy by DECT-­based treat- Nagano H, Hidaka S, Kamimura K, Maemura K,
ment planning is probably the most prominent Arima S, Yoshiura T (2020) Visual enhancement pat-
example along this line. Without any doubt, the tern during the delayed phase of enhanced CT as an
further exploration of dual- and multi-­energy CT independent prognostic factor in stage IV pancreatic
ductal adenocarcinoma. Pancreatology 20:1155–1163.
for improved tumor and organ segmentation, dose https://doi.org/10.1016/j.pan.2020.07.009
reduction and its expansion towards time-resolved Goitein M (1977) The measurement of tissue het-
imaging has high potential to relevantly contribute erodensity to guide charged particle radiotherapy.
to an advanced radiation therapy of tomorrow. Int J Rad Oncol Biol Phys 3:27–33. https://doi.
org/10.1016/0360-­3016(77)90223-­1
Hounsfield GN (1973) Computerized transverse
Compliance with Ethical Standards axial scanning (tomography). 1. Description of
system. Br J Radiol 46:1016–1022. https://doi.
Disclosure of Interests  The authors received individual org/10.1259/0007-­1285-­46-­552-­1016
funding as lecturer from Siemens Healthineers (2018), McCollough CH, Boedeker K, Cody D, Duan X, Flohr T,
which was not related to this study. OncoRay had an Halliburton SS, Hsieh J, Layman RR, Pelc NJ (2020)
346 C. Richter and P. Wohlfahrt

Principles and applications of multienergy CT: report diction in Monte Carlo patient simulations using
of AAPM task group 291. Med Phys 47:e881–e912. stopping-power datasets. Phys Med Biol 65:185004.
https://doi.org/10.1002/mp.14157 https://doi.org/10.1088/1361-­6560/ab9702
McGuire SM, Menda Y, Ponto LLB, Gross B, Juweid M, Peters et al. 2022, Radiother Oncol 166, https://doi.
Bayouth JE (2011) A methodology for incorporating org/10.1016/j.radonc.2021.11.002
functional bone marrow sparing in IMRT planning for Taasti VT, Bäumer C, Dahlgren CV, Deisher AJ,
pelvic radiation therapy. Radiother Oncol 99:49–54. Ellerbrock M, Free J, Gora J, Kozera A, Lomax AJ,
https://doi.org/10.1016/j.radonc.2011.01.025 de Marzi L, Molinelli S, Kevin Teo B-K, Wohlfahrt
Möhler C, Wohlfahrt P, Richter C, Greilich S (2016) P, Petersen JBB, Muren LP, Hansen DC, Richter C
Range prediction for tissue mixtures based on dual-­ (2018) Inter-Centre variability of CT-based stopping-­
energy CT.  Phys Med Biol 61:N268–N275. https:// power prediction in particle therapy: survey-based
doi.org/10.1088/0031-­9155/61/11/N268 evaluation. Phys Imag Rad Oncol 6:25–30. https://doi.
Möhler C, Wohlfahrt P, Richter C, Greilich S (2017) org/10.1016/j.phro.2018.04.006
Methodological accuracy of image-based electron van Elmpt W, Landry G, Das M, Verhaegen F (2016) Dual
density assessment using dual-energy computed energy CT in radiotherapy: current applications and
tomography. Med Phys 44:2429–2437. https://doi. future outlook. Radiother Oncol 119:137–144. https://
org/10.1002/mp.12265 doi.org/10.1016/j.radonc.2016.02.026
Möhler C, Wohlfahrt P, Richter C, Greilich S (2018) Wohlfahrt P (2018) Dual-energy computed tomog-
On the equivalence of image-based dual-energy CT raphy for accurate stopping-power prediction in
methods for the determination of electron density proton treatment planning. Technische Universität
and effective atomic number in radiotherapy. Phys Dresden. Available from: https://nbn-resolving.org/
Imag Rad Oncol 5:108–110. https://doi.org/10.1016/j. urn:nbn:de:bsz:14-qucosa2-317554
phro.2018.03.003 Wohlfahrt P, Richter C (2020) Status and innovations
Noid G, Schott D, Paulson E, Zhu J, Shah J, Li XA (2021) in pre-treatment CT imaging for proton therapy.
Technical note: using virtual noncontrast images from Br J Radiol 93:20190590. https://doi.org/10.1259/
dual-energy CT to eliminate the need of precontrast bjr.20190590
CT for x-ray radiation treatment planning of abdomi- Wohlfahrt P, Möhler C, Hietschold V, Menkel S, Greilich
nal tumors†. Med Phys 48:1365–1371. https://doi. S, Krause M, Baumann M, Enghardt W, Richter C
org/10.1002/mp.14702 (2017) Clinical implementation of dual-energy CT for
Ohira S, Wada K, Hirata T, Kanayama N, Ikawa T, Karino proton treatment planning on pseudo-monoenergetic
T, Nitta Y, Isono M, Ueda Y, Miyazaki M, Koizumi CT scans. Int J Radiat Oncol Biol Phys 97:427–434.
M, Teshima T (2018) Clinical implementation of https://doi.org/10.1016/j.ijrobp.2016.10.022
contrast-enhanced four-dimensional dual-energy com- Wohlfahrt P, Troost EGC, Hofmann C, Richter C, Jakobi
puted tomography for target delineation of pancreatic A (2018) Clinical feasibility of single-source dual-­
cancer. Radiother Oncol 129:105–111. https://doi. spiral 4D dual-energy CT for proton treatment plan-
org/10.1016/j.radonc.2018.01.012 ning within the thoracic region. Int J Radiat Oncol
Paganetti H, Beltran CJ, Both S, Dong L, Flanz JB, Biol Phys 102:830–840. https://doi.org/10.1016/j.
Furutani KM, Grassberger C, Grosshans DR, ijrobp.2018.06.044
Knopf A-C, Langendijk JA, Nyström H, Parodi Wohlfahrt P, Agolli L, Krause M, Pilz K, Richter C, Troost
K, Raaymakers BW, Richter C, Sawakuchi GO, E (2019) PO-1012 dual-energy computed tomography
Schippers JM, Shaitelman SF, Teo K, Unkelbach J, for improved delineation in postoperative brain-tumor
Wohlfahrt P, Lomax AJ (2020) Roadmap: proton ther- patients. Radiother Oncol 133:S559–S560. https://doi.
apy physics and biology. Phys Med Biol. https://doi. org/10.1016/S0167-­8140(19)31432-­X
org/10.1088/1361-­6560/abcd16 Woodard HQ, White DR (1986) The composition of
Permatasari FF, Eulitz J, Richter C, Wohlfahrt P, Lühr body tissues. Br J Radiol 59:1209–1218. https://doi.
A (2020) Material assignment for proton range pre- org/10.1259/0007-­1285-­59-­708-­1209
 The Future of Spectral CT:
Radiomics and Beyond

Bettina Baessler and Davide Cester

Contents
1 Introduction   348
2  uture Advancements in Scanning Techniques and Image
F
Reconstruction   348
2.1  Automated Voltage and Current Optimization Techniques   349
2.2  Automated Patient Positioning and Scan Length Optimization   349
2.3  A Novel Image Quality Metric: The Detectability Index   350
2.4  Advancements in Contrast Media Applications   350
2.5  Reconstruction Techniques   351
3 The Hidden Potential of Underutilized Data   352
3.1   adiomics 
R  353
3.2  Machine Learning for Diagnosis   354
3.3  Opportunistic Screening   355
3.4  Distributed Learning   355
4 Limitations of Radiomics and Artificial Intelligence   356
5 Outlook   356
References   357

Abstract automation, the clinical histories of the

patients and their individual characteristics are
The spectral sensitivity of spectral detectors is going to become even more relevant in the
adding a whole new dimension to the data and determination of the investigation parameters,
this will soon lead to images obtained with while the automated algorithms based on tech-
lower radiation dose, in shorter times, and niques from the field of artificial intelligence
bearing greater diagnostic significance. Image will increasingly benefit from the knowledge
quality is just one aspect of the rapid evolution embedded in billions of archived images.
enabled by modern technologies: through Advancements in the fields of image analysis
and radiomics will soon be able to supply the
B. Baessler (*) · D. Cester radiologists with a growing number of robust
University Hospital Würzburg, Department of
Diagnostic and Interventional Radiology, and standardized indicators, seamlessly inte-
Würzburg, Germany grated into clinical devices and information
e-mail: [email protected]

© The Author(s), under exclusive license to Springer Nature Switzerland AG 2022 347
H. Alkadhi et al. (eds.), Spectral Imaging, Medical Radiology Diagnostic Imaging,
https://doi.org/10.1007/978-3-030-96285-2_24
348 B. Baessler and D. Cester

systems. This chapter summarizes four of the also at reducing the scan time or the dose
main current research areas: the improvement absorbed by the patient.
of the scan conditions, the optimization of the We present here an overview of four of these
applications of contrast media, new methods research areas (Fig.  1): the improvement of the
for image reconstruction and quality assess- scan conditions, the optimization of the applica-
ment, and the increasing use of “hidden data” tions of contrast media, new methods for image
and quantitative imaging biomarkers, includ- reconstruction and quality assessment, and the
ing radiomics, machine learning, and artifi- increasing use of “hidden data” and quantitative
cial/augmented intelligence. imaging biomarkers, including radiomics,
machine learning (ML), and artificial/augmented
intelligence (AI).

1 Introduction
2 Future Advancements
Despite being firstly demonstrated in the 1970s, in Scanning Techniques
CT is still an evolving technology with a lot of and Image Reconstruction
potential for improvement on the hardware (as
demonstrated by the recent introduction of PCCT Compared to the years when CT was introduced,
(see Chap.20)), software, and post-processing there is nowadays much more awareness from the
side. Generation after generation, CT scanners public regarding the biological effects of radia-
haven been able to provide more and better tion. While the use of CT in the last decades has
images. At the same time, the continuous prog- emerged to become standard practice, a growing
ress of computer science has enabled an incredi- proportion of the research effort has been dedi-
ble research effort in the field of software-based cated to dose optimization. However, despite the
image analysis, aiming to provide more informa- introduction of dose registers and dedicated ini-
tion that can be used in the diagnostic process. tiatives, dose levels used in modern radiology are
In addition to the hardware and image analysis still hugely different across countries and institu-
domains, there are other areas where innovation tions. A recent multinational study based on two
can have a significant impact on the efficiency of million CT scans found a four-fold range of
the diagnostic process, not necessarily aimed at effective dose for a standard abdominal CT with
improving the quality of the resulting images, but a 17-fold range for high-dose CT (Smith-­

Fig. 1  Summary of the

HARDWARE & SOFTWARE
different levels of a CT
investigation and the
relative topics discussed Automated voltage and crrent selection
in the chapter ACQUISITION Patient position and scan current optimization
Patient-optimized contrast protocols

RECONSTRUCTION Al-based reconstruction algorithms

IMAGE METRICS Standardization of quality indicators

Radiomics
IMAGE ANALYSIS Opportunistic screening
Distributed learning

DIAGNOSIS, THERAPY MONITORING, PROGNOSIS

The Future of Spectral CT: Radiomics and Beyond 349

Bindman et al. 2019). After adjusting for patient z-axis with a user-provided parameter represent-
characteristics, the major source of dose variation ing the desired contrast-to-noise value (CNR).
was attributed to institutional choices regarding The software then determines the tube voltage
scan parameters. These findings strongly suggest which minimizes the radiation dose while reach-
that the measured differences do not originate ing the requested CNR (Winklehner et al. 2011).
from limitations of CT technology, but rather The dose reduction potential of this technique has
from the lack of consistent and shared standards been reported to be between 13% and 77%
regarding its applications. (Niemann et  al. 2013; Papadakis and Damilakis
The most promising way to achieve the neces- 2019; Winklehner et al. 2011; Yu et al. 2013).
sary homogenization is by means of automated The dose reduction potential of voltage and
procedure optimization and automated evalua- current optimization could be further boosted by
tion metrics. Standardized algorithms could soon the introduction of AI algorithms. As an example,
enable a complete customization of the scan AI models could be trained on large datasets of
parameters in order to minimize the absorbed patients, and the acquisition of scout images
dose depending on the device and patient charac- could be replaced by an automated selection of
teristics. At the same time, novel and standard- voltage and current based on simpler patient
ized metrics will provide an objective evaluation characteristics and the type of examination.
of the image quality in terms of the actual effec-
tiveness of the diagnostic process.
2.2 Automated Patient
Positioning and Scan Length
2.1 Automated Voltage Optimization
and Current Optimization
Techniques Within the process of homogenization of scan
procedures, a fundamental factor which should
Several techniques have been successfully estab- be taken into account is patient positioning. Not
lished in order to limit dose levels in CT imaging only can patient off-centering directly impact
by taking into account patient characteristics in a image quality and radiation dose (Euler et  al.
standardized fashion, the most common ones 2019; Filev et al. 2016; Kaasalainen et al. 2019;
being automatic tube current modulation (ATCM) Li et  al. 2007; Schmidt et  al. 2020; Toth et  al.
and automatic tube voltage selection (ATVS). 2007; Habibzadeh et al. 2012; Toth et al. 2007)
The underlying concept is based on the acquisi- but it can also affect other aspects of the scan pro-
tion of low-dose images of the patient, called cedure; as an example, both ATCM and ATVS
scout images, which enable the calculation of the suffer from patient malpositioning (Euler et  al.
actual size of the patient and the approximated 2019; Filev et al. 2016; Gudjonsdottir et al. 2009;
attenuation profiles. The extracted data and the Kaasalainen et al. 2019; Marsh and Silosky 2017;
user settings are then used to optimize the scan- Matsubara et  al. 2009; Saltybaeva and Alkadhi
ner parameters in order to minimize the patient 2017; Schmidt et al. 2020; Winslow et al. 2018)
dose while reaching the required imaging because off-centering leads to magnification
performance. effects of the scout images.
ATCM works by modulating the current along Even when the patient is perfectly placed on
all three axes depending on the inhomogeneities the table, further optimization of radiation dose
in the attenuation; in particular, the current will be can be achieved by automated planning of the
lower than the average settings along directions scan range. An excessive scan range in up to 80%
where the scout images showed a stronger attenu- of thoracoabdominal CT scans was recently
ation, allowing to achieve dose reduction factors observed by Zanca et  al. (2012), a finding that
up to 68%. ATVS, on the other hand, combines varied substantially among institutions (Schwartz
the estimation of the attenuation profile along the et al. 2018).
350 B. Baessler and D. Cester

The use of three-dimensional cameras based nostic accuracy. Currently, the quality assess-
on infrared light, in combination with dedicated ment of CT images still relies heavily on direct
AI algorithms, has been investigated to address evaluation by radiologists, preventing a complete
both these issues. The optimization of the patient and reproducible standardization of the process.
positioning can be achieved by measuring the The answer to this issue can be found, when
body contour of the patient in order to create a radiation dose and other acquisition parameters
virtual representation, which then can be used to can be formally linked with image quality and
calculate the optimal table height. This method diagnostic accuracy. This approach is called task-­
proved to significantly improve patient centering based image quality assessment and represents a
for CT of the chest and abdomen compared with paradigm shift in the evaluation of CT image
manual positioning using the built-in laser sys- quality. Images are assessed based on their per-
tem (Booij et  al. 2019; Saltybaeva et  al. 2018). formance with respect to a specific diagnostic
By using a similar setup of infrared lights and 3D task and not the quality perceived by a human
cameras in combination with AI algorithms, it observer (Samei et  al. 2019). The detectability
was also possible to estimate the appropriate scan index d’ has been recently proposed and vali-
ranges in order to minimize unnecessary radia- dated in clinical studies as a promising task-based
tion dose (Colevray et al. 2019). In a few more image quality metric (Smith et al. 2017; Solomon
years, the task of correctly positioning the patient et  al. 2015). One study has shown that d’ can
could be completely automated and performed reflect lesion detection and image quality
by the CT device itself instead of being manually perception for clinical CT scans with hepatic
­
encoded into the protocol parameters. metastases (Cheng et al. 2020). When consider-
ing the recent developments of Spectral CT, a
standardized metric like d’ constitutes a highly
2.3  Novel Image Quality Metric:
A interesting tool to perform studies aimed at inves-
The Detectability Index tigating the energy dependence of both new and
existing techniques.
A successful diagnosis is undoubtedly the true Further work will be needed before these indi-
goal of a CT scan, and in principle, any potential cators can be officially integrated into dose man-
improvement resulting from a change in the scan agement workflows. Nevertheless, d’ and
parameters should be measured against its real comparable metrics represent the first steps of a
impact on the effectiveness in the diagnostic pro- necessary patient outcome-centric approach to
cess. This is unfortunately not completely possi- the standardized benchmark of radiation dose.
ble, given the fundamental role played by the
human factor, and researchers have reverted to
use simpler but more measurable and reproduc- 2.4 Advancements in Contrast
ible indicators like image contrast. Media Applications
The administered dose level, or more specifi-
cally its reduction, has become one of the most 2.4.1 Patient-Optimized Contrast
used metrics in the field, given its fundamental Media Protocols
importance in terms of patient safety. The radia- The K-absorption edge of iodine is 33.2  keV, a
tion dose alone, however, does not guarantee the value approximately located at the lower bound-
quality of a CT scan. On the contrary, the con- ary of the X-ray energy range used in CT.  In
tinuous search for lower radiation dose carries proximity of the K-edge, the energy absorption
the risk of worsening the actual detectability of due to photoelectric effect increases abruptly; for
lesions, and therefore negatively impact the this reason, lowering the tube voltage results in
patient outcome. It is clear that the effort to higher attenuation by iodinated contrast media
reduce the dose of the administered radiation has and produces stronger features in the images. The
to be balanced by the need to maintain the diag- main consequence in CT angiography is the pos-
The Future of Spectral CT: Radiomics and Beyond 351

sibility to administer lower volumes of contrast tional test bolus injection, bolus tracking
media while preserving the usual image quality technique is usually preferred as it needs less
(Lell et al. 2017; Martens et al. 2019; Schindera contrast media and a simpler, shorter scan proce-
et al. 2009; Winklehner et al. 2011). When com- dure (Bae 2010). It is not immune from draw-
bined with ATVS, this technique allows to mini- backs, though: bolus tracking uses a fixed trigger
mize the radiation dose received by the patient. delay to start a scan after the attenuation thresh-
The use of ATVS provides another possibility old is reached, and optimizing this delay for each
to reduce the contrast media by increasing the patient and examination would add considerable
tube current (Hendriks et al. 2018; Higashigaito complexity to the procedure.
et al. 2016). While in this case the dose reduction Efforts to introduce automation in this area
potential of ATVS is not fully exploited, the have already been made. A contrast enhancement
quantity of contrast agent can be personalized for prediction (CEP) algorithm was recently pre-
each individual patient (Higashigaito et al. 2016) sented (Korporaal et  al. 2015). This algorithm
which can be of particular importance when deal- incorporates population-averaged blood circula-
ing with patients with reduced kidney functional- tion characteristics and could be used for real-­
ity (Nijssen et al. 2019). time prediction of the scan delay with performance
Further possibilities to reduce the amount of comparable with the traditional approach. The
contrast media and the radiation dose may come introduction of AI algorithms in the field of CT
from future studies exploring the energy dimen- angiography will help to further tailor the scan
sion of the scan parameters enabled by modern parameters to each specific patient, leading to
DECT and PCCT technology, for example, by further reduction of absorbed radiation dose.
investigating the impact of dual-energy scanners
and Virtual Monoenergetic Images (VMI) (Grant
et  al. 2014; Husarik et  al. 2015; Skawran et  al. 2.5 Reconstruction Techniques
2020).
The traditional filtered back projection is still
2.4.2 Patient-Optimized Contrast regarded as the reference technique for CT image
Media Timing reconstruction. However, in recent years, several
As usual in the CT domain, when performing a vendors have introduced different hybrid and
contrasted CT angiography, the scan duration model-based iterative reconstruction (IR) algo-
should be reduced to the necessary minimum. rithms. Many of these IRs are optimized for a
However, this task is made more difficult by the specific application and each of them features its
simultaneous effort of minimizing the quantity of own advantages and trade-offs in terms of image
contrast agent, as discussed before. Optimizing quality (principally SNR), reconstruction speed
the scan timing is therefore an essential technique and dose reduction. Radiation dose reductions up
to reduce radiation dose without compromising to 76% have been reported, depending on the
image quality or the quantity of contrast agent. body region and the reference dose (Willemink
To reduce the overall scan duration, two main et al. 2013). Chest CT and CT angiography have
methods are commonly employed in CT angiog- been found to particularly benefit from IR (Den
raphy. The most common technique is bolus Harder et al. 2015a, b), while in the case of low-­
tracking, in which the scan initiates when the contrast tasks in abdominal CT the benefits are
contrast enhancement on a predefined vessel reported to be lower (Jensen et al. 2018; Schindera
reaches a predefined threshold value. An alterna- et al. 2013) and there could even be a worsening
tive is the test bolus technique; in this method, a of the examination yield when IR is used to
small test bolus of contrast medium (10–20 ml) is achieve high-dose reductions (Schindera et  al.
used to assess the cardiovascular circulation of 2013).
the specific patient before the main scan is per- AI-driven image reconstruction techniques,
formed (Mahnken et al. 2007). Due to the addi- often called deep learning reconstruction or res-
352 B. Baessler and D. Cester

toration (DLR), have recently showed great quality, and reconstruction time. However, CNNs
potential. A DLR is constituted by a DL model are still considered “black boxes” with the power
which is trained using the physics of the CT scan- to alter the image data in an unpredictable way.
ner and the scanned object (i.e., the patient), as The absence of an a priori analytical model for
well as a large sample of clinical CT images. the predictions typical of DL techniques implies
Depending on the training, the model can then be a certain degree of risk that algorithms based on
used for image denoising or to predict missing unsupervised learning might add or subtract
information. Convolutional neural networks structures and lesions from images, leading to
(CNNs) trained on low-dose CT images have wrong diagnoses with potentially severe conse-
been shown to have the potential to generate full-­ quences for the patients. The quantity and variety
dose CT images (Wolterink et al. 2017) and early of training data play by definition a major role in
clinical studies have reported improved image the process of determining this prediction model;
quality compared with IR for abdominal CT the current practice is to train CNNs on small and
(Akagi et al. 2019) and coronary CT angiography properly sampled image data, commensurate to
(Liu et al. 2020; Tatsugami et al. 2019). In situa- the level of complexity encountered in the clini-
tions when sparse-sampling CT techniques must cal routine clinical. Further effort will be needed
be employed and fewer projections are purposely to address all the safety concerns and to enable a
acquired, DLR can be used to generate this miss- successful implementation of DLRs in the clini-
ing information, with the potential net effect of cal practice.
reducing radiation dose while limiting the loss in
image quality (Dong et  al. 2019; Racine et  al.
2020). Based on the same principle of predicting 3  he Hidden Potential
T
missing information, DLR can also be used to of Underutilized Data
decrease streaking artifacts from metal implants
(Gjesteby et al. 2019; Liang et al. 2019). Another Due to its advantages and the continued improve-
study has shown the potential for DL to generate ments of the underlying technology, CT has
high-resolution images from low-resolution enjoyed an emerging popularity in the clinical
images in chest CT (Umehara et  al. 2018). workflow. Not only the use of CT has continu-
Another phantom study reported superior image ously increased since its introduction, but the ten-
noise, spatial resolution, and task-based detect- dency has even accelerated during the last
ability of DLR compared with IR (Higaki et al. decades (Bly et al. 2015; Le Coultre et al. 2016;
2020). Finally, DLR has been successfully Mettler et al. 2009; Pola et al. 2018). Several bil-
applied to improve X-ray scatter estimation, lions of images are generated every year, cover-
which is essential to improve image quality ing a vast variety of diseases and population
(Jiang et al. 2019; Maier et al. 2019). characteristics. Most of the datasets represent a
One of the most important features introduced complete 3D reconstruction for a specific organ
by spectral CT is the possibility to use the energy or an entire body region and potentially consti-
information to separate different materials; how- tutes valuable data for countless studies; how-
ever, this new aspect of image reconstruction ever, the extraction of useful data by trained
requires sophisticated algorithms and a precise human experts represents a huge bottleneck and
physical model. CNNs have the potential to cir- just a minority of the available information is
cumvent this complexity, and they have already actually extracted and used during the clinical
been applied to Multimaterial Decomposition diagnostic process. Moreover, data is usually
(MMD) (Chen and Li 2019) and K-edge imaging locally stored in the so-called data silos: data
(Zimmerman et al. 2020). from each institution can hardly be shared out-
All these findings indicate great potential for side the boundaries of the local IT infrastructure,
deep learning techniques to provide substantial mainly due to ethical, regulatory, and privacy
improvements in terms of dose reduction, image concerns. These restrictions severely limit the
The Future of Spectral CT: Radiomics and Beyond 353

amount of data available to each AI research proj- Since radiomics represents “big data,” analy-
ect, and ultimately limit the exploitation of its sis methods from the field of AI such as ML are
potential. The recent evolution of AI applications usually used to generate knowledge out of these
to medical imaging aims to address both issues. data. As a consequence, radiomics is of high
interest in order to enhance existing diagnostic
processes with additional, quantitative data. In
3.1 Radiomics addition, radiomics—combined with advanced
modeling - might be a valuable tool in providing
One emerging variant of using underutilized data prognostic information in various diseases.
is radiomics. Radiomics represents a quantitative Radiomic analysis in general can be per-
approach to medical imaging, which aims at formed on medical images from all available
enhancing the existing clinical data by means of modalities. When it comes to spectral CT, how-
advanced mathematical and/or statistical analysis. ever, the potential additive diagnostic and prog-
The concept of radiomics, which has most broadly nostic value of spectral CT data with their
been applied in the field of oncology, is based on additional quantitative information on the energy-­
the assumption that biomedical images contain dependent attenuation changes in various tissues
information of disease-specific processes that are (Al Ajmi et al. 2018) is of high interest in order to
imperceptible by the human eye (Mannil et  al. improve predictive models in radiomics studies.
2018) and thus not accessible through traditional While the number of publications on radiomics
visual inspection of an image (and, as a conse- in CT imaging has risen exponentially over the
quence, are underutilized). Through mathematical last few years, only very few proof-of-concept
extraction of the spatial distribution of signal-­ studies applying radiomic analyses on spectral
intensities and pixel-interrelationships, radiomics CT datasets have been published so far. However,
quantifies this textural information (Castellano it is to be expected that more and more studies
et al. 2004; Tourassi 1999) (also known as texture exploring the additive diagnostic potential of the
analysis). Hence, differences in image intensity, spectral information and radiomics will appear.
shape or texture can be quantified by means of Already in 2013, the first application of tex-
radiomics, thus overcoming the subjective nature ture analysis on spectral CT data has been
of image interpretation. An exemplary radiomics reported by Depeursinge et  al. (Depeursinge
workflow for spectral CT data is shown in Fig. 2. et al. 2013). They had noticed that the wealth of

Clinical Question Spectral CT Imaging Image Segmentation Radiomics Data Analysis

Shape

2D ROI

Intensity / Histogram

Iodine Map Therapy / Prognosis

VMIs 3D VOI
Texture ( Texture Analysis”)
”
Manual
Semi-automated
Spectral CT scanner Automated (DL)
Filters, e.g., wavelet

Fig. 2 Exemplary radiomics workflow for Spectral image segmentation, radiomic feature extraction, and
CT.  Schematic illustration of the entire patient journey data-driven patient-specific diagnostic or prognostic
including image acquisition, two- or three-dimensional assessment
354 B. Baessler and D. Cester

the recently introduced spectral CT 4D data (i.e., often used single-energy datasets (in this case
the 3D image series obtained for every X-ray 65  keV) (Al Ajmi et  al. 2018; Forghani et  al.
energy level between 40 and 140  keV) was 2019; Han et al. 2021; Li et al. 2020a, b). In the
mostly discarded by clinicians in clinical routine example of classifying the histopathology of
and that clinicians used only the single-energy benign parotid tumors, classification accuracy
images at 70 keV during diagnostic workup. The increased from 75% using the 65 keV monoener-
group proposed a self-developed computerized getic dataset to 92% when multi-energy datasets
multiscale rotation-covariant texture analysis of were used (Al Ajmi et  al. 2018). It has to be
spectral CT data for local pulmonary perfusion noted, that in general, patient populations in these
assessment in patients with acute pulmonary studies were relatively small, ranging from 42
embolism. This rotation covariant texture analy- (Al Ajmi et al. 2018) to 204 (Li et al. 2020a, b).
sis was aimed at allowing subtle characterization Although there has not been published any
of directional vascular morphological changes study so far on the prognostic value of combining
associated with acute pulmonary embolism. In the spectral information with radiomics, the
their small proof-of-concept study in 19 patients potential of radiomics combined with the infor-
with acute pulmonary embolism and 8 controls, mation on the energy-dependent attenuation
they demonstrated the feasibility of their changes in the tissue for assessment of treatment
approach and an improved diagnostic accuracy of response, outcome prediction, and risk assess-
the multiscale texture analysis model over the ment is huge. It is to be expected, that first studies
standard 70 keV approach (area under the curve will become available in this field within the next
[AUC] in receiver operating curve [ROC] analy- months and years.
sis 0.85 vs. 0.77, respectively).
Further studies reporting the application of
radiomics in spectral CT imaging have been pub- 3.2 Machine Learning
lished over the last 3  years. The time gap from for Diagnosis
2013 to 2018 might be due to the now increasing
availability of easily applicable and—at least to Besides the traditional, “hand-crafted” radiomics
some extent—commercially available radiomics approach, where individual features are extracted
software, whereas previous approaches had been from a drawn region or volume of interest in an
self-developed and self-programmed. image and then fed into some sort of ML algo-
Radiomics has been applied in several differ- rithm for the respective classification task or pre-
ent settings, such as to classify histopathology of dictive modeling, techniques from the ML
benign parotid tumors (Al Ajmi et  al. 2018), to spectrum also can be directly applied on the
predict lymph node metastasis in head and neck images. While such computer vision tasks have
squamous cell cancer (Forghani et al. 2019), pap- been applied broadly on standard CT images, a
illary thyroid cancer (Zhou et al. 2020), and gas- deep learning algorithm based on spectral CT
tric cancer (Li et  al. 2020a, b), to predict images has been reported only for the task of
microsatellite instability in colorectal carcinoma body composition analysis (Zopfs et al. 2020), as
(Wu et  al. 2019), for differentiating benign and further discussed in the section “opportunistic
malignant liver lesions (Homayounieh et  al. screening.” The diagnostic potential of these
2020), to classify renal clear cell carcinoma (Han approaches is yet to be determined and is
et  al. 2021), and to assess biopsy-proven liver expected to be an important research topic for the
fibrosis (Choi et al. 2020). Four of these studies following years.
(two from the same group) additionally investi- Interesting applications recently have been
gated the additive potential of the spectral infor- reported for ML techniques in combination with
mation (i.e., multi-energy virtual monoenergetic low and high keV virtual monoenergetic images
reconstructions) as compared to the clinically (without the use of radiomics or computer vision
The Future of Spectral CT: Radiomics and Beyond 355

tasks). Große Hokamp and colleagues used ML fat quantification (Commandeur et  al. 2018). In
and spectral CT images to facilitate renal stone addition, the assessment of skeletal muscle mass,
characterization (Große Hokamp et al. 2020) in a visceral adipose tissue, and bone mineral density
recently published ex  vivo study. They showed may provide prognostic information for the gen-
that even in compound stones the main compo- eral population as well as for patients with
nent could be reliably determined using spectral comorbidities (Burns et al. 2020; Cano-Espinosa
datasets combined with an ML approach. The et  al. 2018; Dabiri et  al. 2019; Lee et  al. 2017;
identification of the main stone components Wang et  al. 2017; Weston et  al. 2019). Very
thereby was independent of the dose protocol recently, the first deep learning algorithm based
used for image acquisition. on spectral CT images has been reported for the
ML-based CAD systems for a) determination task of body composition analysis (Zopfs et  al.
of preoperative invasion depth of gastric cancer 2020).
(Li et al. 2015) and b) for distinguishing pancre- Another AI field under rapid development is
atic mucinous cystic neoplasms from serous oli- segmentation. Automated segmentation of organ
gocystic adenomas (Li et al. 2016). size and volume may allow to determine
population-­specific size percentiles with greater
precision. The combination of AI-automated
3.3 Opportunistic Screening analysis with access to data from multiple
geographically-­ distributed datacenters would
By enabling a greater extent of the available data provide a unique opportunity to correlate disease
to be processed in a fully automated way (either and patient outcome with body composition on a
using radiomics or techniques from the field of regional or even global scale.
ML, DL, and computer vision), AI can bypass the
current bottleneck in the extraction of relevant
information from images. Several different anal- 3.4 Distributed Learning
yses of CT images could be conducted in back-
ground, while at the same time, the radiologist In order to fully exploit the potential of oppor-
continues to focus on the investigation which tunistic automated image analysis, advance-
originally motivated the scan. The additional ments to individual AI algorithms must be
information could be immediately accessed to complemented by improvements in the way
help the diagnostic process or just be stored for data is stored and accessed between different
later use. This strategy goes under the name of data centers, with special attention to the topics
opportunistic screening. of data governance and privacy issues.
One key feature potentially provided by AI Federated learning (FL), also called distributed
algorithms is the automated quantification of dif- learning, has recently emerged as an effective
ferent human tissues, also called body composi- paradigm to balance data access with protection
tion analysis. The ability to quantitatively of sensible information (Li et  al. 2020a, b;
discriminate the different tissues could enable Rieke et al. 2020). With FL the patient data is
cost-effective opportunistic screening and risk actually never transferred outside of the origi-
stratification, thus maximizing the effectiveness nal institution; instead, the analysis process
of CT scanning performed for other indications. occurs locally, and only the resulting model
Early studies have shown promising results for characteristics are shared. For many AI algo-
cardiovascular risk stratification with automated rithms, the possibility to access data from mul-
assessment of coronary artery stenosis (Hong tiple institutions could finally provide sufficient
et  al. 2019), automated Agatston score calcula- data to enable the transition from research to
tion (Cano-Espinosa et al. 2018), and epicardial clinical practice.
356 B. Baessler and D. Cester

4 Limitations of Radiomics able performance levels (Halevy et  al. 2009).

and Artificial Intelligence This lies in contrast to the usually small patient
cohorts used in diagnostic studies, especially in
Although radiomics, ML, DL, and all abovemen- cases with rare diseases. ML algorithms have a
tioned techniques have been shown to exhibit a large number of parameters to train, and the train-
huge potential for diagnostic, prognostic, and ing typically involves a lot of randomness. This
predictive purposes, the field is facing several poses unique challenges to the reproducibility of
challenges. One of the most important limitations trained algorithms (Beam et  al. 2020; Hutson
in the field of radiomics is the often poor repro- 2018). As a consequence, the interpretation of a
ducibility of radiomic studies due to lack of stan- reported model performance should be under-
dardization, insufficient reporting or limited open taken with caution, since it may be over-­
source code and data. The lack of proper valida- optimistic, especially in the lack of external
tion with the subsequent risk of false positive or validation (Liu et al. 2019; Park and Han 2018).
negative results limits translation to clinical prac-
tice (Chalkidou et  al. 2015). As detailed above,
current radiomics studies in the field of spectral 5 Outlook
CT represent first and small proof-of-concept
studies based on retrospectively collected data, It is to be expected that an increasing number of
whereas prospective randomized controlled stud- studies will come out, underlying the potential of
ies are required in the future to confirm the value the abovementioned techniques especially from the
of radiomics (van Timmeren et al. 2020). field of AI and radiomics. The spectral sensitivity
In most radiomic studies, imaging protocols, of the new detectors is adding a whole new dimen-
including acquisition- and reconstruction set- sion to the data and this will soon lead to images
tings, usually are not controlled or standardized. obtained with lower radiation dose, in shorter
Various studies have demonstrated the high times, and bearing greater diagnostic s­ ignificance.
impact of these technical settings on radiomic Image quality is just one aspect of the rapid evolu-
features or attempted to minimize their influence tion enabled by modern technologies: through
by eliminating features that are sensitive to these automation, the clinical histories of the patients and
variabilities. A summary can be found in van their individual characteristics are going to become
Timmeren et  al. van Timmeren et  al. (2020). even more relevant in the determination of the
While the Image Biomarker Standardization investigation parameters, while the automated
Initiative (IBSI) (Zwanenburg et  al. 2020; algorithms will increasingly benefit from the
Zwanenburg et al. 2016) has made some effort to knowledge embedded in billions of archived
achieve better standardization of radiomics, only images. Advancements in the fields of image anal-
sparse information is available when it comes to ysis and radiomics will soon be able to supply the
the robustness of radiomic features computed radiologists with a growing number of robust and
from spectral CT. Nevertheless, it is crucial, that standardized indicators, seamlessly integrated into
future radiomics studies in spectral CT adhere to clinical devices and information systems.
these guidelines (which was not the case in nearly Most of the improvements presented in this
all of the abovementioned studies). chapter are likely to take a few years, if not a decade,
Apart from the variations in scanners and set- to make their way to the clinical workflow; some of
tings, radiomic feature values are also influenced them may even be replaced by newer techniques
by patient variabilities, e.g. geometry, which before they reach full maturity. In the meantime, the
impact the levels of noise and presence of arti- evolution of detectors and the advancements of
facts in an image (Mühlberg et al. 2020). Computer Science will continue to provide even
ML and DL techniques are highly “data hun- more data and new methods to process it, soon fol-
gry.” For an ML model to be robust, it may lowed by applications to X-ray imaging.
require millions of observations to reach accept- The future of Spectral CT is now!
The Future of Spectral CT: Radiomics and Beyond 357

Compliance with Ethical Standards features: a systematic review. PLoS One 10:e0124165.
https://doi.org/10.1371/journal.pone.0124165
Disclosure of Interests None. Chen Z, Li L (2019) Robust multimaterial decomposi-
Ethical Approval  This article does not contain any stud- tion of spectral CT using convolutional neural net-
ies with human participants performed by any of the works. Opt Eng 58:1. https://doi.org/10.1117/1.
authors. OE.58.1.013104
This article does not contain any studies with animals Cheng Y, Smith TB, Jensen CT, Liu X, Samei E (2020)
performed by any of the authors. Correlation of algorithmic and visual assess-
ment of lesion detection in clinical images. Acad
Informed Consent  Informed consent was not necessary Radiol 27:847–855. https://doi.org/10.1016/j.
since the article does not contain any studies with human acra.2019.07.015
participants. Choi B, Choi IY, Cha SH, Yeom SK, Chung HH, Lee SH,
Cha J, Lee J-H (2020) Feasibility of computed tomog-
raphy texture analysis of hepatic fibrosis using dual-­
energy spectral detector computed tomography. Jpn J
References Radiol. https://doi.org/10.1007/s11604-­020-­01020-­5
Colevray, M., Tatard-Leitman, VM., Gouttard, S., Douek,
Akagi M, Nakamura Y, Higaki T, Narita K, Honda Y, P., Boussel, L., 2019. Convolutional neural network
Zhou J, Yu Z, Akino N, Awai K (2019) Deep learning evaluation of over-scanning in lung computed tomog-
reconstruction improves image quality of abdominal raphy. Diagn Interv Imaging 100, 177–183. doi:https://
ultra-high-resolution CT.  Eur Radiol 29:6163–6171. doi.org/10.1016/j.diii.2018.11.001
https://doi.org/10.1007/s00330-­019-­06170-­3 Commandeur F, Goeller M, Betancur J, Cadet S, Doris
Al Ajmi E, Forghani B, Reinhold C, Bayat M, Forghani R M, Chen X, Berman DS, Slomka PJ, Tamarappoo
(2018) Spectral multi-energy CT texture analysis with BK, Dey D (2018) Deep learning for quantification
machine learning for tissue classification: an investi- of Epicardial and thoracic adipose tissue from non-­
gation using classification of benign parotid tumours contrast CT. IEEE Trans Med Imaging 37:1835–1846.
as a testing paradigm. Eur Radiol 28:2604–2611. https://doi.org/10/df7b
https://doi.org/10.1007/s00330-­017-­5214-­0 Dabiri S, Popuri K, Cespedes Feliciano EM, Caan
Bae KT (2010) Intravenous contrast medium admin- BJ, Baracos VE, Beg MF (2019) Muscle seg-
istration and scan timing at CT: considerations mentation in axial computed tomography (CT)
and approaches. Radiology 256:32–61. https://doi. images at the lumbar (L3) and thoracic (T4) lev-
org/10.1148/radiol.10090908 els for body composition analysis. Comput Med
Beam AL, Manrai AK, Ghassemi M (2020) Challenges Imaging Graph 75:47–55. https://doi.org/10.1016/j.
to the reproducibility of machine learning models in compmedimag.2019.04.007
health care. JAMA 323:305. https://doi.org/10.1001/ Den Harder AM, Willemink MJ, De Ruiter QMB, De
jama.2019.20866 Jong PA, Schilham AMR, Krestin GP, Leiner T, Budde
Bly R, Jahnen A, Järvinen H, Olerud H, Vassileva J, RPJ (2015a) Dose reduction with iterative reconstruc-
Vogiatzi S (2015) Collective effective dose in Europe tion for coronary CT angiography: a systematic review
from X-ray and nuclear medicine procedures. Radiat and meta-analysis. BJR 89:20150068. https://doi.
Prot Dosim 165:129–132. https://doi.org/10.1093/rpd/ org/10.1259/bjr.20150068
ncv094 Den Harder AM, Willemink MJ, de Ruiter QMB,
Booij R, Budde RPJ, Dijkshoorn ML, van Straten M Schilham AMR, Krestin GP, Leiner T, de Jong PA,
(2019) Accuracy of automated patient positioning in Budde RPJ (2015b) Achievable dose reduction using
CT using a 3D camera for body contour detection. iterative reconstruction for chest computed tomogra-
Eur Radiol 29:2079–2088. https://doi.org/10.1007/ phy: a systematic review. Eur J Radiol 84:2307–2313.
s00330-­018-­5745-­z https://doi.org/10.1016/j.ejrad.2015.07.011
Burns JE, Yao J, Chalhoub D, Chen JJ, Summers RM Depeursinge A, Foncubierta-Rodriguez A, Vargas A, Van
(2020) A machine learning algorithm to estimate sar- de Villey D, Platon A, Poletti P-A, Muller H (2013)
copenia on abdominal CT. Acad Radiol 27:311–320. Rotation-covariant texture analysis of 4D dual-energy
https://doi.org/10.1016/j.acra.2019.03.011 CT as an indicator of local pulmonary perfusion.
Cano-Espinosa C, González G, Washko GR, Cazorla M, In: 2013 IEEE 10th International Symposium on
Estépar RSJ (2018) Automated Agatston score com- Biomedical Imaging. Presented at the 2013 IEEE 10th
putation in non-ECG gated CT scans using deep learn- International Symposium on Biomedical Imaging
ing. Proc SPIE Int Soc Opt Eng:10574. https://doi. (ISBI 2013). IEEE, San Francisco, CA, pp 145–148.
org/10/gjpxxq https://doi.org/10.1109/ISBI.2013.6556433
Castellano G, Bonilha L, Li LM, Cendes F (2004) Texture Dong J, Fu J, He Z (2019) A deep learning reconstruc-
analysis of medical images. Clin Radiol 59:1061– tion framework for X-ray computed tomography with
1069. https://doi.org/10.1016/j.crad.2004.07.008 incomplete data. PLoS One 14:e0224426. https://doi.
Chalkidou A, O’Doherty MJ, Marsden PK (2015) False org/10.1371/journal.pone.0224426
discovery rates in PET and CT studies with texture
358 B. Baessler and D. Cester

Euler A, Saltybaeva N, Alkadhi H (2019) How patient vidualized medicine: a prospective clinical study.
off-centering impacts organ dose and image noise in Investig Radiol 53:306–312. https://doi.org/10.1097/
pediatric head and thoracoabdominal CT. Eur Radiol RLI.0000000000000443
29:6790–6793. https://doi.org/10/gjpxxn Higaki T, Nakamura Y, Zhou J, Yu Z, Nemoto T,
Filev PD, Mittal PK, Tang X, Duong P-A, Wang Tatsugami F, Awai K (2020) Deep learning reconstruc-
X, Small WC, Applegate K, Moreno CC (2016) tion at CT: phantom study of the image characteris-
Increased computed tomography dose due to tics. Acad Radiol 27:82–87. https://doi.org/10.1016/j.
Miscentering with use of automated tube voltage acra.2019.09.008
selection: phantom and patient study. Curr Probl Higashigaito K, Husarik DB, Barthelmes J, Plass AR,
Diagn Radiol 45:265–270. https://doi.org/10.1067/j. Manka R, Maisano F, Alkadhi H (2016) Computed
cpradiol.2015.11.003 tomography angiography of coronary artery bypass
Forghani R, Chatterjee A, Reinhold C, Pérez-Lara A, grafts: low contrast media volume protocols adapted
Romero-Sanchez G, Ueno Y, Bayat M, Alexander to tube voltage. Investig Radiol 51:241–248. https://
JWM, Kadi L, Chankowsky J, Seuntjens J, Forghani doi.org/10.1097/RLI.0000000000000233
B (2019) Head and neck squamous cell carcinoma: Homayounieh F, Singh R, Nitiwarangkul C, Lades F,
prediction of cervical lymph node metastasis by dual-­ Schmidt B, Sedlmair M, Saini S, Kalra MK (2020)
energy CT texture analysis with machine learning. Semiautomatic segmentation and Radiomics for dual-­
Eur Radiol 29:6172–6181. https://doi.org/10.1007/ energy CT: a pilot study to differentiate benign and
s00330-­019-­06159-­y malignant hepatic lesions. Am J Roentgenol 215:398–
Gjesteby L, Shan H, Yang Q, Xi Y, Jin Y, Giantsoudi D, 405. https://doi.org/10/gkhjg9
Paganetti H, Man BD, Wang G (2019) A dual-stream Hong Y, Commandeur F, Cadet S, Goeller M, Doris
deep convolutional network for reducing metal streak MK, Chen X, Kwiecinski J, Berman DS, Slomka
artifacts in CT images. Phys Med Biol 64:235003. PJ, Chang H-J, Dey D (2019) Deep learning-based
https://doi.org/10.1088/1361-­6560/ab4e3e stenosis quantification from coronary CT angiogra-
Grant KL, Flohr TG, Krauss B, Sedlmair M, Thomas C, phy. Proc SPIE Int Soc Opt Eng 10949. https://doi.
Schmidt B (2014) Assessment of an advanced image-­ org/10.1117/12.2512168
based technique to calculate virtual Monoenergetic Husarik DB, Gordic S, Desbiolles L, Krauss B, Leschka
computed tomographic images from a dual-energy S, Wildermuth S, Alkadhi H (2015) Advanced vir-
examination to improve contrast-to-noise ratio tual Monoenergetic computed tomography of
in examinations using iodinated contrast media. Hyperattenuating and Hypoattenuating liver lesions:
Investig Radiol 49:586–592. https://doi.org/10.1097/ ex-vivo and patient experience in various body sizes.
RLI.0000000000000060 Investig Radiol 50:695–702. https://doi.org/10.1097/
Große Hokamp N, Lennartz S, Salem J, Pinto dos Santos RLI.0000000000000171
D, Heidenreich A, Maintz D, Haneder S (2020) Dose Hutson M (2018) Artificial intelligence faces repro-
independent characterization of renal stones by means ducibility crisis. Science 359:725–726. https://doi.
of dual energy computed tomography and machine org/10.1126/science.359.6377.725
learning: an ex-vivo study. Eur Radiol 30:1397–1404. Jensen CT, Wagner-Bartak NA, Vu LN, Liu X, Raval B,
https://doi.org/10.1007/s00330-­019-­06455-­7 Martinez D, Wei W, Cheng Y, Samei E, Gupta S (2018)
Gudjonsdottir J, Svensson JR, Campling S, Brennan PC, Detection of colorectal hepatic metastases is superior
Jonsdottir B (2009) Efficient use of automatic expo- at standard radiation dose CT versus reduced dose
sure control systems in computed tomography requires CT. Radiology 290:400–409. https://doi.org/10.1148/
correct patient positioning. Acta Radiol. https://doi. radiol.2018181657
org/10.3109/02841850903147053 Jiang Y, Yang C, Yang P, Hu X, Luo C, Xue Y, Xu L, Hu
Habibzadeh MA, Ay MR, Asl ARK, Ghadiri H, Zaidi X, Zhang L, Wang J, Sheng K, Niu T (2019) Scatter
H (2012) Impact of miscentering on patient dose correction of cone-beam CT using a deep residual
and image noise in x-ray CT imaging: phantom and convolution neural network (DRCNN). Phys Med
clinical studies. Phys Med 28:191–199. https://doi. Biol 64:145003. https://doi.org/10.1088/1361-­6560/
org/10.1016/j.ejmp.2011.06.002 ab23a6
Halevy A, Norvig P, Pereira F (2009) The unreason- Kaasalainen T, Mäkelä T, Kortesniemi M (2019) The
able effectiveness of data. IEEE Intell Syst 24:8–12. effect of vertical centering and scout direction on auto-
https://doi.org/10.1109/MIS.2009.36 matic tube voltage selection in chest CT: a prelimi-
Han, D., Yu, Y., He, T., Yu, N., Dang, S., Wu, H., Ren, nary phantom study on two different CT equipments.
J., Duan, X., 2021. Effect of radiomics from different Eur J Radiol Open 6:24–32. https://doi.org/10.1016/j.
virtual monochromatic images in dual-energy spectral ejro.2018.12.001
CT on the WHO/ISUP classification of clear cell renal Korporaal JG, Mahnken AH, Ferda J, Hausleiter J, Baxa
cell carcinoma. Clin Radiol S0009926021002063. J, Flohr TG, Schmidt BT (2015) Quantitative evalu-
https://doi.org/10/gkhjg5 ation of the performance of a new test BolusYBased
Hendriks BMF, Eijsvoogel NG, Kok M, Martens B, computed tomographic angiography contrast-­
Wildberger JE, Das M (2018) Optimizing pulmonary EnhancementYPrediction algorithm. Investig Radiol
embolism computed tomography in the age of indi- 50:8
The Future of Spectral CT: Radiomics and Beyond 359

Le Coultre R, Bize J, Champendal M, Wittwer D, Ryckx Maier J, Eulig E, Vöth T, Knaup M, Kuntz J, Sawall S,
N, Aroua A, Trueb P, Verdun FR (2016) Exposure Kachelrieß M (2019) Real-time scatter estimation for
of the swiss population by radiodiagnostics: 2013 medical CT using the deep scatter estimation: method
review. Radiat Prot Dosim 169:221–224. https://doi. and robustness analysis with respect to different
org/10.1093/rpd/ncv462 anatomies, dose levels, tube voltages, and data trunca-
Lee H, Troschel FM, Tajmir S, Fuchs G, Mario J, tion. Med Phys 46:238–249. https://doi.org/10.1002/
Fintelmann FJ, Do S (2017) Pixel-level deep segmen- mp.13274
tation: artificial intelligence quantifies muscle on com- Mannil M, von Spiczak J, Manka R, Alkadhi H (2018)
puted tomography for body morphometric analysis. J Texture analysis and machine learning for detect-
Digit Imaging 30:487–498. https://doi.org/10/gjpxxp ing myocardial infarction in noncontrast low-dose
Lell MM, Fleischmann U, Pietsch H, Korporaal JG, computed tomography: unveiling the invisible.
Haberland U, Mahnken AH, Flohr TG, Uder M, Investig Radiol 53:338–343. https://doi.org/10.1097/
Jost G (2017) Relationship between low tube volt- RLI.0000000000000448
age (70 kV) and the iodine delivery rate (IDR) in CT Marsh RM, Silosky MS (2017) The effects of patient posi-
angiography: an experimental in-vivo study. PLoS tioning when interpreting CT dose metrics: a phantom
One 12:e0173592. https://doi.org/10.1371/journal. study. Med Phys 44:1514–1524. https://doi.org/10/
pone.0173592 gjpxxm
Li C, Lin X, Hui C, Lam KM, Zhang S (2016) Computer-­ Martens B, Hendriks BMF, Eijsvoogel NG, Wildberger
aided diagnosis for distinguishing pancreatic mucinous JE, Mihl C (2019) Individually body weight–adapted
cystic neoplasms from serous Oligocystic adenomas in contrast media application in computed tomography
spectral CT images. Technol Cancer Res Treat 15:44– imaging of the Liver at 90 kVp. Investig Radiol 54:177–
54. https://doi.org/10.1177/1533034614563013 182. https://doi.org/10.1097/RLI.0000000000000525
Li C, Shi C, Zhang H, Hui C, Lam KM, Zhang S (2015) Matsubara K, Koshida K, Ichikawa K, Suzuki M, Takata
Computer-aided diagnosis for preoperative invasion T, Yamamoto T, Matsui O (2009) Misoperation of
depth of gastric cancer with dual-energy spectral CT automatic tube current modulation systems with
CT imaging. Acad Radiol 22:149–157. https://doi. inappropriate patient centering: phantom studies. Am
org/10.1016/j.acra.2014.08.006 J Roentgenol 192:862–865. https://doi.org/10.2214/
Li J, Dong D, Fang M, Wang R, Tian J, Li H, Gao J (2020a) AJR.08.1472
Dual-energy CT–based deep learning radiomics can Mettler FA, Bhargavan M, Faulkner K, Gilley DB, Gray
improve lymph node metastasis risk prediction for JE, Ibbott GS, Lipoti JA, Mahesh M, McCrohan JL,
gastric cancer. Eur Radiol 30:2324–2333. https://doi. Stabin MG, Thomadsen BR, Yoshizumi TT (2009)
org/10/gkhjgt Radiologic and nuclear medicine studies in the United
Li J, Udayasankar UK, Toth TL, Seamans J, Small WC, States and worldwide: frequency, radiation dose, and
Kalra MK (2007) Automatic patient centering for comparison with other radiation sources—1950–2007.
MDCT: effect on radiation dose. Am J Roentgenol Radiology 253:520–531. https://doi.org/10.1148/
188:547–552. https://doi.org/10.2214/AJR.06.0370 radiol.2532082010
Li L, Fan Y, Tse M, Lin K-Y (2020b) A review Mühlberg A, Katzmann A, Heinemann V, Kärgel R,
of applications in federated learning. Comput Wels M, Taubmann O, Lades F, Huber T, Maurus
Ind Eng 149:106854. https://doi.org/10.1016/j. S, Holch J, Faivre J-B, Sühling M, Nörenberg D,
cie.2020.106854 Rémy-Jardin M (2020) The Technome  - a predictive
Liang K, Zhang L, Yang H, Yang Y, Chen Z, Xing Y (2019) internal ­calibration approach for quantitative imag-
Metal artifact reduction for practical dental computed ing biomarker research. Sci Rep 10:1103. https://doi.
tomography by improving interpolation-based recon- org/10.1038/s41598-­019-­57325-­7
struction with deep learning. Med Phys 46:e823–e834. Niemann T, Henry S, Faivre J-B, Yasunaga K, Bendaoud
https://doi.org/10.1002/mp.13644 S, Simeone A, Remy J, Duhamel A, Flohr T, Remy-­
Liu P, Wang M, Wang Y, Yu M, Wang Y, Liu Z, Li Y, Jin Jardin M (2013) Clinical evaluation of automatic
Z (2020) Impact of deep learning-based optimization tube voltage selection in chest CT angiography.
algorithm on image quality of low-dose coronary CT Eur Radiol 23:2643–2651. https://doi.org/10.1007/
angiography with noise reduction: a prospective study. s00330-­013-­2887-­x
Acad Radiol 27:1241–1248. https://doi.org/10.1016/j. Nijssen EC, Nelemans PJ, Rennenberg RJ, Theunissen
acra.2019.11.010 RA, van Ommen V, Wildberger JE (2019) Prophylaxis
Liu Y, Chen P-HC, Krause J, Peng L (2019) How to read in high-risk patients with eGFR < 30 mL/min/1.73
articles that use machine learning: users’ guides to m2. Investig Radiol 54(9):580–588
the medical literature. JAMA 322:1806. https://doi. Papadakis AE, Damilakis J (2019) Automatic tube cur-
org/10.1001/jama.2019.16489 rent modulation and tube voltage selection in pediatric
Mahnken AH, Rauscher A, Klotz E, Mühlenbruch G, Das computed tomography: a phantom study on radiation
M, Günther RW, Wildberger JE (2007) Quantitative dose and image quality. Investig Radiol 54:265–272.
prediction of contrast enhancement from test bolus https://doi.org/10.1097/RLI.0000000000000537
data in cardiac MSCT.  Eur Radiol 17:1310–1319. Park SH, Han K (2018) Methodologic guide for evalu-
https://doi.org/10.1007/s00330-­006-­0486-­9 ating clinical performance and effect of artificial
360 B. Baessler and D. Cester

intelligence technology for medical diagnosis and Eur J Radiol 102:49–54. https://doi.org/10.1016/j.
prediction. Radiology 286:800–809. https://doi. ejrad.2018.03.005
org/10.1148/radiol.2017171920 Skawran S, Angst F, Blüthgen C, Eberhard M, Kälin
Pola A, Corbella D, Righini A, Torresin A, Colombo PE, P, Kobe A, Nagy D, Szucs-Farkas Z, Alkadhi H,
Vismara L, Trombetta L, Maddalo M, Introini MV, Euler A (2020) Dual-energy low-keV or single-
Tinelli D, Strohmenger L, Garattini G, Munari A, energy low-­ kV CT for Endoleak detection?: a
Triulzi F (2018) Computed tomography use in a large 6-reader study in an aortic aneurysm phantom.
Italian region: trend analysis 2004-2014 of emergency Investig Radiol 55:45–52. https://doi.org/10.1097/
and outpatient CT examinations in children and adults. RLI.0000000000000606
Eur Radiol 28:2308–2318. https://doi.org/10.1007/ Smith TB, Solomon JB, Samei E (2017) Estimating
s00330-­017-­5225-­x detectability index in  vivo: development and valida-
Racine D, Becce F, Viry A, Monnin P, Thomsen B, Verdun tion of an automated methodology. JMI 5:031403.
FR, Rotzinger DC (2020) Task-based characterization https://doi.org/10.1117/1.JMI.5.3.031403
of a deep learning image reconstruction and com- Smith-Bindman R, Wang Y, Chu P, Chung R, Einstein AJ,
parison with filtered back-projection and a partial Balcombe J, Cocker M, Das M, Delman BN, Flynn
model-based iterative reconstruction in abdominal M, Gould R, Lee RK, Yellen-Nelson T, Schindera S,
CT: a phantom study. Phys Med 76:28–37. https://doi. Seibert A, Starkey J, Suntharalingam S, Wetter A,
org/10.1016/j.ejmp.2020.06.004 Wildberger JE, Miglioretti DL (2019) International
Rieke N, Hancox J, Li W, Milletarì F, Roth HR, Albarqouni variation in radiation dose for computed tomography
S, Bakas S, Galtier MN, Landman BA, Maier-Hein K, examinations: prospective cohort study. BMJ 364.
Ourselin S, Sheller M, Summers RM, Trask A, Xu https://doi.org/10.1136/bmj.k4931
D, Baust M, Cardoso MJ (2020) The future of digital Solomon J, Mileto A, Ramirez-Giraldo JC, Samei E
health with federated learning. NPJ Digit Med 3:119. (2015) Diagnostic performance of an advanced mod-
https://doi.org/10.1038/s41746-­020-­00323-­1 eled iterative reconstruction algorithm for low-­contrast
Saltybaeva N, Alkadhi H (2017) Vertical off-centering detectability with a third-generation dual-source mul-
affects organ dose in chest CT: evidence from Monte tidetector CT scanner: potential for radiation dose
Carlo simulations in anthropomorphic phantoms. Med reduction in a multireader study. Radiology 275:735–
Phys 44:5697–5704. https://doi.org/10/ghpwdd 745. https://doi.org/10.1148/radiol.15142005
Saltybaeva N, Schmidt B, Wimmer A, Flohr T, Alkadhi Tatsugami F, Higaki T, Nakamura Y, Yu Z, Zhou J, Lu Y,
H (2018) Precise and automatic patient positioning in Fujioka C, Kitagawa T, Kihara Y, Iida M, Awai K (2019)
computed tomography: avatar modeling of the patient Deep learning–based image restoration algorithm for
surface using a 3-dimensional camera. Investig Radiol coronary CT angiography. Eur Radiol 29:5322–5329.
53:641–646. https://doi.org/10/ggrmfc https://doi.org/10.1007/s00330-­019-­06183-­y
Samei E, Bakalyar D, Boedeker KL, Brady S, Fan J, Leng van Timmeren J, Cester D, Tanadini-Lang S, Alkadhi
S, Myers KJ, Popescu LM, Giraldo JCR, Ranallo F, H, Baessler B (2020) Radiomics in medical imag-
Solomon J, Vaishnav J, Wang J (2019) Performance ing  – “how-to” guide and critical reflection. Insights
evaluation of computed tomography systems: sum- Imaging 11(1):1–16
mary of AAPM task group 233. Med Phys 46:e735– Toth T, Ge Z, Daly MP (2007) The influence of patient
e756. https://doi.org/10.1002/mp.13763 centering on CT dose and image noise. Med Phys
Schindera ST, Graca P, Patak MA, Abderhalden S, 34:3093–3101. https://doi.org/10.1118/1.2748113
von Allmen G, Vock P, Szucs-Farkas Z (2009) Tourassi GD (1999) Journey toward computer-aided
Thoracoabdominal-Aortoiliac multidetector-row CT diagnosis: role of image texture analysis. Radiology
angiography at 80 and 100 kVp: assessment of image 213:317–320. https://doi.org/10.1148/radiology.213.2
quality and radiation dose. Investig Radiol 44:650– .r99nv49317
655. https://doi.org/10.1097/RLI.0b013e3181acaf8a Umehara K, Ota J, Ishida T (2018) Application of super-­
Schindera ST, Odedra D, Raza SA, Kim TK, Jang H-J, resolution convolutional neural network for enhancing
Szucs-Farkas Z, Rogalla P (2013) Iterative reconstruc- image resolution in chest CT. J Digit Imaging 31:441–
tion algorithm for CT: can radiation dose be decreased 450. https://doi.org/10.1007/s10278-­017-­0033-­z
while low-contrast detectability is preserved? Wang Y, Qiu Y, Thai T, Moore K, Liu H, Zheng B (2017)
Radiology 269:511–518. https://doi.org/10.1148/ A two-step convolutional neural network based
radiol.13122349 computer-aided detection scheme for automatically
Schmidt C, Baessler B, Nakhostin D, Das A, Eberhard segmenting adipose tissue volume depicting on CT
M, Alkadhi H, Euler A (2020) Dual-energy CT-based images. Comput Methods Prog Biomed 144:97–104.
iodine quantification in liver tumors–impact of scan-, https://doi.org/10/gbgxvq
patient-, and position-related factors. Acad Radiol Weston AD, Korfiatis P, Kline TL, Philbrick KA,
28(6):783–789. S1076633220302142. https://doi. Kostandy P, Sakinis T, Sugimoto M, Takahashi N,
org/10.1016/j.acra.2020.04.021 Erickson BJ (2019) Automated abdominal segmenta-
Schwartz F, Stieltjes B, Szucs-Farkas Z, Euler A (2018) tion of CT scans for body composition analysis using
Over-scanning in chest CT: comparison of practice deep learning. Radiology 290:669–679. https://doi.
among six hospitals and its impact on radiation dose. org/10/ggv9fz
The Future of Spectral CT: Radiomics and Beyond 361

Willemink MJ, Leiner T, de Jong PA, de Heer LM, Zhou Y, Su G-Y, Hu H, Ge Y-Q, Si Y, Shen M-P, Xu X-Q,
Nievelstein RAJ, Schilham AMR, Budde RPJ (2013) Wu F-Y (2020) Radiomics analysis of dual-energy
Iterative reconstruction techniques for computed CT-derived iodine maps for diagnosing metastatic cer-
tomography part 2: initial results in dose reduction and vical lymph nodes in patients with papillary thyroid
image quality. Eur Radiol 23:1632–1642. https://doi. cancer. Eur Radiol 30:6251–6262. https://doi.org/10/
org/10.1007/s00330-­012-­2764-­z gkhjhj
Winklehner A, Goetti R, Baumueller S, Karlo C, Schmidt Zimmerman KC, Sharma G, Parchur AK, Joshi A,
B, Raupach R, Flohr T, Frauenfelder T, Alkadhi H Schmidt TG (2020) Experimental investigation of
(2011) Automated attenuation-based tube potential neural network estimator and transfer learning tech-
selection for Thoracoabdominal computed tomog- niques for K-edge spectral CT imaging. Med Phys
raphy angiography: improved dose effectiveness. 47:541–551. https://doi.org/10.1002/mp.13946
Investig Radiol 46:767–773. https://doi.org/10.1097/ Zopfs D, Bousabarah K, Lennartz S, dos Santos DP,
RLI.0b013e3182266448 Schlaak M, Theurich S, Reimer RP, Maintz D,
Winslow J, Zhang Y, Koweek L, Samei E (2018) Haneder S, Große Hokamp N (2020) Evaluating
Dependency of prescribed CT dose on table height, body composition by combining quantitative
patient size, and localizer acquisition for one clini- spectral detector computed tomography and
cal MDCT.  Phys Med 55:56–60. https://doi.org/10/ deep learning-based image segmentation. Eur
gfqd6h J Radiol 130:109153. https://doi.org/10.1016/j.
Wolterink JM, Leiner T, Viergever MA, Išgum I (2017) ejrad.2020.109153
Generative adversarial networks for noise reduction Zwanenburg, A., Leger, S., Vallières, M., Löck, S., 2016.
in low-dose CT. IEEE Trans Med Imaging 36:2536– Image biomarker standardisation initiative
2545. https://doi.org/10.1109/TMI.2017.2708987 Zwanenburg A, Vallières M, Abdalah MA, Aerts HJWL,
Wu J, Zhang Q, Zhao Y, Liu Y, Chen A, Li X, Wu T, Andrearczyk V, Apte A, Ashrafinia S, Bakas S,
Li J, Guo Y, Liu A (2019) Radiomics analysis of Beukinga RJ, Boellaard R, Bogowicz M, Boldrini L,
iodine-based material decomposition images with Buvat I, Cook GJR, Davatzikos C, Depeursinge A,
dual-energy computed tomography imaging for pre- Desseroit M-C, Dinapoli N, Dinh CV, Echegaray S,
operatively predicting microsatellite instability status El Naqa I, Fedorov AY, Gatta R, Gillies RJ, Goh V,
in colorectal cancer. Front Oncol 9:1250. https://doi. Götz M, Guckenberger M, Ha SM, Hatt M, Isensee
org/10.3389/fonc.2019.01250 F, Lambin P, Leger S, Leijenaar RTH, Lenkowicz J,
Yu L, Fletcher JG, Grant KL, Carter RE, Hough DM, Lippert F, Losnegård A, Maier-Hein KH, Morin O,
Barlow JM, Vrtiska TJ, Williamson EE, Young PM, Müller H, Napel S, Nioche C, Orlhac F, Pati S, Pfaehler
Goss BC, Shiung M, Leng S, Raupach R, Schmidt EAG, Rahmim A, Rao AUK, Scherer J, Siddique
B, Flohr T, McCollough CH (2013) Automatic selec- MM, Sijtsema NM, Socarras Fernandez J, Spezi E,
tion of tube potential for radiation dose reduction Steenbakkers RJHM, Tanadini-Lang S, Thorwarth
in vascular and contrast-enhanced abdominopelvic D, Troost EGC, Upadhaya T, Valentini V, van Dijk
CT.  Am J Roentgenol 201:W297–W306. https://doi. LV, van Griethuysen J, van Velden FHP, Whybra
org/10.2214/AJR.12.9610 P, Richter C, Löck S (2020) The image biomarker
Zanca F, Demeter M, Oyen R, Bosmans H (2012) Excess standardization initiative: standardized quantitative
radiation and organ dose in chest and abdominal CT Radiomics for high-throughput image-­based pheno-
due to CT acquisition beyond expected anatomi- typing. Radiology:191145. https://doi.org/10.1148/
cal boundaries. Eur Radiol 22:779–788. https://doi. radiol.2020191145
org/10.1007/s00330-­011-­2332-­y
 Photon-Counting CT: Initial
Clinical Experience

Victor Mergen, André Euler, Kai Higashigaito,

Matthias Eberhard, and Hatem Alkadhi

Contents
1 Introduction   363
2 Technical Specifications   364
3 Cardiovascular Imaging   365
4 Abdominal Imaging   367
5 Lung Imaging   369
6 Skeletal Imaging   372
7 Conclusions   373
References   373

1 Introduction mation of every individually detected photon

enabling intrinsic spectral imaging in every CT
Photon-counting detector computed tomography scan.
(PCD-CT) is an emerging technology and repre- The clinical benefits of PCD-CTs are the elim-
sents the next milestone in CT developments. ination of classical electronic noise (Yu et  al.
Compared to conventional energy-integrating 2016; Sartoretti et al. 2020; Sartoretti et al. 2021;
detectors in which an indirect conversion tech- Symons et  al. 2018; Flohr et  al. 2020a; Ferda
nology is used to detect incident photons, PCD et  al. 2021), a higher spatial resolution (Flohr
technology uses semiconductors that directly et al. 2020a; Leng et al. 2016; Pourmorteza et al.
convert X-ray photons to an electrical signal. 2018; von Spiczak et  al. 2018; Mannil et  al.
This generated signal includes the energy infor- 2018), reduction of metal artifacts (Do et  al.
2020; Zhou et  al. 2019), improved iodine
contrast-­to-noise ratio (CNR) (Symons et  al.
V. Mergen · A. Euler · K. Higashigaito · M. Eberhard 2018; Flohr et  al. 2020a; Gutjahr et  al. 2016a;
H. Alkadhi (*)
Institute of Diagnostic and Interventional Radiology, Schmidt 2009; Symons et al. 2019), and improved
University Hospital Zurich, Zurich, Switzerland radiation dose efficiency (Symons et  al. 2019;
e-mail: [email protected]; [email protected]; Alkadhi and Euler 2020; Leng et  al. 2018).
[email protected]; [email protected]; Simultaneous multi-energy acquisition at a single
[email protected]

© The Author(s), under exclusive license to Springer Nature Switzerland AG 2022 363
H. Alkadhi et al. (eds.), Spectral Imaging, Medical Radiology Diagnostic Imaging,
https://doi.org/10.1007/978-3-030-96285-2_25
364 V. Mergen et al.

X-ray tube potential also permits new ways of secondary visible light photons. These are
advanced data processing. The energy of every absorbed by the photodiode and are converted
transmitted photon is allocated between multiple into an electrical signal. The intensity of the gen-
energy thresholds, called bins, leading to energy-­ erated electrical signal depends on the amount of
based attenuation profiles of tissue. This allows, incident photons and is proportional to the total
for example, for the simultaneous detection of energy deposited during a measurement interval.
one or more k-edge contrast agents. The specific Based on their detection principle, these detec-
visualization of exogenous contrast agent enables tors are called “energy-integrating detectors” and
single-scan multiphase imaging demonstrating a do not provide energy-resolved signals. The EIDs
new way of functional imaging (Tao et al. 2019; are separated by thin, optically intransparent col-
Si-Mohamed et al. 2019; Symons et al. 2017a). limator blades to prevent optical cross-talk (Flohr
In addition, virtual non-contrast reconstructions et al. 2020a; Alkadhi and Euler 2020; Leng et al.
can be computed obviating the need for addi- 2019; Willemink et al. 2018; Flohr et al. 2020b).
tional non-enhanced scans thus substantially Significant progress in the fields of detector
reducing the radiation dose of the examination materials, electronics and software allowed the
(Alkadhi and Euler 2020; Tao et al. 2019; Symons development of PCD-CTs during the last few
et  al. 2017a; Leng et  al. 2019). PCD-CT offers years. In 2021, Siemens Healthineers, Forchheim
new options for material decomposition and (Germany), presented the world’s first PCD-CT
quantification. For example, this novel imaging for full clinical use, the NAEOTOM Alpha, her-
technique is capable of quantifying bone mineral alding a new era of CT imaging.
density from localizer radiographs (Nowak et al. The underlying principle in PCD-CT is the
2021), assessing the macrophage burden in ath- use of a semiconductor diode capable of directly
erosclerotic plaques using gold nanoparticles converting the incident photons into an electrical
(Si-Mohamed et  al. 2021) and differentiating signal. Research focused on cadmium telluride
between blood and iodine (Riederer et al. 2019) (CdTe), cadmium zinc telluride (CdZnTe), and
or between calcium pyrophosphate and hydroxy- silicon (Si) as semiconductor material. Between
apatite deposits (Stamp et al. 2019). the cathode at the top of the thick layer of semi-
In April 2021, we got the first worldwide conductor material (1.4–30  mm depending on
PCD-CT system for full clinical use installed in material) and the pixelated anode at the bottom a
our radiology department. In this chapter, we strong electric field is applied. During imaging
briefly explain the technical aspects of PCD-CT the incident photon is absorbed in the photocon-
and present our first experience with this new CT ductor and creates an electron-hole pair. The
system with benefits highlighted with several electrons are immediately attracted by the anode
clinical case examples. For a detailed description and induce short currents of a few nanoseconds
of the underlying physical principles of PCD-CT (10−9 s). The height of the voltage pulse is directly
please refer to Chap. 6. proportional to the amount of absorbed charge
and is counted when exceeding a defined energy
threshold level. The strong electrical field
2 Technical Specifications between the cathode and the pixelated anode
obviates the need for collimator blades and
Current medical CT systems use energy-­ increases geometrical dose efficiency. Moreover,
integrating detectors (EID) in which the incident current pulses are counted only when exceeding a
photons are converted into an electrical signal preset energy threshold level, set above the elec-
during a two-step detection process. The EID tronic noise level, but lower than pulses gener-
consist of a scintillation crystal attached to a pho- ated by striking photons reducing electronic
todiode made of semiconducting material. noise in the generated images. By defining sev-
During the detection process, the incident X-rays eral threshold levels, PCD can assign the incom-
first strike the scintillation crystal and generate ing photons to precise energy bins, thus
Photon-Counting CT: Initial Clinical Experience 365

generating energy-based attenuation profiles of Table 1  Technical specifications of first-generation dual-­

source photon-counting CT
tissue (Tao et  al. 2019; Leng et  al. 2019;
Willemink et al. 2018; Flohr et al. 2020b). Detector type 2 × 6 cm photon-counting detector,
equipped with four energy thresholds
Current challenges with PCD-CT systems are
Rotation time 0.25 s
cross-talk and pile-up. Due to physical effects a Temporal 66 ms
single incident photon may be erroneously regis- resolution
tered in two different detectors referred to as Scan speed 737 mm/s
cross-talk. If the photon is absorbed at the border Tube type 2 × VECTRON tubes
of neighboring PCD charge sharing may occur Power 2 × 120 kW
during which the generated charge is detected by Bore size 82 cm
more than one PCD with only part of the original
energy registered in each. In addition, secondary 0.151 mm
photons may be generated during the absorption z S1 0.176 mm
process in the PCD, or because of Compton scat- M4 at iso-center
ter or in the form of fluorescent X-rays. These
secondary photons diverge randomly and can be
registered in neighbored detectors. Cross-talk
deteriorates the image’s spatial resolution and
contrast-to-noise ratio as well as its energy reso-
lution. During medical imaging, PCD need a fast
read-out of the incident photons. If consecutive
photons hit the PCD too closely in time, the elec- Fig. 1 Photon-counting detector layout of the
NAEOTOM Alpha, Siemens Healthineers, Forchheim,
trical pulses will superimpose on each other, a Germany
phenomenon called pulse pile-up. Pule pile-up
impairs image noise and measured signal as less
photons contribute to the image and influences with a z-coverage of 57.6  mm and a spectral
energy resolution (Flohr et  al. 2020a; Alkadhi image acquisition considering two energy levels
and Euler 2020; Tao et al. 2019; Willemink et al. is used. Applying the ultra-high-resolution mode,
2018; Flohr et al. 2020b). UHR mode, 120 pixels measuring 0.2 × 0.2 mm2
The NAEOTOM Alpha is the world’s first with a z-coverage of 24 mm and a spectral image
dual-source PCD-CT system for full clinical use acquisition considering two energy levels are
and was installed in April 2021 at our Department used. A third research mode is available using 96
in the University Hospital Zurich, Switzerland. macro-pixels and differentiating between four
This system uses cadmium telluride (CdTe) as energy thresholds.
semiconductor material. The X-ray tubes can be
operated at voltages up to 140 kVp, the tube cur-
rent can be set to values between 10 and 1300 mA 3 Cardiovascular Imaging
and the shortest rotation time of the system is
0.25  s. Further technical specifications of this The first-generation dual-source PCD-CT scanner
system can be found in Table  1. The PCD con- has the advantages of a third-generation dual-­
sists of sub-pixels, the so-called S1-pixels, mea- source CT with high temporal resolution (gantry
suring 0.2 × 0.2 mm2. Figure 1 depicts the precise rotation time 0.25  s, temporal resolution 66  ms)
detector layout. 2 × 2 sub-pixels form a “macro-­ and fast volume coverage (high pitch with
pixel,” the so-called M4-pixel, measuring 737  mm/s). In addition, the PCD-CT system
0.4 × 0.4 mm2. The detector provides two energy offers—through inherent multi-energy capabili-
thresholds per sub-pixel. The sub-pixels can be ties—the possibility of direct reconstruction of vir-
combined and read-out with different weightings tual monoenergetic images, virtual non-iodine (i.e.
(Table  2). In standard mode 144 macro-pixels non-enhanced) images, iodine maps, and of virtual
366 V. Mergen et al.

Table 2  Acquisition modes of the NAEOTOM Alpha, Siemens Healthineers, Forchheim, Germany
Detectors Detector size z-coverage
Standard mode 1376 M4—pixels 144 × 0.4 mm 57.6 mm
Ultra-high-resolution 2752 S1—pixels 120 × 0.2 mm 24 mm
(UHR) mode
Research mode 1376 M4—pixels 96 × 0.4 mm 38.4 mm

a b c

Fig. 2  Cardiac photon-counting detector CT in a 53-year-­ and (b) at 65  keV. (c) shows corresponding iodine map
old male patient with atypical chest pain. (a) Transverse from material decomposition
reformation of a virtual monoenergetic image at 45 keV

a b c d

Fig. 3  Photon-counting CT angiography of the chest and artifacts. (c) Transverse axial images at the level of the
abdomen in a 69-year-old male patient after endovascular kidneys showing a partially calcified abdominal aneurysm
repair. (a) Maximum intensity projection of virtual before and (d) after calcium subtraction (virtual non-­
monoenergetic images at 190 keV and (b) 3D cinematic calcium image)
rendering of the same dataset show the prosthesis without

non-calcium images from each dataset (Figs. 2 and subjects for imaging of the major arteries of the
3). Also, the multi-energy option opens the door to head and neck with PCD-CT and found an
the potential application of new contrast media in improved image quality of carotid and intracranial
cardiovascular CT imaging (Mergen et al. 2022). arteries along with fewer artifacts as compared to
Symons et al. (2019) showed in their study in a conventional single-energy CT with energy-inte-
calcium phantom, in ex vivo human hearts, and in grating detectors. In that study, a head phantom
asymptomatic volunteers the potential of PCD-CT was used to validate iodine concentration measure-
to improve the image quality of coronary artery ments in PCD-CT showing an excellent correlation
calcium scoring with the potential to reduce radia- between actual and measured iodine concentra-
tion dose at a constant image quality. In another tions and in addition a higher CNR in iodine maps
study, Symons et al. (2018) included asymptomatic compared with non-spectral PCD-CT images.
Photon-Counting CT: Initial Clinical Experience 367

Sartoretti et al. (2020) investigated the poten- yielded superior in-stent lumen delineation of
tial of a preclinical prototype PCD-CT scanner stents as compared to conventional scans with
with a tungsten-based contrast medium for energy-integrating detectors. These results could
carotid artery imaging using a human ex  vivo be further improved when using dedicated sharp
specimen with the aim of differentiating between tissue convolution kernels yielding superior qual-
the contrast-enhanced lumen and the calcified itative and quantitative image characteristics of
vessel wall. Authors could show that PCD-CT the in-stent lumen (von Spiczak et al. 2018).
employing the multi-energy bin option in combi- Our first study included 40 patients who
nation with tungsten as contrast medium enabled underwent clinical PCD-CT angiography of the
an improved carotid artery imaging with respect thoraco-abdominal aorta and who had a previous
to lumen and plaque visualization as well as CT angiography on a conventional EID-CT sys-
image noise. In another study, Sartoretti et  al. tem with automatic tube voltages selection (Euler
(2021) systematically evaluated the potential of et al. 2021). Radiation dose at both CT scanners
PCD-CT for investigational contrast media for was kept intentionally at the same level. We
subtraction of calcified plaques in a small vessel found that high-pitch PCD-CT angiography of
phantom. Five contrast media with iodine, bis- the aorta with reconstruction of virtual monoen-
muth, hafnium, holmium, and tungsten at equal ergetic images at 40 and 45 keV resulted in a sig-
mass concentrations were tested, and authors nificantly higher contrast-to-noise-ratio, an effect
found that contrast maps with tungsten and, to a which was pronounced in overweight patients.
lesser extent, with hafnium yielded superior This indicates the potential for further radiation
image noise properties and improved vessel dose and/or contrast media volume reductions
lumen visualization including an improved sub- with PCD-CT.
jective image quality as compared with the refer-
ence standard with iodinated contrast media.
Si-Mohamed et  al. (2021) performed k-edge 4 Abdominal Imaging
imaging with a prototype PCD-CT system using
both iodinated contrast media and gold nanopar- Dual-energy CT has shown beneficial applications
ticles to detect and quantify the macrophage bur- in abdominal imaging in the last decades with
den within the atherosclerotic aortas of rabbits improvements in, e.g., lesion detection or conspi-
in  vivo. Authors found that PCD-CT imaging cuity (Shuman et  al. 2014; Darras et  al. 2019),
with gold nanoparticles allowed for the noninva- lesion characterization (Nagayama et  al. 2020),
sive evaluation of both molecular and anatomic tumor treatment response monitoring (Parakh
information of atherosclerotic plaques. et  al. 2018; Jiang et  al. 2017; Aoki et  al. 2016),
Dangelmeier et al. (2018) evaluated in an ex vivo evaluation of liver parenchyma (Marri et al. 2021;
phantom study simulating an abdominal aortic Elbanna et  al. 2020), detection of gallstones
aneurysm a mixture of iodine and gadolinium (Uyeda et al. 2017), or urinary stone characteriza-
and could show that PCD-CT was able to differ- tion (Zheng et  al. 2016; Habashy et  al. 2016).
entiate the distributions within different compart- However, dual-energy CT exams commonly have
ments filled with iodine, gadolinium and with to be prospectively ordered and protocolled, par-
calcifications. ticularly on the widely prevalent scanners using
The potential of higher spatial resolution of fast kVp-switching or dual-source dual-energy
PCD-CT deserves further note. In a phantom CT.  The introduction of PCD-CT improves the
study, Mannil et al. (2018) included 18 different workflow of CT protocolling because spectral
coronary stent types, filled with iodinated con- information is available from each image acquisi-
trast media, and imaged in different orientations tion and can be retrospectively accessed and
with a prototype PCD-CT scanner. At matched reconstructed if needed. This saves time and
CT scan protocol settings and identical image resources for CT protocolling and improves stan-
reconstruction parameters, the PCD system dardization. This holds great potential to improve
368 V. Mergen et al.

55  keV to improve contrast in vascular imaging

and 60  keV to improve CNR in parenchymal
imaging). Our first study included 39 patients
who underwent clinical abdominal PCD-CT in
the portal venous phase who also had a foregoing
abdominal CT on a conventional EID-CT system
with automatic tube voltages selection
(Higashigaito et al. 2021). Radiation dose between
systems was intentionally kept constant. Results
indicated that PCD-CT with reconstruction of
VMI at 50 keV yielded significantly higher con-
trast-to-noise-ratio in various abdominal organs
and vessels as compared to EID-CT. Another
recent study showed that image quality of portal-
venous phase PCD-CT is further improved by
applying the new iterative reconstruction tech-
nique of the scanner (i.e., quantum iterative recon-
struction) (Sartoretti et al. 2022b).
In addition, the reconstruction of polychro-
Fig. 4  Iodine map of the right liver lobe showing a hem- matic images, the so-called T3D images, is avail-
angioma with increased iodine concentration measured at able. These images are reconstructed by using the
the lesion’s periphery (4.5  mg iodine per mL) as com- data of the lowest energy threshold at 20 keV in a
pared to the normal liver parenchyma (2.5 mg iodine per
mL). With PCD-CT, the reconstruction of iodine maps has conventional, non-spectral image reconstruction
become available from each scan and aim to emulate the image perception of a
conventional polychromatic image at
120 kV. Compared to EID-CT, PCD-CT has the
the characterization of incidental lesions, avoid advantage to decrease image noise at low X-ray
additional imaging, and decrease patient anxiety. photon flux by excluding electronic noise
Spectral imaging is automatically available in the (Willemink et al. 2018). This potentially benefits
standard scan mode, the so-called Quantum Plus imaging at very low radiation doses and in large-­
mode, with a collimation of 144 × 0.4 mm and cur- sized patients.
rently utilizing two energy levels . PCD-CT has the unique capability to dis-
Classic post-processing options, e.g., virtual criminate among different contrast agents, e.g.,
non-contrast reconstructions (Sartoretti et al. iodine, gadolinium, and bismuth (Tao et  al.
2022a), iodine maps (Fig. 4), liver iron or fat frac- 2019; Ren et al. 2020). This enables simultane-
tion estimation as well as characterization of kid- ous multi-contrast agent imaging (Si-Mohamed
ney stones or gout are available on PCD-CT and et  al. 2019; Symons et  al. 2017a; Willemink
might benefit from the improved spectral separa- et  al. 2018). This in combination with a split-
tion provided by the novel detector. First preclini- bolus injection and virtual non-contrast imag-
cal studies have indicated improved ing could be used to transform traditional
characterization of small urinary stones of less three-phase exams on traditional EID-CT
than 3  mm as compared to dual-energy EID-CT (unenhanced, arterial, portal venous phase) to a
(Marcus et  al. 2018). In addition to these tradi- single-phase acquisition on PCD-CT
tional options, the vendor has introduced the rou- (Willemink et  al. 2018). To date, however,
tine use of virtual monoenergetic images (VMIs) multi-contrast imaging on PCD-CT has been
as the routine diagnostic reading series (Fig. 5). limited to animal models due to the current off-
VMIs-energies (in keV) are optimized based label use of different contrast agents in CT
on the desired imaging task and body region (e.g., imaging of humans.
Photon-Counting CT: Initial Clinical Experience 369

Fig. 5  Axial virtual monoenergetic CT images at differ- substantially increasing CT attenuation of the ring
ent energies of a 55-year-old male patient with an intrahe- enhancement of the lesion with decreasing keV level
patic cholangiocarcinoma of the left liver lobe. Note the

5 Lung Imaging energy-integrating detector CT (Ferda et  al.

2021). Symons et al. showed in a study with 30
CT has a fundamental role in noninvasive imag- asymptomatic volunteers that PCD-CT chest CT
ing of the chest. Within a single breath-hold, CT images have an approximately 20% higher CNR
enables the acquisition of isotropic, high-­ of lung nodules compared to conventional
resolution images providing detailed anatomic energy-integrating detector CT (Symons et  al.
information about thoracic disease. Inherent 2017b). Lower noise levels may facilitate radia-
opportunities of the PCD-CT, such as availability tion dose reduction with PCD-CT, which may be
of spectral data, ultra-high-resolution imaging, especially beneficial in lung cancer screening
and reduction of image noise may further improve examinations or follow-up chest CT. Comparable
this technique (Flohr et  al. 2020a; Willemink to the use of tin filtration on a dual-source chest
et al. 2018). CT, the application of tin filters may further lower
In the last two decades evolution of scanner noise levels with the potential for radiation dose
technology and iterative image reconstruction reduction (Martini et al. 2015).
have led to an improvement in diagnostic quality As stated above, the photon-counting detector
of chest CT while being able to reduce radiation enables the possibility to scan with an ultra-high-­
dose and noise levels. In contrast to energy-­ resolution mode, with 120 pixels measuring
integrating detector CT images, the fraction of 0.2 × 0.2 mm2. In contrast to an energy-integrated
lower energy photons contributing to PCD-CT detector, there is no need for a mechanical sepa-
images is higher due to the photon binning pro- ration of detector cells (Flohr et  al. 2020a).
cess (Farhadi et al. 2021). Moreover, the exclu- Bartlett et al. (2019) as well as Ferda et al. (2021)
sion of high-energy photons in monoenergetic showed an improved visualization and delinea-
image reconstruction further improve image con- tion of third-, fourth- and fifth-order bronchi
trast and CNR in PCD-CT compared to energy-­ using ultra high-resolution lung PCD-CT with a
integrating detector CT (Symons et  al. 2019). 1024 image matrix reconstruction compared to
Cadaver studies (Gutjahr et al. 2016a) as well as conventional chest EID-CT (Fig. 6). As a draw-
first patient experience with a single-source, full back, increased spatial resolution is inherently
field-of-view PCD-CT have shown significantly accompanied by increased noise or the need for
lower image noise with PCD-CT compared to higher radiation doses to maintain noise levels of
370 V. Mergen et al.

a b

c d

Fig. 6 Comparison of photon-counting detector CT CTDIvol, 1.25  mGy; Kernel, Bl64; Matrix 512  ×  512).
images (PCD-CT; left column) and conventional energy-­ The bottom row shows images of an 83-year-old female
integrating detector CT images (EID-CT; right column). patient with systemic sclerosis. The left image (c) was
The upper row shows images of a 46-year-old female acquired with a dual-source PCD-CT (NAEOTOM Alpha,
patient with mixed connective tissue disease and basal Siemens); tube voltage, 100  kV with tin filtration;
predominant lung fibrosis. The left image (a) was acquired CTDIvol, 0.51 mGy; Kernel, BI64; Matrix 512 × 512), the
with a dual-source PCD-CT (NAEOTOM Alpha, right image (d) was acquired with a third-generation,
Siemens); tube voltage, 100  kV with tin filtration; dual-source EID-CT (SOMATOM Force, Siemens); tube
CTDIvol, 0.68 mGy; Kernel, BI64; Matrix 512 × 512), the voltage, 100  kV; CTDIvol, 1.70  mGy; Kernel, Bl64;
right image (b) was acquired with a single-source EID-CT Matrix 512 × 512). Note the improved spatial resolution
(SOMATOM Edge Plus, Siemens); tube voltage, 100kV; of PCD-CT despite of lower radiation doses
Photon-Counting CT: Initial Clinical Experience 371

a b

Fig. 7 Images of a 43-year-old female patient with subsegmental branch of the posterobasal segmental artery
chronic thromboembolic pulmonary hypertension under- (green arrow) and rarefication of subsegmental arteries of
going photon-counting detector CT pulmonary angiogra- the apico-posterior segment of the left lung (blue arrow).
phy. Maximum intensity reconstruction of iodine-weighted The iodine map (b) illustrates the corresponding segmen-
images (a) illustrates the sudden change of caliber of a tal perfusion defects

images with lower spatial resolution. Future thromboembolism. The calculation of iodine
research may show whether improved evaluation maps may assist in detecting subsegmental perfu-
of the morphology of lung nodules and lung sion defects (Fig. 7). Masy et al. have shown that
masses may improve differentiation between dual-energy CT-derived iodine maps for the
benign and malignant as well as characterization assessment of lung perfusion show excellent
of lesions and outweigh the potential drawbacks. agreement with V/Q scintigraphy in diagnosing
In our experimental study using an anthromor- chronic thromboembolic pulmonary hypertension
phic chest phantom containing various sized pul- with a kappa of 0.8 (Masy et  al. 2018). In their
monary nodules we found that image quality of study, the combination of CT pulmonary angiog-
PCD-CT was superior to EID-CT while showing raphy and iodine maps enabled to correctly diag-
comparably lower image noise (Jungblut et  al. nose all patients with chronic thromboembolic
2021). pulmonary hypertension (Masy et  al. 2018).
Directly counting the energy of an incident Furthermore, spectral information in CT pulmo-
photon, PCD-CT can be utilized for material nary angiography may enable the reduction of
decomposition of tissues with equal densities but contrast media by providing low keV monoener-
different elemental composition (Flohr et  al. getic images with improved contrast attenuation
2020a). Spectral information may help to amplify (Meier et al. 2016). Using dual-­energy CT, Meier
subtle attenuation differences as well as the distri- et al. could show that CT pulmonary angiography
bution of enhancement for detection and charac- is feasible with only 6 g of iodine corresponding
terization of lung nodules and tumors (Chae et al. to 15 ml of contrast media at a concentration of
2010). In chest imaging, material decomposition 400 mg iodine/ml (Meier et al. 2016).
is frequently used in the evaluation of pulmonary CT pulmonary angiography is especially sus-
vessels in patients with suspected acute or chronic ceptible to the transient interruption of contrast
372 V. Mergen et al.

inflow from the superior vena cava due to the such as monosodium urate detection in gout
increase of intrathoracic pressure by patients tak-arthropathy, detection of bone marrow edema in
ing a deep breath before scan initiation. In these fractures or metal artifact reduction using virtual
cases, advanced monoenergetic reconstructions monoenergetic images (Rajiah et  al. 2019). In
may help to increase the attenuation of pulmo- theory, PCD-CT enables perfect spectral separa-
nary arteries compared to single-energy scans tion which would result in superior performance
and avoid unnecessary repetition of CT studies in of those applications compared to dual-energy
patients with suboptimal opacification of pulmo- CT. However, in practice, spectral separation of
nary arteries. PCD-CT is impaired by various effects such as
K-escape, charge sharing or detector polarization
(Gutjahr et  al. 2016b; Taguchi and Iwanczyk
6 Skeletal Imaging 2013). So far, only limited literature is available
about the performance of PCT-CT in musculo-
The novel PCD-CT system offers various bene- skeletal imaging. Stamp et al. were able to detect
fits in musculoskeletal imaging such as the con- monosodium urate with PCT-CT using a proto-
stant availability of spectral information or the type scanner in a cadaver finger. Recently, Zhou
higher spatial resolution of bone structures using et  al. demonstrated the superior metal artifact
(UHR) mode. reduction capability of PCD-CT using tin filter in
The main advantage of PCD-CT in musculo- combination with high-energy threshold com-
skeletal imaging is the constant availability of pared to EID-CT (Zhou et  al. 2019). Our first
spectral information. Spectral information from clinical experience of PCD-CT in musculoskele-
PCD-CT can be used for various advanced imag- tal imaging is limited to virtual non-calcium
ing processing application similar to dual-energy imaging and gout detection. Using virtual non-­
CT. In musculoskeletal imaging, numerous dual-­ calcium imaging, we were able to reliably detect
energy applications are already well established, bone marrow edema adjacent to fractures (Fig. 8).

a b c d

Fig. 8  Photon-counting detector CT of the lower extrem- in coronal (c) and sagittal (d) plane. Note the bone mar-
ity of a 28-year-old male patient in coronal (a) and sagittal row edema (white arrow) adjacent to the fractures
plane (b) with corresponding virtual non-calcium images
Photon-Counting CT: Initial Clinical Experience 373

Another promising advantage of PCD-CT is material decomposition and quantification. This

the so-called ultra-high-resolution (UHR) mode. feature allows for the identification of contrast
High spatial resolution of CT plays a crucial role media and the computation of virtual non-­
in the assessment of delicate osseous structures contrast reconstructions or iodine maps.
such as the incudo-mallear joint in the temporal Moreover, precise material characterization is
bone. Spatial resolution in conventional EID-CT capable of identifying the composition of, e.g.,
is limited due to the larger size of the detector cell urinary stones or differentiating between crystal
and the requirement of septa between each cell. arthropathies. Feasibility of CT imaging simulta-
To overcome this limitations, spatial resolution in neously using different k-edge contrast agents or
EID-CT can be improved by using an additional targeted nanoparticles has already been proved in
attenuation comb filter in front of the detector experimental studies but is yet to be assessed in
which reduces the detector aperture but also clinical studies. Undoubtedly, this new imaging
decreases dose efficiency of EID (Leng et  al. approach will reveal previously unseen informa-
2015). In contrast, PCD is equipped with smaller tion and offer a new dimension of functional
detector cells (0.2 × 0.2 mm2) and has no require- imaging.
ment for septa or attenuation comb which results In conclusion, PCD-CT will push the bound-
in higher spatial resolution and higher dose effi- aries of future CT imaging and opens new, cur-
ciency compared to EID.  In temporal bone CT, rently unpredictable areas of application.
higher spatial resolution and higher dose effi-
ciency of prototype PCD-CT compared to Compliance with Ethical Standards
EID-CT has been already demonstrated (Zhou
et al. 2018; Rajendran et al. 2020). For example, Disclosure of Interests The manuscript has not been
Zhou et  al. demonstrated a substantially better submitted to more than one publication simultaneously.
The submitted work is original and has not been published
delineation of the anatomy of the temporal bones elsewhere in any form or language (partial or complete).
of ten cadavers scanned on PCD-CT with UHR The authors have followed the rules for obtaining, select-
mode compared to EID-CT using identical radia- ing, and processing image data and have not presented any
tion dose (Zhou et al. 2018) and Rajendran et al. data or text of other authors as if they were their own.
demonstrated a radiation dose reduction of
greater than 80% in temporal bone CT using
PCD-CT compared to EID-CT maintaining simi- References
lar image quality (Rajendran et al. 2020).
Alkadhi H, Euler A (2020) The future of computed tomog-
raphy: personalized, functional, and precise. Investig
Radiol 55(9):545–555. https://doi.org/10.1097/
7 Conclusions RLI.0000000000000668
Aoki M, Hirose K, Sato M, Akimoto H, Kawaguchi H,
Hatayama Y, Fujioka I, Tanaka M, Ono S, Takai Y
Our initial experience with the first PCD-CT sys- (2016) Prognostic impact of average iodine density
tem for full clinical use shows promising results assessed by dual-energy spectral imaging for pre-
in several clinical application fields. The low dicting lung tumor recurrence after stereotactic body
radiotherapy. J Radiat Res 57(4):381–386. https://doi.
electronic noise and high spatial resolution org/10.1093/jrr/rrv100
improve diagnostic accuracy especially in cardio- Bartlett DJ, Koo CW, Bartholmai BJ, Rajendran K,
vascular, lung, and musculoskeletal imaging. The Weaver JM, Halaweish AF, Leng S, McCollough CH,
superior iodine CNR is particularly valuable in Fletcher JG (2019) High-resolution chest computed
tomography imaging of the lungs: impact of 1024
abdominal imaging depicting lesions of paren- matrix reconstruction and photon-counting detector
chymatous organs as well as in cardiovascular computed tomography. Investig Radiol 54(3):129–
imaging increasing vessel attenuation. Large 137. https://doi.org/10.1097/RLI.0000000000000524
potential of PCD-CT lies in the intrinsic spectral Chae EJ, Song JW, Krauss B, Song KS, Lee CW, Lee
HJ, Seo JB (2010) Dual-energy computed tomog-
imaging properties in every CT scan enabling
374 V. Mergen et al.

raphy characterization of solitary pulmonary nod- Gutjahr R, Halaweish AF, Yu Z, Leng S, Yu L, Li Z,

ules. J Thorac Imaging 25(4):301–310. https://doi. Jorgensen SM, Ritman EL, Kappler S, McCollough
org/10.1097/RTI.0b013e3181e16232 CH (2016a) Human imaging with photon-counting-­
Dangelmaier J, Bar-Ness D, Daerr H, Muenzel D, based CT at clinical dose levels: contrast-to-noise ratio
Si-Mohamed S, Ehn S, Fingerle AA, Kimm MA, and cadaver studies. Investig Radiol 51(7):421. https://
Kopp FK, Boussel L, Roessl E, Pfeiffer F, Rummeny doi.org/10.1097/RLI.0000000000000251
EJ, Proksa R, Douek P, Noël PB (2018) Experimental Gutjahr R, Halaweish AF, Yu Z, Leng S, Yu L, Li Z,
feasibility of spectral photon-counting computed Jorgensen SM, Ritman EL, Kappler S, McCollough
tomography with two contrast agents for the detec- CH (2016b) Human imaging with photon counting-­
tion of endoleaks following endovascular aortic repair. based computed tomography at clinical dose levels:
Eur Radiol 28(8):3318–3325. https://doi.org/10.1007/ contrast-to-noise ratio and cadaver studies. Investig
s00330-­017-­5252-­7 Radiol 51(7):421–429. https://doi.org/10.1097/
Darras KE, Clark SJ, Kang H, Mohammed MF, Barrett rli.0000000000000251
S, Chang SD, Harris AC, Nicolaou S, McLaughlin Habashy D, Xia R, Ridley W, Chan L, Ridley L
PD (2019) Virtual monoenergetic reconstruc- (2016) Impact of dual energy characteriza-
tion of contrast-­enhanced CT scans of the abdo- tion of urinary calculus on management. J Med
men and pelvis at 40 keV improves the detection Imaging Radiat Oncol 60(5):624–631. https://doi.
of peritoneal metastatic deposits. Abdom Radiol org/10.1111/1754-­9485.12497
(NY) 44(2):422–428. https://doi.org/10.1007/ Higashigaito K, Euler A, Eberhard M, Flohr TG, Schmidt
s00261-­018-­1733-­7 B, Alkadhi H (2021) Contrast-enhanced abdominal
Do T, Sawall S, Heinze S, Reiner T, Ziener C, Stiller W, CT with clinical photon-counting detector CT: assess-
Schlemmer H, Kachelrieß M, Kauczor H, Skornitzke ment of image quality and comparison with energy-­
S (2020) A semi-automated quantitative compari- integrating detector CT.  Acad Radiol. https://doi.
son of metal artifact reduction in photon-counting org/10.1016/j.acra.2021.06.018
computed tomography by energy-selective thresh- Jiang C, Yang P, Lei J, Li J, Yan K, Li F, Yan R, Xia L
olding. Sci Rep 10(1):1–10. https://doi.org/10.1038/ (2017) The application of iodine quantitative infor-
s41598-­020-­77904-­3 mation obtained by dual-source dual-energy com-
Elbanna KY, Mansoori B, Mileto A, Rogalla P, Guimarães puted tomography on Chemoradiotherapy effect
SL (2020) Dual-energy CT in diffuse liver disease: is monitoring for cervical cancer: a preliminary study.
there a role? Abdom Radiol (NY) 45(11):3413–3424. J Comput Assist Tomogr 41(5):737–745. https://doi.
https://doi.org/10.1007/s00261-­020-­02702-­4 org/10.1097/rct.0000000000000603
Euler A, Higashigaito K, Mergen V, Sartoretti T, Zanini Jungblut L, Blüthgen C, Polacin M, Messerli M, Schmidt
B, Schmidt B, Flohr TG, Ulzheimer S, Eberhard M, B, Euler A, Alkadhi H, Frauenfelder T, Martini K
Alkadhi H (2021) High-pitch photon-counting detec- (2021) First performance evaluation of an artificial
tor computed tomography angiography of the aorta: intelligence–based computer-aided detection sys-
Intraindividual comparison to energy-integrating tem for pulmonary nodule evaluation in dual-source
detector computed tomography at equal radia- photon-counting detector CT at different low-dose
tion dose. Investig Radiol. https://doi.org/10.1097/ levels. Investig Radiol. https://doi.org/10.1097/
rli.0000000000000816 rli.0000000000000814
Farhadi F, Rajagopal JR, Nikpanah M, Sahbaee P, Leng S, Bruesewitz M, Tao S, Rajendran K, Halaweish
Malayeri AA, Pritchard WF, Samei E, Jones EC, AF, Campeau NG, Fletcher JG, McCollough CH
Chen MY (2021) Review of technical advance- (2019) Photon-counting detector CT: system design
ments and clinical applications of photon-counting and clinical applications of an emerging technol-
computed tomography in imaging of the thorax. J ogy. Radiographics 39(3):729–743. https://doi.
Thorac Imaging 36(2):84–94. https://doi.org/10.1097/ org/10.1148/rg.2019180115
RTI.0000000000000569 Leng S, Diehn FE, Lane JI, Koeller KK, Witte RJ, Carter
Ferda J, Vendiš T, Flohr T, Schmidt B, Henning A, RE, McCollough CH (2015) Temporal bone CT:
Ulzheimer S, Pecen L, Ferdová E, Baxa J, Mírka improved image quality and potential for decreased radi-
H (2021) Computed tomography with a full FOV ation dose using an ultra-high-resolution scan mode with
photon-­ counting detector in a clinical setting, the an iterative reconstruction algorithm. Am J Neuroradiol
first experience. Eur J Radiol 137:109614. https://doi. 36(9):1599–1603. https://doi.org/10.3174/ajnr.A4338
org/10.1016/j.ejrad.2021.109614 Leng S, Rajendran K, Gong H, Zhou W, Halaweish
Flohr T, Petersilka M, Henning A, Ulzheimer S, Ferda AF, Henning A, Kappler S, Baer M, Fletcher JG,
J, Schmidt B (2020a) Photon-counting CT review. McCollough CH (2018) 150-μm spatial resolution
Phys Med 79:126–136. https://doi.org/10.1016/j. using photon-counting detector computed tomography
ejmp.2020.10.030 technology: technical performance and first patient
Flohr T, Ulzheimer S, Petersilka M, Schmidt B (2020b) images. Investig Radiol 53(11):655–662. https://doi.
Basic principles and clinical potential of photon-­ org/10.1097/RLI.0000000000000488
counting detector CT. Chin J Acad Radiol 3(1):19–34. Leng S, Yu Z, Halaweish A, Kappler S, Hahn K, Henning
https://doi.org/10.1007/s42058-­020-­00029-­z A, Li Z, Lane J, Levin DL, Jorgensen S (2016) Dose-­
Photon-Counting CT: Initial Clinical Experience 375

efficient ultrahigh-resolution scan mode using a pho- Parakh A, Patino M, Muenzel D, Kambadakone A,
ton counting detector computed tomography system. Sahani DV (2018) Role of rapid kV-switching dual-­
J Med Imag 3(4):043504. https://doi.org/10.1117/1. energy CT in assessment of post-surgical local
JMI.3.4.043504 recurrence of pancreatic adenocarcinoma. Abdom
Mannil M, Hickethier T, von Spiczak J, Baer M, Henning Radiol (NY) 43(2):497–504. https://doi.org/10.1007/
A, Hertel M, Schmidt B, Flohr T, Maintz D, Alkadhi s00261-­017-­1390-­2
H (2018) Photon-counting CT: high-resolution imag- Pourmorteza A, Symons R, Henning A, Ulzheimer S,
ing of coronary stents. Investig Radiol 53(3):143–149. Bluemke DA (2018) Dose efficiency of quarter-­
https://doi.org/10.1097/rli.0000000000000420 millimeter photon-counting computed tomography:
Marcus RP, Fletcher JG, Ferrero A, Leng S, Halaweish first-in-human results. Investig Radiol 53(6):365–372.
AF, Gutjahr R, Vrtiska TJ, Wells ML, Enders FT, https://doi.org/10.1097/RLI.0000000000000463
McCollough CH (2018) Detection and character- Rajendran K, Voss BA, Zhou W, Tao S, DeLone DR,
ization of renal stones by using photon-counting-­ Lane JI, Weaver JM, Carlson ML, Fletcher JG,
based CT.  Radiology 289(2):436–442. https://doi. McCollough CH, Leng S (2020) Dose reduction for
org/10.1148/radiol.2018180126 sinus and temporal bone imaging using photon-count-
Marri UK, Das P, Shalimar M, Kalaivani DNS, ing detector CT with an additional tin filter. Investig
Madhusudhan KS (2021) Noninvasive staging of Radiol 55(2):91–100. https://doi.org/10.1097/
liver fibrosis using 5-minute delayed dual-energy CT: rli.0000000000000614
comparison with US Elastography and correlation Rajiah P, Sundaram M, Subhas N (2019) Dual-energy CT
with histologic findings. Radiology 298(3):600–608. in musculoskeletal imaging: what is the role beyond
https://doi.org/10.1148/radiol.2021202232 gout? AJR Am J Roentgenol 213(3):493–505. https://
Martini K, Higashigaito K, Barth BK, Baumueller S, Alkadhi doi.org/10.2214/ajr.19.21095
H, Frauenfelder T (2015) Ultralow-dose CT with tin Ren L, Rajendran K, Fletcher JG, McCollough CH, Yu
filtration for detection of solid and sub solid pulmonary L (2020) Simultaneous dual-contrast imaging of
nodules: a phantom study. Br J Radiol 88(1056):20150389. small bowel with iodine and bismuth using photon-­
https://doi.org/10.1259/bjr.20150389 counting-­detector computed tomography: a feasibility
Masy M, Giordano J, Petyt G, Hossein-Foucher C, animal study. Investig Radiol 55(10):688–694. https://
Duhamel A, Kyheng M, De Groote P, Fertin M, doi.org/10.1097/rli.0000000000000687
Lamblin N, Bervar JF, Remy J, Remy-Jardin M (2018) Riederer I, Si-Mohamed S, Ehn S, Bar-Ness D, Noël
Dual-energy CT (DECT) lung perfusion in pulmonary PB, Fingerle AA, Pfeiffer F, Rummeny EJ, Douek P,
hypertension: concordance rate with V/Q scintigraphy Pfeiffer D (2019) Differentiation between blood and
in diagnosing chronic thromboembolic pulmonary iodine in a bovine brain—initial experience with spec-
hypertension (CTEPH). Eur Radiol 28(12):5100– tral photon-counting computed tomography (SPCCT).
5110. https://doi.org/10.1007/s00330-­018-­5467-­2 PLoS One 14(2):e0212679. https://doi.org/10.1371/
Meier A, Higashigaito K, Martini K, Wurnig M, Seifert journal.pone.0212679
B, Keller DI, Frauenfelder T, Alkadhi H (2016) Sartoretti T, Eberhard M, Nowak T, Gutjahr R, Jost
Dual energy CT pulmonary angiography with 6g G, Pietsch H, Schmidt B, Flohr T, Alkadhi H,
iodine-a propensity score-matched study. PLoS One Euler A (2021) Photon-counting multienergy com-
11(12):e0167214. https://doi.org/10.1371/journal. puted tomography with spectrally optimized con-
pone.0167214 trast Media for Plaque Removal and Stenosis
Mergen V, Sartoretti T, Klotz E, Schmidt B, Jungblut Assessment. Investig Radiol. https://doi.org/10.1097/
L, Higashigaito K, Manka R, Euler A, Kasel M, rli.0000000000000773
Eberhard M, Alkadhi H (2022) Extracellular vol- Sartoretti T, Eberhard M, Rüschoff JH, Pietsch H, Jost
ume quantification with cardiac late enhancement G, Nowak T, Schmidt B, Flohr T, Euler A, Alkadhi
scanning using dual-source photon-counting detec- H (2020) Photon-counting CT with tungsten as con-
tor CT. Invest Radiol. https://doi.org/10.1097/ trast medium: experimental evidence of vessel lumen
RLI.0000000000000851. Epub ahead of print. PMID: and plaque visualization. Atherosclerosis 310:11–16.
35066531 https://doi.org/10.1016/j.atherosclerosis.2020.07.023
Nagayama Y, Inoue T, Oda S, Tanoue S, Nakaura T, Sartoretti T, Mergen V, Higashigaito K, Eberhard M,
Ikeda O, Yamashita Y (2020) Adrenal adenomas Alkadhi H, Euler A (2022a) Virtual noncontrast
versus metastases: diagnostic performance of dual-­ imaging of the liver using photon-counting detector
energy spectral CT virtual noncontrast imaging and computed tomography: a systematic phantom and
iodine maps. Radiology 296(2):324–332. https://doi. patient study. Invest Radiol. https://doi.org/10.1097/
org/10.1148/radiol.2020192227 RLI.0000000000000860. Epub ahead of print. PMID:
Nowak T, Eberhard M, Schmidt B, Frey D, Distler O, 35136003
Saltybaeva N, Alkadhi H, Euler A (2021) Bone mineral Sartoretti T, Landsmann A, Nakhostin D, Eberhard M,
density quantification from localizer radiographs: accu- Röeren C, Mergen V, Higashigaito K, Raupach R,
racy and precision of energy-integrating detector CT and Alkadhi H, Euler A (2022b) Quantum ­ iterative
photon-counting detector CT.  Radiology 298(1):147– reconstruction for abdominal photon-counting
152. https://doi.org/10.1148/radiol.2020202767 detector ct improves image quality. Radiology.
376 V. Mergen et al.

https://doi.org/10.1148/radiol.211931. Epub ahead of vascular imaging of the head and neck: first in  vivo
print. PMID: 35103540 human results. Investig Radiol 53(3):135–142. https://
Schmidt TG (2009) Optimal “image-based” weighting doi.org/10.1097/rli.0000000000000418
for energy-resolved CT. Med Phys 36(7):3018–3027. Symons R, Sandfort V, Mallek M, Ulzheimer S,
https://doi.org/10.1118/1.3148535 Pourmorteza A (2019) Coronary artery calcium scor-
Shuman WP, Green DE, Busey JM, Mitsumori LM, Choi ing with photon-counting CT: first in  vivo human
E, Koprowicz KM, Kanal KM (2014) Dual-energy experience. Int J Cardiovasc Imaging 35(4):733–739.
liver CT: effect of monochromatic imaging on lesion https://doi.org/10.1007/s10554-­018-­1499-­6
detection, conspicuity, and contrast-to-noise ratio of Taguchi K, Iwanczyk JS (2013) Vision 20/20: sin-
hypervascular lesions on late arterial phase. AJR Am J gle photon counting x-ray detectors in medical
Roentgenol 203(3):601–606. https://doi.org/10.2214/ imaging. Med Phys 40(10):100901. https://doi.
ajr.13.11337 org/10.1118/1.4820371
Si-Mohamed S, Tatard-Leitman V, Laugerette A, Sigovan Tao S, Rajendran K, McCollough CH, Leng S (2019)
M, Pfeiffer D, Rummeny EJ, Coulon P, Yagil Y, Feasibility of multi-contrast imaging on dual-source
Douek P, Boussel L (2019) Spectral photon-counting photon counting detector (PCD) CT: an initial phan-
computed tomography (SPCCT): in-vivo single-­ tom study. Med Phys 46(9):4105–4115. https://doi.
acquisition multi-phase liver imaging with a dual org/10.1002/mp.13668
contrast agent protocol. Sci Rep 9(1):1–8. https://doi. Uyeda JW, Richardson IJ, Sodickson AD (2017) Making
org/10.1038/s41598-­019-­44821-­z the invisible visible: improving conspicuity of noncal-
Si-Mohamed SA, Sigovan M, Hsu JC, Tatard-Leitman V, cified gallstones using dual-energy CT. Abdom Radiol
Chalabreysse L, Naha PC, Garrivier T, Dessouky R, (NY) 42(12):2933–2939. https://doi.org/10.1007/
Carnaru M, Boussel L, Cormode DP, Douek PC (2021) s00261-­017-­1229-­x
In vivo molecular K-edge imaging of atherosclerotic Willemink MJ, Persson M, Pourmorteza A, Pelc
plaque using photon-counting CT. Radiology:203968. NJ, Fleischmann D (2018) Photon-counting
https://doi.org/10.1148/radiol.2021203968 CT: technical principles and clinical prospects.
von Spiczak J, Mannil M, Peters B, Hickethier T, Baer Radiology 289(2):293–312. https://doi.org/10.1148/
M, Henning A, Schmidt B, Flohr T, Manka R, Maintz radiol.2018172656
D, Alkadhi H (2018) Photon counting computed Yu Z, Leng S, Kappler S, Hahn K, Li Z, Halaweish
tomography with dedicated sharp convolution kernels: AF, Henning A, McCollough CH (2016) Noise per-
tapping the potential of a new Technology for Stent formance of low-dose CT: comparison between an
Imaging. Investig Radiol 53(8):486–494. https://doi. energy integrating detector and a photon counting
org/10.1097/rli.0000000000000485 detector using a whole-body research photon count-
Stamp LK, Anderson NG, Becce F, Rajeswari M, Polson ing CT scanner. J Med Imag 3(4):043503. https://doi.
M, Guyen O, Viry A, Choi C, Kirkbride TE, Raja AY org/10.1117/1.JMI.3.4.043503
(2019) Clinical utility of multi-energy spectral photon-­ Zheng X, Liu Y, Li M, Wang Q, Song B (2016) Dual-­
counting computed tomography in crystal arthritis. energy computed tomography for characterizing
Arthritis Rheumatol 71(7):1158–1162. https://doi. urinary calcified calculi and uric acid calculi: a meta-­
org/10.1002/art.40848 analysis. Eur J Radiol 85(10):1843–1848. https://doi.
Symons R, Krauss B, Sahbaee P, Cork TE, Lakshmanan org/10.1016/j.ejrad.2016.08.013
MN, Bluemke DA, Pourmorteza A (2017a) Photon-­ Zhou W, Bartlett DJ, Diehn FE, Glazebrook KN, Kotsenas
counting CT for simultaneous imaging of multiple con- AL, Carter RE, Fletcher JG, McCollough CH, Leng
trast agents in the abdomen: an in vivo study. Med Phys S (2019) Reduction of metal artifacts and improve-
44(10):5120–5127. https://doi.org/10.1002/mp.12301 ment in dose efficiency using photon-counting detec-
Symons R, Pourmorteza A, Sandfort V, Ahlman MA, tor computed tomography and tin filtration. Investig
Cropper T, Mallek M, Kappler S, Ulzheimer S, Mahesh Radiol 54(4):204–211. https://doi.org/10.1097/
M, Jones EC, Malayeri AA, Folio LR, Bluemke DA rli.0000000000000535
(2017b) Feasibility of dose-reduced chest CT with Zhou W, Lane JI, Carlson ML, Bruesewitz MR, Witte RJ,
photon-counting detectors: initial results in humans. Koeller KK, Eckel LJ, Carter RE, McCollough CH,
Radiology 285(3):980–989. https://doi.org/10.1148/ Leng S (2018) Comparison of a photon-counting-­
radiol.2017162587 detector CT with an energy-integrating-detector CT
Symons R, Reich DS, Bagheri M, Cork TE, Krauss B, for temporal bone imaging: a cadaveric study. AJNR
Ulzheimer S, Kappler S, Bluemke DA, Pourmorteza Am J Neuroradiol 39(9):1733–1738. https://doi.
A (2018) Photon-counting computed tomography for org/10.3174/ajnr.A5768