Image Fusion

in Preclinical
Applications

Claudia Kuntner-Hannes
York Haemisch
Editors

123
Image Fusion in Preclinical Applications
Claudia Kuntner-Hannes • York Haemisch
Editors

Image Fusion in Preclinical

Applications
Editors
Claudia Kuntner-Hannes York Haemisch
Center for Health & Bioresources Technical Sales & Business Development
AIT Austrian Institute of Technology Direct Conversion GmbH
GmbH München
Seibersdorf Germany
Austria

ISBN 978-3-030-02972-2    ISBN 978-3-030-02973-9 (eBook)

https://doi.org/10.1007/978-3-030-02973-9

Library of Congress Control Number: 2018964720

© Springer Nature Switzerland AG 2019

This work is subject to copyright. All rights are reserved 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, express 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
Preface

In the last decade, our understanding of the function of the human body and the
interaction of drugs with certain molecular targets has increased significantly. This
deepened insight is a result of the wider availability of new diagnostic tools such as
those developed by the different “-omics” disciplines (genomics, proteomics, etc.),
but largely also of the application of noninvasive, in vivo imaging techniques apply-
ing molecular targets and markers. In vivo imaging modalities have different char-
acteristics based on their individual mechanisms of tissue contrast or function,
specific sensitivity, and spatial and temporal resolution in relationship to diseases or
biological processes. Multimodality imaging using two or more imaging modalities
simultaneously or sequentially allows the combination of the strengths of individual
modalities while overcoming their limitations. Anatomical imaging techniques such
as (X-ray) computed tomography (CT) and functional imaging such as magnetic
resonance imaging (MRI) provide excellent structural detail and resolution.
Molecular imaging techniques such as positron emission tomography (PET) and
single-photon emission computed tomography (SPECT) allow tracing of biochemi-
cal and biological processes at the molecular level. By combining anatomical and/
or functional with molecular imaging, complementary information can be obtained
leading to an improved understanding of the physiological mechanisms at the
molecular and cellular level.
In vivo imaging modalities have emerged from their separate use to application
in combination, generating multimodal datasets. The combination of multimodal
datasets by image fusion has led to a new understanding of the biology underlying
multiple diseases and biological processes. Multimodal imaging techniques can be
based on the acquisition of images at different times, on the acquisition of images
using different techniques, or on simultaneously acquired images. Especially in pre-
clinical imaging, using animal (mostly rodent) models of human diseases, a variety
of imaging tools are available, ranging from anatomical imaging with CT and ana-
tomical and functional imaging with MRI to the molecular imaging techniques
based on the use of isotopes such as PET and SPECT. Especially for imaging small
rodents, these techniques are complemented by optical or acoustic imaging tech-
niques such as optical imaging, ultrasound, and optoacoustic imaging. Using this
enormous toolset of techniques allowing noninvasive insights into objects, imaging
has the potential to improve the identification and development of new diagnostic

v
vi Preface

and therapeutic drugs and to facilitate the translation of preclinical findings to appli-
cations in the clinics.
This book attempts to provide an overview of the different ingredients that are
important in preclinical multimodal imaging such as image fusion techniques, data
analysis, imaging systems hardware, and animal handling. Applications of various
combinations of imaging techniques are explained exemplarily on different case
studies in cardiovascular diseases, inflammatory diseases, radiotherapy, and drug
development applications. The reader might find it particularly interesting to learn
about the use of an imaging modality ranging from macroscopic (tomographic)
imaging of whole animals down to microscopic imaging of cells using
optoacoustics.
While this compilation of experiences of acknowledged authors in this field can
only point some spotlights onto interesting aspects of preclinical in vivo imaging,
we still hope that this book will enable the reader to obtain an understanding of the
different factors and aspects to be considered when planning and performing pre-
clinical imaging experiments. We also wish that the case studies provided might
serve as examples of how to apply multimodal imaging techniques adapted to dif-
ferent research questions and challenges in many application areas and might trig-
ger inspiration and ideas on when and how to apply multimodal preclinical imaging
in your own research studies.

Seibersdorf, Austria Claudia Kuntner-Hannes

München, Germany  York Haemisch
Contents

1 High-Level Story: Data Analysis in Multimodal Preclinical

Imaging—Methods and Tools ������������������������������������������������������������������   1
Gabriel Tobon, Jacob Hesterman, Shil Patel, and Christian Lackas
2 Instrumentation Challenges in (S)PE(C)T Systems�������������������������������� 25
David Brasse and Frederic Boisson
3 Influence of Animal Handling and Housing
on Multimodality Imaging������������������������������������������������������������������������ 55
David Stout
4 Multimodal Optoacoustic Imaging���������������������������������������������������������� 69
Murad Omar, Dominik Soliman, and Vasilis Ntziachristos
5 Small Animal Imaging in Oncology Drug Development������������������������ 101
Joseph D. Kalen and James L. Tatum
6 Investigation of Transporter-Mediated Drug-Drug
Interactions Using PET/MRI�������������������������������������������������������������������� 117
Thomas Wanek, Alexander Traxl, Claudia Kuntner-Hannes,
and Oliver Langer
7 Multimodality Preclinical Imaging in Inflammatory Diseases�������������� 135
Paul D. Acton
8 Preclinical Multimodality Imaging and Image
Fusion in Cardiovascular Disease������������������������������������������������������������ 161
James T. Thackeray
9 Dual-Energy SPECT Imaging with Contrast-Enhanced CT:
A Case Study���������������������������������������������������������������������������������������������� 183
Emily B. Martin, Alan Stuckey, Stephen J. Kennel,
and Jonathan S. Wall
10 Multimodality Imaging in Small Animal Radiotherapy������������������������ 197
Christian Vanhove and Stefaan Vandenberghe

vii
High-Level Story: Data Analysis
in Multimodal Preclinical 1
Imaging—Methods and Tools

Gabriel Tobon, Jacob Hesterman, Shil Patel,

and Christian Lackas

1.1 Introduction

Preclinical research has long been at the forefront of software and methodological
innovation for multimodal imaging. In vivo imaging, ex vivo imaging, and the com-
bination of the two offer a wide variety of complementary image modalities for
deriving biological insight. It is increasingly common for research applications to
use images from hybrid modality scanners, images from multiple scanners, and
derived images in tandem for advanced analysis and quantitation. A long-standing
example of the value of multimodal imaging comes from positron emission tomog-
raphy (PET) and single-photon emission computed tomography (SPECT) imaging,
where it is a common practice to acquire a computed tomography (CT) image for
attenuation correction (AC) of the reconstructed signal [1]. It demonstrates how
combining image modalities can not only provide complementary information but
also enhance the quality and reliability of one or more of the modalities. In the sub-
sequent analysis of PET and SPECT images, it is common for a researcher to quan-
tify the sum or concentration of signal within a region of interest. As “functional”
modalities, they may only exhibit contrast to background in regions where the tar-
geted function takes place; however, more regions than those visible may be desired
a priori for quantification. The quantitation of such regions greatly benefits from
fusing an anatomical modality with a functional one for use in region
segmentation.
To maximize the insight derived from multimodal applications, careful align-
ment of images and their corresponding fusion is required for visualization. It is

G. Tobon · J. Hesterman · C. Lackas (*)

Invicro—A Konica Minolta Company, Boston, MA, USA
e-mail: [email protected]
S. Patel
Eisai Inc., Andover, MA, USA

© Springer Nature Switzerland AG 2019 1

C. Kuntner-Hannes, Y. Haemisch (eds.), Image Fusion in Preclinical
Applications, https://doi.org/10.1007/978-3-030-02973-9_1
2 G. Tobon et al.

software’s function to assist the user in establishing the spatial correspondence

across images and present their combined information appropriately. In the simplest
case, hybrid scanner vendors provide hardware-level alignment parameters, a fixed
transform that can be used by the software to align images for fusion. Otherwise, if
modalities are acquired on separate scanners, an inherent balance exists between
images containing complementary information for added insight and shared infor-
mation to support adequate alignment. Sophisticated alignment methods using
mutual information-based similarity metrics support the co-registration with mini-
mal shared spatial patterns. In yet another level of difficulty, it may be necessary to
have pixel-wise alignment across subjects in a multi-subject study. Functional mag-
netic resonance imaging (fMRI) and atlas-based approaches often require pixel-
wise spatial correspondence across subjects to normalize spatial patterns to a
template image matrix. In other cases, precise pixel-wise alignment may not be
necessary, yet template co-registration can more clearly normalize structure when
comparing subjects visually. Indeed, significant efforts have been made to combine
multimodal images acquired across patients, species, and imaging sites to improve
understanding and translation in research studies [2]. A truly multimodal software
package should be able to load and fuse image data stored in varied formats, matrix
sizes, voxel information, units, and dynamic ranges and leverage a variety of co-
registration algorithms.
This chapter is divided into three sections. In the first section, popular software
options used for preclinical research, core functions provided by these software
packages for image fusion, and the application of these tools on recent preclinical
studies are reviewed. In the second section, methods and examples are provided
within which multimodal data are utilized for the extraction of quantitative data.
This section focuses especially on how image registration tools, such as those
available within the software platforms described in the first section, enable appli-
cation of anatomical data to improve estimations from functional data. The final
section provides examples of how multimodal data, often across a variety of spa-
tial or temporal scales, can provide a more comprehensive understanding of a
biological question. This section focuses particularly on how software tools are
used to help combine and visualize such data. Finally, samples of novel analysis
and visualization routines used to combine multimodal data derived from recent
research in machine learning and medical image processing are introduced and
looked at.

1.2 Software Options

Although many clinical viewing software packages can be repurposed for preclini-
cal imaging applications, preclinical image data introduce a varied set of formats
and unique challenges not commonly supported by clinical imaging software tools.
In preclinical research, the focus on diagnostic accuracy is supplanted with issues of
spatial resolution and noise reduction. In addition, image data formats vary in their
levels of compliance to the DICOM standard and can often be proprietary. Table 1.1
1 High-Level Story: Data Analysis in Multimodal Preclinical Imaging—Methods… 3

Table 1.1 A list of commonly used software in preclinical radiology research

Software Author Licensing Application area Platforms
VivoQuanta Invicro Commercial General Windows, Mac OS, GNU/
[3] Linux
ImageJ/Fiji Community Free (BSD/ General Windows, Mac OS, GNU/
[4] GPL) Linux
SPM [5] UCL Free (GPL) MRI, PET/ Windows, Mac OS, GNU/
SPECT, EEG/ Linux on MATLAB
MEG
FSL [6] Oxford Free (FSL) MRI Mac OS, GNU/Linux
Amide [7] Stanford Free (GPL) General Windows, Mac OS, GNU/
Linux
Analyze [8] AnalyzeDirect Commercial General Windows, Mac OS, GNU/
Linux
AFNI/SUMA NIH Free (GPL) MRI Windows (Cygwin), Mac OS,
[9] GNU/Linux
Paraview Kitwareb Free (BSD) General Windows, Mac OS, GNU/
[10] Linux
ITK SNAP UPenn/Utah/ Free (GPL) General Windows, Mac OS, GNU/
[11] Kitware Linux
PMOD [12] PMOD Commercial PET/SPECT, Windows, Mac OS, GNU/
Technologies CT, MRI Linux
3D Slicer Community Free (Slicer) General Windows, Mac OS, GNU/
[13] Linux
FreeSurfer MGH Martinos Free MRI, PET Mac OS, GNU/Linux
[14] Center (Freesurfer)
Amira [15] FEI Commercial General Windows, Mac OS, GNU/
Linux
Osirix [16] Pixmeo Commercial General Mac OS
(Free Lite
version)
The authors listed are not comprehensive; however, they indicate the primary development con-
tributors to the application. Free licenses are marked by name and range from copyleft, to free for
academic use, and flexible BSD-style varieties. Application areas and supported platforms are
marked according to how the software product is advertised; however, it is common that different
software packages specialize or lack in methodologies not indicated in this table
a
VivoQuant was used in the generation of all figures used in this chapter
b
ParaView contributors also include Sandia National Laboratories, Los Alamos National
Laboratory, Army Research Laboratory, and CSimSoft

catalogs some of the most prevalent imaging software tools used by preclinical
researchers.
Each software specializes in certain analysis and visualization applications, and
users must consider data format compatibility between their imaging equipment and
the analysis software. Many camera vendors will provide a custom-developed soft-
ware not mentioned in Table 1.1 for the visualization and analysis of images from
their corresponding instrument; however, these tools often lack capabilities required
for multimodal multi-scanner analysis applications. Examples of such software
4 G. Tobon et al.

include ParaVision® from Bruker and Living Image® from PerkinElmer. Users also
have a choice between commercial and free applications. Free applications with
open-source licensing often receive novel contributions from the research commu-
nity. By exposing their source code, they provide users with the deepest level of
flexibility and customization, thus making them a strong choice for method devel-
opment researchers. By contrast, commercial applications tend to hide their algo-
rithmic complexity in an effort to make workflows more accessible to general users
and less susceptible to user errors, and their development teams are more likely to
invest in government regulatory compliance. In this chapter, all example images and
analyses are generated by VivoQuant®, an application specializing in multimodal
visualization and analysis for CT, PET, SPECT, MRI, ultrasound, autoradiography,
and fluorescence imaging with the ability to load a variety of open and proprietary
preclinical imaging formats.
Much of the software in Table 1.1 relies on core technology developed by con-
tributors to open-source and proprietary medical imaging libraries. These libraries
include, but are not limited to, the Insight Toolkit [17] for segmentation and co-
registration, Visualization Toolkit [18] and Open Inventor [19] for rendering, Qt
[20] for user interfaces and networking, DCMTK for DICOM support, and Java
[21] for cross-platform deployment. Indeed, many software developers in the imag-
ing community both collaborate and compete to advance the tools for imaging sci-
entists. To be effective for multimodal applications, tools should support a variety of
image formats, fuse images together with sophisticated co-registration and visual-
ization algorithms, and provide both generalized and domain-specific processing
tools.

1.3 Extraction of Data for Quantitation

The ability to extract estimates of physical or physiological quantities is a powerful

feature of imaging. Quantitative accuracy depends on many factors, such as scanner
calibration, noise, spatial and temporal resolutions, quality of segmentation, and
application of appropriate corrections and models. Image registration, both intra-
and inter-modality, plays a critical role in accurate quantitation. This importance is
often manifested through the use of image registration to enable the generation and/
or application of anatomical regions of interest (ROIs).

1.3.1 Preprocessing

Most image processing pipelines include some form of preprocessing or data har-
monization, followed by a ROI generation procedure and extraction of quantitative
data from the generated ROIs. In some cases, those extracted data undergo addi-
tional processing (e.g., tracer kinetic modeling). While several data processing steps
(e.g., intensity normalization, resampling, unit conversion) are unimodal in nature,
multimodal data are critical in many applications to improve the accuracy and
robustness of this data extraction process.
1 High-Level Story: Data Analysis in Multimodal Preclinical Imaging—Methods… 5

The preprocessing step for which multimodal data are most critical is that of
image alignment or co-registration. In some multimodal scanners acquiring images
simultaneously, different modalities are natively co-registered via its reconstruction
algorithm. Commonly, however, modalities must be acquired in separate scanning
sessions, perhaps on different scanners entirely, during which the patient/animal has
been moved or adjusted. In such cases, a post hoc co-registration is needed to align
images prior to further analysis. In the case of preclinical imaging, it is not uncom-
mon for researchers to use rodent “hotels” where multiple animals can be scanned
simultaneously in one session. These images must be cropped and aligned appropri-
ately across modalities to analyze animals individually. Figure 1.1 shows an exam-
ple where the PET image and a CT image of a mouse hotel were acquired on
separate scanners. The individual animals were segmented using the CT scan, and
those segmentations were applied to crop both the PET and CT images. This is a

b
a

Fig. 1.1 Axial views of a PET+CT mouse hotel acquisition of three mice with one scan. Before
analyzing each rodent separately, the PET and CT images are (a) co-registered using a rigid trans-
form and (b) segmented using an automatic laplacian-based algorithm
6 G. Tobon et al.

canonical approach for multimodal software—derive an image transformation from

one modality and apply it to one or more additional modalities.
A co-registration between two images is typically defined with a spatial transfor-
mation applied to one image, the “moving” image, to align it more closely to the
other, the “fixed” image. To visualize or analyze an aligned image, the moving
image must be resampled spatially using the co-registration transform.

1.3.1.1 Transforms

Linear
Linear spatial transformations can be subcategorized into rigid and affine trans-
forms based on their mathematical guarantees. Rigid transforms preserve absolute
distances between any two points of the image. A rigid transform parameterized by
rotation and translation is often the best choice for multiple scans of one animal
with stable structural anatomy, such as multiple scans of the brain. Thanks to the
equivalence of absolute distances, it is expected that volumes, angles, and pixel
quantitation will all be stable following the application of a rigid transform. Indeed,
in practice, these values are stable for many naturally occurring structures within
some sampling error inherent to the discrete nature of image data. More details on
the effects of sampling error can be found in Image Registration Pitfalls.
Affine transforms guarantee the preservation of relative distances within an
image and define a superset of rigid transforms. An affine transform may include
parameters for scaling or shearing an image and thus loses many of the aforemen-
tioned guarantees provided by rigid transforms.

Nonlinear
In many imaging research applications, a linear transform will not be sufficient to
provide adequate alignment. Group-level voxel-wise analysis and atlas-based vol-
ume estimation are commonly used cases for nonlinear spatial alignment, where
similar modalities are co-registered across different patients [22–24] . This is often
called “spatial normalization” because it normalizes scans from different animals
with respect to their anatomical differences. Certain MRI routines, such as echo-
planar imaging, introduce significant spatial distortion that can be partially recov-
ered from using multimodal nonlinear alignment [25] or even alignment of serial
acquisitions taken in opposite phase-encoding directions [26]. These artifacts can
appear more severely in preclinical MRI than in clinical applications [27], further
justifying the use of nonlinear spatial alignment. The potential gains possible from
proper utilization of nonlinear registration are illustrated in the example shown in
Fig. 1.2.
Various state-of-the-art nonlinear co-registration transforms and optimization
algorithms exist for accurate alignment [28]. These transforms are often high
dimensional and provide minimal guarantees for quantification. In most cases, the
transform itself is spatially regularized during optimization to both avoid erratic
transforms and preserve locally similar relative distances. Within a small enough
local window of the image, the transform may be approximated linearly where
1 High-Level Story: Data Analysis in Multimodal Preclinical Imaging—Methods… 7

Fig. 1.2 Example co-registration of T1 MRI maps of feline brains. From left to right, a check-
board overlay view of (1) two brains from different feline patients prior to co-registration, (2) the
brains aligned with a translation transform, (3) the brains rigidly aligned, (4) the brains aligned
with an affine transform, and (5) the brains aligned with a nonlinear deformation field computed
using the symmetric normalization algorithm

relative changes in size are stable. Accurate transform estimation often requires that
the images already be linearly aligned and that they contain dense anatomical con-
trast information. Indeed, nonlinear transforms are most accurately estimated from
modalities with significant intensity heterogeneity, such as MRI and white light
photography, although they have been successfully used with other modalities, such
as CT and ultrasound [29, 30].

1.3.1.2 Similarity Metrics

Co-registration algorithms are effective implementations of cost function optimiza-
tion. The cost function to be minimized is typically a negated similarity metric. For
example, the similarity between two images may be defined as the summed
Euclidean distance between pairs of landmarks on the image defined by the user.
Pinpointing consistent landmarks can be a challenging and manual task without
automatically identifiable multimodal fiducials, so often a more global approach is
used. An efficient landmark-based approach will often have consistent fiducials vis-
ible in both modalities and images to be co-registered as in Fig. 1.3.
For the following metrics, we first define two images—fixed image If and mov-
ing image Im—with intensity values If(x) or Im(x) where x is the n-dimensional spa-
tial location of a voxel. The normalized correlation ρ computes the spatial intensity
correlation between images:

C ov ( I m ,I f )
r=
Std ( I m ) ´ Std ( I f )

where Cov is the sample spatial covariance function and Std is the sample spatial
standard deviation function. This similarity measure shares the same advantages
and pitfalls it would with two one-dimensional functions. The metric works well
with similar image histograms, that is, similar modality and acquisition details.
Without prior histogram equalization, a pair of T1- and T2-weighted images will
not co-register well using normalized correlation.
8 G. Tobon et al.

Fig. 1.3 Fiducials suitable for landmark-based co-registration of ex vivo images. On the left are
3D reconstructions of white light images and fluorescence images, each with fiducials made
through the acquisition block and used for 3D alignment. On the right, similarly embedded fidu-
cials are matched across different slice photographs from a block. Co-registration of these images
is done using the Euclidean distance between fiducial markers

Another commonly used metric is mutual information. The motivation behind

using mutual information as a similarity metric is statistical: it measures how read-
ily the fixed image’s voxel intensities can be predicted from the moving image.
This metric’s computation begins by estimating the spatial probability density
function (PDF) of intensities for each image. These PDFs are used to compute the
statistical mutual information measure between the two images. Mutual informa-
tion is better suited for multimodal images and heterogeneous acquisition param-
eters for co-registration. It is the default metric for co-registrations done in
VivoQuant.

1.3.1.3 Image Registration Pitfalls

The above paragraphs have outlined the utility of registration transforms in intro-
ducing efficiency and consistency gains to quantitation. However, because images
exist on discrete grids, information content is not perfectly preserved upon applica-
tion of a transform.

MR-Based Susceptibility Artifact Correction

Functional imaging data often refers to the nuclear modalities of PET and
SPECT. However, several MRI modalities may also provide functional information.
These functional MRI modalities often require specialized sequences which sacri-
fice the resolution and contrast of anatomical data to capture the functional data of
interest. As an example, consider diffusion-weighted imaging (DWI). These scans
often suffer from susceptibility artifacts at tissue boundaries which manifest them-
selves in image distortion. This distortion is limited to the phase-encoding direction.
Using specific image registration steps, both linear and nonlinear, but specifically
restricted to the phase-encoding direction, enables correction of
1 High-Level Story: Data Analysis in Multimodal Preclinical Imaging—Methods… 9

susceptibility-induced distortion in the DWI image through the use of the anatomi-
cal information in a T1-weighted scan.

 pplication of Image Registration for Quantitative Extraction

A
We can define a region of interest in the moving image with volume Vm and total
intensity Im and thus a concentration:

Im
Cm =
Vm

Given a transform T(⋅), the software aligns the moving image into the fixed
image space to produce new values for concentration Cf, total intensity If, and vol-
ume Vf. For a rigid transform, Vf = Vm, If = Im, and Cf = Cm within resampling error.
An affine transform, however, may scale the region such that Vf ≠ Vm; given the
above relationship, it is clear to see that either Cf = Cm or If = Im in this case, but not
both. Expressed in other terms, either the mean intensity value or the total intensity
value will be stable in the transformed region. It is important to have clarity over
which has been preserved when an affine transform has been applied or avoid the
potential confusion altogether by using a rigid transform.
The linearity property of rigid and affine transforms does afford the user accu-
rate relative quantification. Vf = |T| × Vm; thus anatomical structures change with
a constant multiplicative factor across the entire image. This constant factor drops
out of the equation when normalizing a region of interest with a reference region.
This can be demonstrated by comparing the behavior of SUV and SUVR after
nonrigid affine transformation. A standardized uptake value (SUV) is equal to the
concentration within a volume (pixel or region) normalized by the injected dose
per body mass. SUVR is the ratio between SUV within a tracer’s target region and
the SUV within a chosen nontarget, or “reference,” region. A unit of voxel inten-
sity such as SUV would require qualification if the image underwent affine co-
registration but not rigid, while SUVR would be stable with any linear
transformation.
Post-reconstruction, care must be taken in the subsequent analysis software to
maintain the fidelity of quantitative data [31]. Imaging software users expect total
uptake in a region of interest to stay constant after resampling. Although this is not
generally true, many practical applications maintain equivalent region uptake and
concentration values within a small amount of unbiased error. The magnitude of this
error is controlled in many applications by the size and shape of the region itself.
For a spherical region with radius r, it is expected that errors in the uptake will occur
most at the surface of the sphere. The error is bound by the quantity of voxels at the
surface and is proportional to r2 compared to a total quantity of voxels proportional
to r3. The case for resampling is also aided by the fact that images are generally
spatially smooth; thus the mixing of voxels inside and outside the region would not
dramatically change local uptake values. Countermeasures can be taken to mini-
mize the effect of resampling on quantitative analysis, including the use of 3D vs.
2D regions of interest, larger regions, partial voxel quantitation, fixed voxel sizes
10 G. Tobon et al.

across data sets, and fixed volume regions of interest. Increasingly popular is the use
of partial volume correction [32], which should be done prior to any needed resam-
pling to avoid the introduction of errors into the procedure.

PET/SPECT Dopamine Example

Consider estimation of signal in the striatum of nuclear medicine (NM) images such
as PET or SPECT acquired using a dopamine tracer. A test image is co-registered to
an NM template, which may be either a single reference NM image or a composite
reference (see Atlases). The ROIs, striatal regions in this case, will have either been
generated on this NM template or manually co-registered to it. With each test data
set co-registered to the NM template, this common set of ROIs may be applied to
extract summary statistics. Variability is reduced through the use of these common
ROIs, and visualization is aided by the mapping of all data to a common space.
There are potential issues with this approach. For example, consider normal bio-
logical variability in structure. Nonlinear registration transforms are required to
capture this variability in mapping to the template space. However, the low resolu-
tion and lack of anatomical structure in many functional modalities prohibit suc-
cessful application of nonlinear registration methods. Additionally, consider the
situation of pathology, often of interest, has resulted in diminished signal in the
regions of interest. While stiffness of elastic solvers may be tuned to prevent over-
fitting, this step is not typically sufficient to allow optimal registration to a template
space using functional image data alone.
Therefore, multimodal registration techniques are critical. Structural imaging
modalities, particularly T1-weighted or T2-weighted MRI, provide superior resolu-
tion and tissue contrast to those of functional image data. Atlases are created and
aligned to MRI-based templates. Functional data are acquired in conjunction with
MRI data for a test subject and aligned via rigid registration, assuming the func-
tional and MRI data are acquired closely enough in time that structural changes will
not have occurred. The test MRI may then be nonlinearly registered to the MRI
template, and this transformation may be applied to the functional data to map it to
the template space. This approach has become the standard practice in many pre-
clinical and clinical applications.

1.3.2 Atlases

Atlases are population-level representations of anatomical structures and, as such,

are a powerful tool for generating quantitative output in specific regions of interest
(ROIs). This section describes some methods utilized for atlas construction and
application.
Atlas construction relies on development of a reference data set and generation
of labels. One approach to generating the reference data set is via combination of
multiple individual data sets drawn from a representative population of interest.
These data sets are combined through linear and nonlinear image registration to
produce a composite template image that is representative of the overall population.
1 High-Level Story: Data Analysis in Multimodal Preclinical Imaging—Methods… 11

An additional step of composing the template image in a well-defined coordinate

space is often applied to better enable cross-study comparison and dissemination
across multiple imaging centers. Anatomical labels are typically manually defined
by one or more expert anatomists on the template image.
Multimodal/functional atlases are utilized to apply ROIs to data sets in cases
where the ROIs are not necessarily well-represented by the signal present in the test
image data. In other words, it may be the case where signal characteristics prohibit
direct segmentation or identification of a particular ROI on a test data set so the ROI
is defined based on registration of the test data set and a pre-existing atlas. Such
atlases are often constructed with and applied to multimodal data.
Probabilistic atlases are generated similarly to anatomical atlases. However, atlas
labels are not deterministic in the sense that region boundaries are clearly defined
and region membership is binary (inside or outside). Rather, on construction, labels
from multiple individual data sets are combined to provide a probability value for a
given voxel to belong to a particular region. Probabilistic atlases have the advantage
of accounting for partial volume effects but can be more challenging to interpret and
require a separate atlas instance for each region.
Table 1.2 provides examples of anatomical, functional/multimodal, and probabi-
listic atlases for preclinical and clinical use. The list is not comprehensive but pro-
vides a sample of the types of atlas approaches utilized in preclinical research.

1.3.2.1 Multi-atlas Segmentation

Multi-atlas segmentation (MAS), also known as label fusion, methods utilize image
registration techniques to map a population of reference subjects and their associ-
ated subject-specific region segmentations to a test subject, producing a probability
map that may be used to segment regions within the test subject [50]. The MAS
approach requires some initial setup time for population of a reference library but is
powerful as a general segmentation approach across species, modality, and disease
area. The segmentation method benefits from a multi-resolution approach. A cas-
cade of registration and bounding box operations are utilized to map reference
library data sets to an input test data set. For example, in a whole-body image, this
process is achieved through a whole-body affine registration, followed by axial sub-
region affine registration, and, finally, region-specific nonlinear registration. The
resulting affine and nonlinear transforms are combined and applied to the reference
library ROIs. Reference library data sets are sorted by registration performance. A
predefined number of registered reference library ROIs are combined to generate an
ROI probability map representing the probability that any given voxel belongs to the
test data set region of interest. A deterministic ROI may be generated from the ROI
probability map using a variety of methods such as thresholding or maximum likeli-
hood approaches [50].

1.3.2.2 Brain Region Segmentation Without Soft Tissue Contrast

The anatomical imaging modality accompanying many preclinical nuclear imaging
scanners is x-ray CT rather than MRI. CT images have many uses but tend to have
less soft tissue contrast than anatomical MRI data. This lack of soft tissue contrast
12 G. Tobon et al.

Table 1.2 A list of atlases available to preclinical researchers, including the species, modality,
and accessibility information
Construction (voxel/surface,
Atlas Species automatic/manual) Availability
Digimouse [33] Mouse Voxel model built from co-registered Free
CT, PET, and cryosection data
MOBY [34] Mouse Based on MRI and MRM data and Licensing fee
modeled with NURBS. Also has through Johns
mouse skeleton atlas, cardiac, and Hopkins
respiratory functions University
Shan T2 MR [35] Rat Voxel model email author:
shanbc@ihep.
ac.cn
Allen Brain Atlas Mouse, human Coronal and sagittal mouse sections Online tools
[36]
XCAT Phantom Human NURBS surfaces, designed for 4D Licensing fee
[37] through Johns
Hopkins
Paxinos and Rat brain Stereotaxic coordinates Book
Watson [38]
INIA19 [39] Rhesus Volume, from 100 T1 MRI scans from Online for
macaque brain 19 animals download,
Attribution
License
MNI Macaque Macaca Stereotaxic reference frame. Combines Online for
Atlas [40] fascicularis, aspects of NeuroMaps Atlas download
Macaca
mulatta
112RM-SL [41] Rhesus T1 and T2 anatomical templates Online for
macaque brain download
Visible Mouse Mouse (MR) Effective volume resolution down to Umlaut Software
Atlas [42] 100 × 100 × 100 μm (1 × 10−3 mm3)
Scalable Brain Various Fully web-based display engine for Online viewing
Atlas [43] brain atlases
Duke/Calabrese Rhesus Diffusion tensor MRI atlas based on Online for
MRI + DTI Rhesus macaque brain postmortem scans of ten rhesus download
Macaque Atlas [44] macaques. Detailed segmentation with
242 anatomical structures
MRI/DTI-Based Ferret brain Population-based MRI and DTI Freely available
Ferret Brain Atlas templates of adult ferret brain and
[45] tools for voxel-wise analysis; in vivo
and ex vivo; 48 regions
Valdés-Hernández Rat brain T2 MRI template set for morphometry, Freely available
MR Rat Brain tissue segmentation, and fMRI
Templates [46] localization in rats; includes white and
gray matter probabilistic
segmentations; fMRI
Waxholm Space Sprague MRI/DTI volumes in NIfTI ures Freely available
Atlas [47] Dawley rat
brain
1 High-Level Story: Data Analysis in Multimodal Preclinical Imaging—Methods… 13

Table 1.2 (continued)

Construction (voxel/surface,
Atlas Species automatic/manual) Availability
T1 MRI Sheep Sheep Population-averaged T1 MR Freely available
Atlas [48] template—gray matter, white matter,
and CSF
Digital Atlas of Mixed-breed 1 mm and 0.33 mm isotropic Free to use, but
Dog Brain [49] dog diffeomorphic templates of the canine cannot be
brain, as well as cortical surface distributed for
representations suitable for use in commercial gain
FreeSurfer. In vivo, ex vivo, and
surface templates. T1 MR

can be particularly noticeable in brain applications that require separation of gray

matter (GM), white matter (WM), and cerebrospinal fluid (CSF). In this situation,
the application of multimodal data and the previously described multi-atlas segmen-
tation method provide powerful tools for improving quantitation. By imaging repre-
sentative subjects with both CT and anatomical MRI within a short temporal
window (such that anatomy is constant within subject) and segmenting the GM,
WM, and CSF from the MRI data sets, a reference library may be constructed for
these regions. When a new PET/CT or SPECT/CT data set is acquired, the MAS
tool may be used with a reference library of CT data sets but MR-derived atlas
regions to build subject-specific segmentations of GM, WM, and CSF even in the
absence of soft tissue contrast.
Consider an example study in which GM, WM, and CSF PET signals were to be
evaluated longitudinally in several cohorts of mice acquired using a dedicated PET/
CT scanner. Prior to execution of the PET study, the MR and CT reference library
data were constructed. High-resolution T2-weighted MR images (Biospec
7 T/20 cm, Bruker, T2-weighted RARE, TR = 2600 ms, TE = 74 ms, 10 averages,
15 1-mm slices, FOV = 1.7 × 1.6 cm, matrix = 128 × 128), with co-registered CT
images (Inveon PET/CT, Siemens, 80 kVp, 500 uA, 215 ms per projection, 0.21 ×
0.21 × 0.21 mm voxel size using filtered back projection with Shepp filter at Nyquist
point), were acquired in 12 healthy wild-type mice. GM, WM, and CSF regions
were manually segmented. Example segmentations from three coronal slices, dis-
played as probability maps in the template space, are shown in Fig. 1.4a. Using a
leave-one-out approach, the GM/WM/CSF regions were segmented for each of the
individual MR and CT data sets using the MAS methodology. The left panel of
Fig. 1.4b shows a coronal MR slice for the individual mouse. The middle panel
shows the probability map for the WM region, generated using MR data (the gold
standard). The right panel shows the same probability map generated using CT data,
which is largely devoid of soft tissue contrast. As shown in Fig. 1.4c, while segmen-
tation performance was superior using MR data, Dice coefficients of nearly 0.8
were achieved using only CT data. This performance enabled successful application
of a CT + ROI reference library to study PET/CT data for segmentation of complex
 14

a b

c
CT: Dice Coefficient MR: Dice Coefficient MR: Precision & Sensitivity CT: Precision & Sensitivity
0.79 0.85 1 1
01 01 01 01
02 0.845 02 04 04
0.78 03 03 07 0.95 07
04 04 0.95 09 09
05 0.84 05 11 11
0.77 06 06 0.9
07 0.835 07
08 08 0.9
0.76 09 09 0.85
0.83
10 10
11 11 0.8
0.75 0.825 0.85

Precision
Precision

0.82 0.75
0.74

Dice Coefficient
0.8

Dice coefficient
0.815
0.73 0.7
0.81
0.75
0.72 0.805 0.65

0.71 0.8 0.7 0.6

0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.7 0.75 0.8 0.85 0.9 0.95 1 0.6 0.65 0.7 0.75 0.8 0.85 0.9 0.95 1
Threshold Threshold Sensitivity Sensitivity

Fig. 1.4 Segmentation of gray matter (GM), white matter (WM), and cerebrospinal fluid (CSF) regions from CT data is challenging due to the lack of soft
tissue contrast. A cohort of 12 mice was imaged using MRI and CT, followed by expert manual segmentation of GM, WM, and CSF structures from the MRI.
(a) Visualization of probability maps for GM (red), WM (green), and CSF (blue) regions from the cohort of 12 animals, following nonlinear co-registration to
a template space. (b) The original MR (left) is segmented using a multi-atlas segmentation (MAS) approach to generate probability maps of the WM using (b)
MRI data, and (c) CT data as the reference library. Even without the soft tissue contrast provided by MRI, a CT-only approach provides nearly as accurate
segmentation of the WM region. (c) Comparisons of registration performance between MRI- and CT-based reference libraries in terms of Dice coefficient,
G. Tobon et al.

sensitivity, and precision

1 High-Level Story: Data Analysis in Multimodal Preclinical Imaging—Methods… 15

brain structures in the absence of soft tissue contrast. Subsequently, this approach
now has the utility of more confidently interrogating noninvasive methods targeting
WM and/or CSF in addition to GM regions.

1.4  ombining Multimodal Data to Provide

C
a Comprehensive Picture

In addition to their utility for extracting accurate information from images, multi-
modal data are pivotal in their ability to inform study outcomes by spanning multi-
ple orders of magnitude both spatially and temporally. Software tools are critical in
this capacity for visualization and image processing.

1.4.1 Image Fusion Visualization

MRI is the gold standard for 3D segmentation; however, even CT can be particu-
larly helpful for segmentation of soft tissue organs, including the heart, lungs, and
kidneys. In preclinical research, this is particularly important when one treatment
group shows very low levels of PET or SPECT uptake in a region of interest. Such
an organism can be invisible on PET/SPECT but still identifiable in CT/MRI.
The fusion process is used to visualize an overlay between two or more images.
It is one way to qualitatively assess the co-registration; however, it is also a useful
technique for a variety of research applications. In tandem, multimodal images can
be used to simultaneously outline different features of an image. For example, a
PET image can be used to clearly identify a xenograft tumor, while its co-registered
CT image delineates other anatomical features, such as the lungs, heart, and
kidneys.
A number of implementation details play an important role in how two- and
three-dimensional intensity matrices are converted to red, green, and blue pixel val-
ues on a display. Most reconstructed image formats define a physical patient coor-
dinate system in which distances can be measured in millimeters. For most
simultaneous multimodal scanners, this coordinate system is shared across modali-
ties and clearly defined in the reconstructed image headers, allowing the software to
co-register the images without any transform optimization. In the event that images
without aligned coordinate systems are loaded, the software should be able to rea-
sonably display them together, for example, by reasonably translating one image
coordinate system such that it is centered in the other image’s field of view. Even
when image coordinate systems are physically aligned, multimodal pairs of images
will rarely share the same number of slices, voxel sizes, or field of view. For fused
cross-sectional slice views, slices from each image are sampled along the same two-
dimensional plane in physical coordinates, and each resulting sample is fused across
images. In some preclinical MRI applications, the resolution along one axis will be
significantly different to the other two. When these thick slices are sampled, and
fused, any cross section along the low-resolution axis will appear much coarser.
16 G. Tobon et al.

Resampling an image with anisotropic voxels is necessary for visualization but

may be undesirable for analysis. Image Registration Pitfalls describes how resam-
pling may introduce quantitation errors. In addition, resampling an image to be
isotropic in memory may greatly increase the total voxel count and correspondingly
increase the runtime required for most analysis routines without increasing the qual-
ity or information content in the output. Where possible, anisotropic images should
be filtered with physical distances and neighborhoods or processed with 2D routines
on each high-resolution slice.
The typical color transfer function in medical imaging has a simple linear form.
That is, a defined array of colors is indexed linearly based on the voxel intensity. It
should be noted that equivalent differences along the color array indices correspond
approximately to equivalent differences in voxel intensity, within error due to color
discretization. In research applications, parametric derived images or autoradiogra-
phy images are examples where a logarithmic scale may be more appropriate for its
color transfer function. Individual image visualization and especially image fusion
are most informative when intensity contrasts of interest map to a linear difference
in color. If a linear color scale is used for logarithmically scaled intensity values, the
image will appear to have little to no contrast and may saturate the image relative to
others when it is fused.
Once intensity values are resampled and mapped to colors, image fusion is exe-
cuted with a chosen blending algorithm. The most common blending technique is
alpha blending, a linear combination between each color channel parameterized by
their corresponding opacity values. The higher the alpha value, the larger that
image’s contribution will be to the final blended pixel color. When fusing more than
two images together, a typical implementation will fuse the first two, fuse the result
with the third, and so on. The alpha compositing process implies that the final color
can depend on the order in which the three colors are blended. DWI is a special case
where it is common for the three components of the voxel’s tensor vector to be
mapped linearly to red, green, and blue channels of a final fused image.

1.4.1.1 Volume Rendering

Cross-sectional views provide 2D slices of 3D image matrices and form the stan-
dard views for clinical assessments. In research and preclinical imaging, it is also
customary for software to provide 3D graphical projections to give visualizations a
3D appearance on the screen. In a traditional 3D graphics scene, objects are repre-
sented with surface-based objects with particular reflectance and lighting proper-
ties. Ray casting is used to project the appearance of these surfaces to a 2D
representation on the screen [51]. Ray casting is also used for volume rendering,
albeit with a slightly different approach. Instead of intersecting with surfaces on the
screen, rays are projected into the image volume, and a resulting projected pixel
color is a function of all voxels in the volume intersected by the ray. A maximum
intensity projection computes the maximum voxel intensity along the ray and maps
this value through the image’s color transfer function to produce the visible color.
This has the effect of rendering the “brightest” features within the volume. Similarly,
color mappings along the said ray can be composited from pixel to pixel during
traversal with a compositing function that can be optimized to highlight particular
1 High-Level Story: Data Analysis in Multimodal Preclinical Imaging—Methods… 17

Fig. 1.5 Fused volume

renderings of 3D
reconstructed fluorescence
images and white light
images using composite
volume rendering

tissue intensities. An example of the fusion between two composite renderings is

shown in Fig. 1.5.
Whereas cross-sectional views only allow traversal along the plane’s perpendic-
ular axis, a projection can be manipulated in all three dimensions by transforming
the 3D volume in the scene prior to ray casting. The ray casting algorithm itself is
highly parallelizable, allowing its implementation to be executed on modern graph-
ics hardware with interactive performance. Graphic processing units (GPU) typi-
cally have their own dedicated memory hardware, often with less capacity than
random access memory on the machine. This makes GPU-accelerated rendering of
particularly large image volumes intractable on most computers without prior
down-sampling done by the software. The Visualization Toolkit provides rendering
implementations both on the CPU and the GPU for maximal cross-platform
compatibility.
Fusion for volume rendering is typically done using alpha blending of the 2D
projections from multiple images. Algorithmically computed opacity values can
often enhance the appearance of surface boundaries within the volume when the
gradient is used.

1.5 Applications of Preclinical Image Fusion

One of the primary advantages of multimodal imaging is the generation of comple-

mentary imaging around a common biological question. The ability to combine data
from various sources of imaging information is an area of active research and can
depend strongly on the specific task or question. In this section, we describe
18 G. Tobon et al.

examples from three scenarios. First, multimodal data across scales can be com-
pared directly. Second, multimodal data across disparate scales or types can be com-
bined to tell a more complete biological story. Third, data from multiple modalities
can be incorporated into complex models, often using machine learning methods to
improve our ability to perform classification and estimation tasks.

1.5.1 Cryofluoresence Tomography (CFT)

For example, consider the case of fluorescence imaging. A fluorophore may be

administered intravenously or intrathecally and imaged longitudinally in vivo to
provide some low-resolution spatial distribution information as well as in vivo tem-
poral information about biodistribution. That same subject can then be sacrificed,
and organs can be prepared for subsequent whole-body or focused organ ex vivo
CFT to provide higher-resolution distribution information at a macroscopic level.
During CFT acquisition, individual slices may be transferred to tape or slide to
enable fluorescence microscopy. Depending on the compounds and studies of inter-
est, this approach can be extended to other modalities. For example, if a compound
can be labeled with both a PET tracer and a fluorophore, then high temporal resolu-
tion PET dynamic imaging can be used to support pharmacokinetic modeling appli-
cations through in vivo imaging prior to higher-resolution, single timepoint ex vivo
CFT imaging. Alternately, for even higher-resolution imaging of focused regions,
transfer of individual sections to the appropriate medium enables the use of conven-
tional immunohistochemistry (IHC) staining. Consider the intrathecal administra-
tion of a Cy7-conjugated compound. A subject’s brain is blocked and imaged via
CFT to produce a high-resolution brain-wide distribution of signal as shown in
Fig. 1.6. Individual sections are transferred to tape for IHC staining as shown in
Fig. 1.6. Using these approaches, a single subject may be used to generate temporal,
macroscopic, and microscopic data—such approaches maximize scarce resources
while minimizing variability.

a b c

Fig. 1.6 The ability to utilize a single animal for imaging across multiple scales is often beneficial
as it removes intersubject variability. (a) Whole-brain CFT is used to visualize and assess fluoro-
phore distribution at the whole-brain level. Transfer of individual sections during CFT processing
enables application of fluorescence microscopy in specific subsections of the brain as shown here
at (b) ~30× magnification, and (c) ~40× magnification
1 High-Level Story: Data Analysis in Multimodal Preclinical Imaging—Methods… 19

As another example, consider a surgery study in hounds for study of a pancarpal

arthrodesis model. In this study, two cohorts of five hounds received different con-
centrations of a bone morphogenetic protein (BMP) co-labeled to 124I and 125I. In
vivo 124I PET/CT imaging of the paw region was performed out to 2 weeks post-
surgery. 125I planar imaging was performed in vivo out to 5 weeks. Through multiple
image processing and calibration steps, the percent injected activity in multiple paw
regions could be combined across the entire 5-week in vivo imaging period to
enable more accurate estimates of residence times and other cohort comparisons.
After 5 weeks, ex vivo 3D quantitative autoradiography and accompanying white
light imaging was performed. This imaging approach enabled quantitative analysis
and qualitative visualization of BMP distribution at high resolution as shown in
Fig. 1.7. The combination of these far-reaching multimodal imaging techniques
yielded a richer data set and greater statistical power from ten canines when com-
pared to a previous non-imaging study that utilized fifty canines.
Utility can also be gained from direct quantitative comparison of multimodal
imaging data across different spatial scales. For example, in vivo SPECT and PET
imaging are commonly used to estimate drug concentration within the brain.

a b 110 Subject 7510: Joints

100
90
Normalized Pct Inj Dose

80
70
60 l124 Data
50
l124 Fit
l125 Data
40
30
20
10
0
0 100 200 300 400 500 600 700 800 900
Time (h)

c d e

Fig. 1.7 A combined in vivo/ex vivo imaging pipeline provides quantitative and qualitative infor-
mation to describe the behavior of a BMP in a canine pancarpal arthrodesis model. (a) In vivo
PET/CT (PET in purple/blue, CT in gray) and planar imaging are utilized to estimate 124I and 125I
signal, respectively, out to 5 weeks post-surgery. (b) Information from these two modalities is
combined into a single quantitative time-activity curve, enabling pharmacokinetic analysis
between cohorts. (c–e) 3D cryo-imaging quantitative autoradiography is combined with white
light imaging to provide high-resolution qualitative localization of 125I signal post-sacrifice
20 G. Tobon et al.

d SPECT Day 28, Correlation = 0.91

a 40
Anim 1 Anim 2 Anim 3 Anim 4 Anim 5

35

Area Sum aB Hotspot

30

25

20

15

10

0
0 0.005 0.01 0.015 0.02 0.025 0.03
Entire Corr Hemi Percent ID/g (%ID/g)

b c

Fig. 1.8 Pipeline to enable correlative analysis of quantitative SPECT and IHC data in the mouse
brain. (a) Upon acquisition of SPECT/CT data, the test CT data are co-registered to a template CT.
(b) A template MR exists in the space of the template CT. The Bregma coordinate system is uti-
lized to identify the MR slice corresponding to each IHC section (vertical red lines). (c) The cor-
responding SPECT slice (left) is extracted, and its mean signal in %ID/g is compared to the
corresponding IHC section’s (right) total beta-amyloid burden. (d) SPECT and IHC summary
statistics are compared across multiple animals and sections, providing a strong validation between
estimation methods

Analysis of these distributions through ex vivo staining can provide useful valida-
tion but is often limited to visual, qualitative comparisons. In [52], in vivo SPECT/
CT images were co-registered and aligned to a template space. In the template
space, individual slices at different stereotaxic coordinates were extracted, and a
total %ID/g value within a hemisphere was computed. This %ID/g value was
directly compared to the estimates of plaque load derived via IHC. Hemispheric
estimates of IHC Aβ load and SPECT %ID/g were linearly related with an R2 = 0.83
as shown in Fig. 1.8.

1.5.2 Integration of Multimodal Imaging with Machine Learning

Computational tools derived from machine learning research have had great success
in image segmentation, data classification, and multimodal processing applications
[53, 54]. Machine learning is increasingly featured in imaging journals and confer-
ences with applications ranging from diagnosis to segmentation and longitudinal
prediction [55–57]. Imaging itself offers a variety of modalities which can be
1 High-Level Story: Data Analysis in Multimodal Preclinical Imaging—Methods… 21

further integrated with data from genomics and other biological assays to create
sophisticated models of disease and biomarkers [58, 59]. Indeed, the goal of this
active area of research is to leverage multimodal data at a larger scale to derive new
insights moving forward.
Machine learning models can be trained in a supervised manner when ground
truth results are available or in an unsupervised manner for blind discovery of
underlying patterns in the data. Image segmentation is an example where supervised
models have been trained with great success. Large software companies have con-
tributed methods and software tools which they use at large scales internally for
image classification and processing [60, 61]. Challenges have arisen in the applica-
tion of these methods to medical images [57]. 2D convolutional neural networks do
not leverage spatial correlations in the third dimension of typical 3D matrix sizes.
Furthermore, when 3D convolution kernels are used, the computational resource
demands exceed the capacity of individual workstations and do not scale efficiently
with increasing kernel sizes. Nevertheless, a number of applications have been dis-
covered where these models outperform classical methods [62]. The disadvantage
of some of these “deep learning” approaches can be that the highly parameterized
model does not take full advantage of relevant domain knowledge and may not shed
much light on the image generation process itself.

1.6 Conclusion

Preclinical multimodal data have tremendous utility. The use of the appropriate soft-
ware tools and resources, such as atlases, is critical to practical use of multimodal
images for the extraction of quantitative data and for the combination of data to
facilitate meaningful interpretation in the context of complex biological questions.
Available software tools are well-designed for handling of multimodal data across
multiple temporal and spatial scales. These tools have been specifically designed to
enable seamless visualization as well as to support an array of image processing
operations. Currently, the most important of these operations as it pertains to multi-
modal data is image registration with mature methods supporting a variety of tasks
and data types. As illustrated in the examples throughout this chapter, such tools are
already aiding in resource optimization, ethical use of animals through a reduction
in necessary animal number, and improved data utilization and outcome measures.
The burgeoning fields of machine learning and artificial intelligence are rapidly
growing in importance for multimodal analysis and will further increase the utility
of preclinical multimodal data as they transition from academia-heavy development
environments to more turn-key commercial availability. As imaging continues to
cement itself within the drug development cycle from discovery through late phase
clinical trials, the need for and uses of multimodal data will continue to expand. In
support of these needs, it is critical that innovation around software tools for image
fusion continues and performs robustly across different species and disease
models.
22 G. Tobon et al.

References
1. Kinahan PE, Townsend DW, Beyer T, Sashin D. Attenuation correction for a combined 3D
PET/CT scanner. Med Phys. 1998;25:2046–53.
2. Fox GB, Chin C-L, Luo F, et al. Translational neuroimaging of the CNS: novel pathways to
drug development. Mol Interv. 2009;9:302.
3. Invicro VivoQuant. http://www.vivoquant.com. Accessed 2 Nov 2017.
4. Schindelin J, Arganda-Carreras I, Frise E, et al. Fiji: an open-source platform for biological-
image analysis. Nat Methods. 2012;9:676–82. https://doi.org/10.1038/nmeth.2019.
5. SPM software—Statistical Parametric Mapping. http://www.fil.ion.ucl.ac.uk/spm/software/.
Accessed 2 Nov 2017.
6. Jenkinson M, Beckmann CF, Behrens TEJ, et al. FSL. NeuroImage. 2012;62:782–90. https://
doi.org/10.1016/j.neuroimage.2011.09.015.
7. Loening AM, Gambhir SS. AMIDE: a free software tool for multimodality medical image
analysis. Mol Imaging. 2003;2:131–7.
8. AnalyzeDirect.com. https://analyzedirect.com/. Accessed 2 Nov 2017.
9. Cox RW. AFNI: software for analysis and visualization of functional magnetic resonance neu-
roimages. Comput Biomed Res Int J. 1996;29:162–73.
10. Ayachit U. The ParaView guide: updated for ParaView version 4.3, Full color version. Los
Alamos: Kitware; 2015.
11. Yushkevich PA, Piven J, Cody Hazlett H, et al. User-guided 3D active contour segmenta-
tion of anatomical structures: significantly improved efficiency and reliability. NeuroImage.
2006;31:1116–28.
12. PMOD Technologies LLC—PMOD Technologies. https://www.pmod.com. Accessed 2 Nov
2017.
13. Kikinis R, Pieper SD, Vosburgh KG. 3D slicer: a platform for subject-specific image analysis,
visualization, and clinical support. In: Jolesz FA, editor. Intraoperative imaging and image-
guided therapy. New York, NY: Springer; 2014. p. 277–89.
14. FreeSurfer. http://surfer.nmr.mgh.harvard.edu/. Accessed 3 Nov 2017.
15. Amira 3D Software for Life Sciences|Thermo Fisher Scientific. https://www.fei.com/software/
amira-for-life-sciences/. Accessed 3 Nov 2017.
16. Rosset A, Spadola L, Ratib O. OsiriX: an open-source software for navigating in multidimen-
sional DICOM images. J Digit Imaging. 2004;17:205–16.
17. Avants BB, Tustison NJ, Stauffer M, et al. The insight ToolKit image registration framework.
Front Neuroinform. 2014;8:44. https://doi.org/10.3389/fninf.2014.00044.
18. VTK—The Visualization Toolkit. https://www.vtk.org/. Accessed 2 Nov 2017.
19. Open Inventor®|Open Inventor 3D SDK. http://www.openinventor.com/. Accessed 3 Nov
2017.
20. Qt|Cross-platform software development for embedded & desktop. https://www.qt.io/.
Accessed 3 Nov 2017.
21. Java Software|Oracle. https://www.oracle.com/java/index.html. Accessed 3 Nov 2017.
22. Dubois A, Hérard A-S, Delatour B, et al. Detection by voxel-wise statistical analysis of sig-
nificant changes in regional cerebral glucose uptake in an APP/PS1 transgenic mouse model of
Alzheimer’s disease. NeuroImage. 2010;51:586–98.
23. Lebenberg J, Hérard A-S, Dubois A, et al. Validation of MRI-based 3D digital atlas registration
with histological and autoradiographic volumes: an anatomofunctional transgenic mouse brain
imaging study. NeuroImage. 2010;51:1037–46.
24. Worsley KJ, Liao CH, Aston J, et al. A general statistical analysis for fMRI data. NeuroImage.
2002;15:1–15.
25. Murgasova MK, Lockwood-Estrin G, Nunes RG, et al. Distortion correction in fetal EPI using
non-rigid registration with a Laplacian constraint. IEEE Trans Med Imaging. 2018;37(1):12–9.
26. Hédouin R, Commowick O, Bannier E, et al. Block-matching distortion correction of
echo-planar images with opposite phase encoding directions. IEEE Trans Med Imaging.
2017;36:1106–15.
1 High-Level Story: Data Analysis in Multimodal Preclinical Imaging—Methods… 23

27. Jezzard P, Clare S. Sources of distortion in functional MRI data. Hum Brain Mapp. 1999;8:80–5.
28. Klein A, Andersson J, Ardekani BA, et al. Evaluation of 14 nonlinear deformation algorithms
applied to human brain MRI registration. NeuroImage. 2009;46:786–802.
29. Cheng J, Qiu W, Yuan J, et al. Accurate quantification of local changes for carotid arteries in
3D ultrasound images using convex optimization-based deformable registration. In: Medical
imaging: image processing. Proc SPIE. 2016;9784:978448.
30. Elfarnawany M, Alam SR, Agrawal SK, Ladak HM. Evaluation of non-rigid registration
parameters for atlas-based segmentation of CT Images of Human Cochlea. In: SPIE medical
imaging. International Society for Optics and Photonics. Proc SPIE. 2017;10133:101330Z.
31. Pierce LA, Elston BF, Clunie DA, et al. A digital reference object to analyze calculation accu-
racy of PET standardized uptake value. Radiology. 2015;277:538–45.
32. Erlandsson K, Buvat I, Pretorius PH, et al. A review of partial volume correction techniques
for emission tomography and their applications in neurology, cardiology and oncology. Phys
Med Biol. 2012;57:R119.
33. Dogdas B, Stout D, Chatziioannou AF, Leahy RM. Digimouse: a 3D whole body mouse
atlas from CT and cryosection data. Phys Med Biol. 2007;52:577–87. https://doi.
org/10.1088/0031-9155/52/3/003.
34. Khmelinskii A, Groen HC, Baiker M, et al. Segmentation and visual analysis of whole-body
mouse skeleton microSPECT. PLoS One. 2012;7:e48976.
35. Nie B, Chen K, Zhao S, et al. A rat brain MRI template with digital stereotaxic atlas of fine
anatomical delineations in paxinos space and its automated application in voxel‐wise analysis.
Hum Brain Mapp. 2013;34:1306–18.
36. Lein ES, Hawrylycz MJ, Ao N, et al. Genome-wide atlas of gene expression in the adult mouse
brain. Nature. 2007;445:168–76.
37. Segars WP, Sturgeon G, Mendonca S, et al. 4D XCAT phantom for multimodality imaging
research. Med Phys. 2010;37:4902–15.
38. Paxinos G, Franklin KB. The mouse brain in stereotaxic coordinates. Houston, TX: Gulf
Professional Publishing; 2004.
39. Rohlfing T, Kroenke CD, Sullivan EV, et al. The INIA19 template and NeuroMaps atlas for
primate brain image parcellation and spatial normalization. Front Neuroinform. 2012;6:27.
40. Frey S, Pandya DN, Chakravarty MM, et al. An MRI based average macaque monkey stereo-
taxic atlas and space (MNI monkey space). NeuroImage. 2011;55:1435–42.
41. McLaren DG, Kosmatka KJ, Oakes TR, et al. A population-average MRI-based atlas collec-
tion of the rhesus macaque. NeuroImage. 2009;45:52–9.
42. Petiet AE, Kaufman MH, Goddeeris MM, et al. High-resolution magnetic resonance histology
of the embryonic and neonatal mouse: a 4D atlas and morphologic database. Proc Natl Acad
Sci. 2008;105:12331–6.
43. Bakker R, Tiesinga P, Kötter R. The scalable brain atlas: instant web-based access to public
brain atlases and related content. Neuroinformatics. 2015;13:353–66.
44. Calabrese E, Badea A, Coe CL, et al. A diffusion tensor MRI atlas of the postmortem rhesus
macaque brain. NeuroImage. 2015;117:408–16.
45. Hutchinson EB, Schwerin SC, Radomski KL, et al. Population based MRI and DTI templates
of the adult ferret brain and tools for voxelwise analysis. NeuroImage. 2017;152:575–89.
46. Valdes Hernandez PA, Sumiyoshi A, Nonaka H, et al. An in vivo MRI template set for
morphometry, tissue segmentation, and fMRI localization in rats. Front Neuroinform.
2011;5:26.
47. Papp EA, Leergaard TB, Calabrese E, et al. Waxholm space atlas of the sprague dawley rat
brain. NeuroImage. 2014;97:374–86.
48. Nitzsche B, Frey S, Collins LD, et al. A stereotaxic, population-averaged T1w ovine brain atlas
including cerebral morphology and tissue volumes. Front Neuroanat. 2015;9:69.
49. Datta R, Lee J, Duda J, et al. A digital atlas of the dog brain. PLoS One. 2012;7:e52140.
50. Wang H, Suh JW, Das SR, et al. Multi-atlas segmentation with joint label fusion. IEEE Trans
Pattern Anal Mach Intell. 2013;35:611–23.
51. Glassner AS. An introduction to ray tracing. Palo Alto: Elsevier; 1989.
24 G. Tobon et al.

52. Patel S. SPECT imaging of an 125I-labeled antibody against amyloid plaques and determina-
tion of correlation with amyloid immunostaining. Poster presented at the 10th International
symposium on functional neuroreceptor mapping of the living brain, May 2014, Egmond aan
Zee, The Netherlands.
53. LeCun Y, Bengio Y, Hinton G. Deep learning. Nature. 2015;521:436–44.
54. Libbrecht MW, Noble WS. Machine learning applications in genetics and genomics. Nat Rev
Genet. 2015;16:321–32.
55. Huynh T, Gao Y, Kang J, et al. Estimating CT image from MRI data using structured random
forest and auto-context model. IEEE Trans Med Imaging. 2016;35:174–83.
56. Parmar C, Grossmann P, Bussink J, et al. Machine learning methods for quantitative radiomic
biomarkers. Sci Rep. 2015;5:13087.
57. Chen J, Yang L, Zhang Y, et al. Combining fully convolutional and recurrent neural networks
for 3D biomedical image segmentation. In: Lee DD, Sugiyama M, Luxburg UV, et al., editors.
Advances in neural information processing systems 29. Red Hook, NY: Curran Associates,
Inc.; 2016. p. 3036–44.
58. Batmanghelich NK, Dalca A, Quon G, et al. Probabilistic modeling of imaging, genetics and
diagnosis. IEEE Trans Med Imaging. 2016;35:1765–79.
59. Yao X, Yan J, Kim S, et al. Genetic findings using ADNI multimodal quantitative phenotypes:
a 2014 update. Alzheimers Dement J Alzheimers Assoc. 2015;11:P426.
60. Shotton J, Sharp T, Kipman A, et al. Real-time human pose recognition in parts from single
depth images. Commun ACM. 2013;56:116–24.
61. Abadi M, Agarwal A, Barham P, et al (2016) Tensorflow: large-scale machine learning on
heterogeneous distributed systems. arXiv:1603.04467
62. Prasoon A, Petersen K, Igel C, et al. Deep feature learning for knee cartilage segmentation
using a triplanar convolutional neural network. In: Mori K, Sakuma I, Sato Y, Barillot C,
Navab N, editors. Medical image computing and computer-assisted intervention—MICCAI
2013: 16th International Conference, Nagoya, Japan, September 22-26, 2013, Proceedings,
Part II. Berlin: Springer; 2013. p. 246–53.
Instrumentation Challenges in (S)PE(C)T
Systems 2
David Brasse and Frederic Boisson

2.1 Introduction

To date, molecular imaging is globally recognized as a powerful tool to assess

in vivo anatomical and functional structures within a living subject [1, 2]. A
wide range of imaging modalities based on different physical principles is avail-
able. Used individually, the value of modalities such as X-ray computed tomog-
raphy (CT), single-photon emission computed tomography (SPECT), positron
emission tomography (PET), magnetic resonance imaging (MRI), and ultra-
sound (US) is well recognized, both in clinical and preclinical fields [3–12].
Although established for decades, these modalities are still subject to much
research, both on the development of new detection components and the design
of new probes to artificially improve the natural contrast induced by physical
processes.
With different performances in terms of spatial and temporal resolutions, sensi-
tivity, and specificity but also their ability to provide quantitative results, all of the
previously mentioned imaging modalities allow to carry out a great variety of stud-
ies. However, the need for and relevance of obtaining a maximum of information
relating to the same subject is undeniable. The last two decades have thus seen the
advent of a full-fledged research portfolio on multimodal in vivo imaging systems,
leading to specific instrument developments.
Despite the certain enthusiasm for multimodal imaging, it is important to high-
light the instrumental developments specific to each imaging modality, in particu-
lar those concerning emission tomography modalities. Since the introduction of
the Anger camera in the 1950s [13], instrumentation (hardware) has become one

D. Brasse (*) · F. Boisson

Université de Strasbourg, CNRS, IPHC, Strasbourg, France
e-mail: [email protected]

© Springer Nature Switzerland AG 2019 25

C. Kuntner-Hannes, Y. Haemisch (eds.), Image Fusion in Preclinical
Applications, https://doi.org/10.1007/978-3-030-02973-9_2
26 D. Brasse and F. Boisson

essential element for the design of imaging systems. The gamma photon detection
principle based on the use of scintillating crystals, photomultiplier tubes (PMTs),
and readout electronics has been the basis of almost all clinical and preclinical
PET and SPECT systems. At the moment, we witness a rapid technology change
in the area of photodetectors with more and more systems using semiconductor
photodetectors such as avalanche photodiodes (APDs) [14, 15] and silicon PMs
(SiPMs). The change has first been triggered by the attempt to combine PET and
MRI, since PMTs are not compatible with magnetic fields. A second motivation
for the transition to SiPM photodetectors in particular is their better timing behav-
ior, important for the development of time-of-flight (TOF) (clinical) PET systems.
For preclinical applications, their compactness and the ability to read out very
small pixels make them particularly attractive. As the use of SiPMs increases,
their costs are coming down to competitive levels. In the same way, semiconduc-
tor detectors based on CdTe or CdZnTe have emerged as the alternative of choice
for SPECT system developments, thanks to their high energy and intrinsic spatial
resolutions.
This chapter focuses on instrumentation developments for nuclear emission
tomography imaging, without claiming completeness. It presents recent develop-
ments of PET and SPECT imaging technology by focusing on their origin triggered
by scientific questions but also characteristics that are specific to each of these
modalities. Thus, common notions to PET and SPECT imaging, such as the posi-
tioning of the interaction at the level of the detection module or the key elements of
the acquisition chain, are discussed. Special attention will be paid to instrumenta-
tion developments aiming at an insertion of PET or SPECT systems into MRI
systems.

2.2 Instrumentation in PET

At the early stage of this century, clinically available medical imaging technologies
have been adapted for use in preclinical investigations. The tremendous effort in
biological research focusing on the observation of the living at the molecular scale
pushed forward the development of such dedicated positron emission tomography
(PET) systems.
Techniques and methods used in molecular imaging combined with drug devel-
opment gave rise to a wide range of biological processes observed and measured
quantitatively using PET.
A number of important issues have been addressed while designing dedicated
preclinical PET systems. The main challenges have been to optimize both detection
efficiency and spatial resolution to correctly extract biological information from a
single living animal model facilitating the bridge between basic science and clinical
research.
This section presents an overview of instrumentation developments in PET in
terms of detection module and readout strategies.
2 Instrumentation Challenges in (S)PE(C)T Systems 27

2.2.1 Basic Components of a Preclinical PET Module

PET imaging is based on the simultaneous detection of two opposite gamma rays
generated by the annihilation of a positron with an electron within the volume of
interest. The detected photons contribute to identify a coincidence event occurring
in a volume described by the line joining the two triggered detectors called a line of
response (LOR). Over the last 20 years, hardware and software methods have been
proposed to improve the definition of each LOR (spatial resolution) and to increase
both the number of LOR and the number of coincidences acquired inside each LOR
(detection efficiency). PET molecular imaging dedicated to small-animal studies
requires both high spatial resolution and high sensitivity [16].
In most recent imaging systems, the geometry of the detector module is based on
a block detector structure. The following picture (Fig. 2.1) is an illustration of a
block structure introduced by M Casey and R Nutt [17] where a matrix of 64 crys-
tals is read by 4 individual photomultiplier tubes (PMTs). The same concept has
been transferred to a detection module dedicated to preclinical PET systems where
individual PMTs have been replaced by position-sensitive PMTs (PS-PMTs) or
multi-anode PMTs (MA-PMTs). In this configuration, crystals are identified using
a flood histogram.
This histogram image is obtained experimentally by irradiating the surface of the
detector with 511 keV photons. The location of each detected event is then calcu-
lated and stored in a two-dimensional image as illustrated in Fig. 2.2 for a matrix of
27 × 26 crystals read by MA-PMT.
The role of the scintillating crystal is to convert the high energy photons imping-
ing the detector to a number of optical photons proportional to the deposited energy.
Scintillating crystals can be organic or inorganic compounds and are characterized

Fig. 2.1 ECAT EXACT

HR+ (Siemens) detector
block composed of 64
crystals and read by 4
PMTs
28 D. Brasse and F. Boisson

Fig. 2.2 Illustration of a

flood histogram obtained
on an IRIS PET module
composed of 27 × 26
LYSO crystals, 1.7 mm
pitch, read by a H8500
MA-PMT (Image courtesy
of Liji Cao, Inviscan)

by intrinsic properties such as their stopping power, their light yield, and their scin-
tillation decay. The main requirements for crystals used in preclinical PET modules
are (1) a high efficiency to convert the excitation energy to fluorescent radiation, (2)
a transparency to its fluorescent radiation to optimize the transmission of light, (3)
the emission of light in a spectral range detectable by photodetectors, (4) a short
decay time to achieve fast response, and (5) a good radiation hardness.
Scintillating crystals used for preclinical PET are inorganic, mainly cerium-­
doped and lutetium-based. One drawback using lutetium-based crystals is the fact
that 2.6% of naturally occurring lutetium is 176Lu, a long-lived radioactive isotope
including beta and gamma decays resulting in a small natural radiation background
increasing random events during PET acquisition. This radiation baseline can how-
ever be beneficial to, for example, energy calibration or daily quality control with-
out any external radioactive source [18].
Table 2.1 lists the principal properties of bismuth germinate (Bi4Ge3O12 or BGO),
cerium-doped gadolinium oxyorthosilicate (Gd2SiO5[Ce] or GSO), cerium-doped
lutetium oxyorthosilicate (Lu2SiO5[Ce] or LSO), cerium-doped lutetium orthoalu-
minate (LuAlO3[Ce] or LuAP), cerium-doped lutetium yttrium orthosilicate (Lu2(1-­
x)Y2xSiO5[Ce] or LYSO), and cerium-doped mixed lutetium and gadolinium
orthosilicate (Lu1.8Gd0.2SiO5[Ce] or LGSO) crystals. Additional information on
scintillators can be found in [19, 20].
Historically, PET scanners used photomultiplier tubes to convert the scintilla-
tion light into an electric current. The basic principle of a PMT is the multiplica-
tion of photoelectron with a series of dynodes in a vacuum environment. The
scintillation light is transmitted through an entrance window of the vacuum tube
and interacts with a thin layer of material called photocathode. Each optical
2 Instrumentation Challenges in (S)PE(C)T Systems 29

Table 2.1 Properties of main scintillating crystals used for preclinical PET imaging
BGO GSO LSO LuAP LYSO LGSO
Density (g cm−3) 7.13 6.70 7.40 8.34 7.11 7.3
Effective atomic number 75 59 66 65 65 63
μ at 511 keV (cm−1) 0.89 0.72 0.88 0.95 0.82 0.80
Photofraction (%) 37.2 24.9 32.4 30.4 31.1 34
Index of refraction 2.15 1.89 1.81 1.94 1.81 1.8
Light yield (ph MeV−1) 8200 9000 25,000 11,300 33,000 41,000
Scintillation decay (ns) 300 56 40 184 41 31–47
λmax (nm) 480 430 420 365 420 420

photon can then produce a photoelectron via photoelectric effect. The photoelec-
tron is then accelerated by a high potential difference and focalized on a first
dynode. The acquired energy during the acceleration process allows the next dyn-
ode, coated with an emissive material, to release three to four new electrons when
the photoelectron interacts with it. After several stages of dynode amplification,
each initial photoelectron gives birth to 106 electrons. A PMT can be assumed as
a current source where an optical photon can generate a charge of 160 pC in few
nanoseconds.
An alternative to PMTs is the avalanche photodiodes (APDs). This photodetector
is based on a silicon photodiode. The absorption of an incident optical photon gen-
erates electron-hole pairs. An applied reverse bias voltage on the junction creates a
strong internal electric field which accelerates the electron through the diode and
produces secondary electrons by impact ionization. The resulting electron ava-
lanche can produce up to several hundreds of electrons.
More recently, SiPMs have been implemented as photodetectors for PET [21].
Further capitalizing on the idea of semiconductor avalanche photodiodes, SiPMs
are matrices of miniaturized APDs (called “single-photon avalanche diodes”
(SPADs)) operating in Geiger mode, showing the same compactness and non-­
sensitivity to magnetic field. Due to the Geiger-mode operation, SiPMs provide
electronic gains up to 106, independent of the incident signal and thus the ability to
count single photons. The number of SPADs forming a SiPM determines its dynamic
range, the size and density of SPADs, and the sensitivity and timing behavior of the
SiPM. The linearity of a SiPM response is strongly correlated with the number of
SPADs in a single pixel. Therefore, in a properly designed configuration, the num-
ber of SPADs corresponds to the expected number of optical photons produced by
the scintillation process and transferred to the photodetector. In order to assess the
energy, the remaining non-linearities have to be corrected [22, 23]. Optimizing both
the energy and timing resolution requires to deal with the trade-off between the size
and number of SPADs within a single SiPM pixel. It is interesting to observe the
evolution of this emerging technology, particularly from the perspective of the crys-
tal choice and their dimensions.
In this “analog” SiPM, the signals from individual SPADs are summed up, and
the energy and timing information of each event are mainly derived from leading
30 D. Brasse and F. Boisson

Table 2.2 Properties of main analog photodetectors used for preclinical PET imaging
PMT APD SiPM
Photodetection efficiency at 420 nm 20–40% 60–70% 25–75%
Gain 106–107 101–103 105–106
Rise time (ns) ∼1 ∼5 ∼1
Bias (V) >800 300–1000 30–80
Temperature sensitivity (%/K) <1 ∼3 1–8
Compactness − +++ +++
Magnetic field sensitivity − +++ +++

edge discrimination and signal integration. Digital approach has been introduced
where every photoelectron detected in a SPAD is recorded with its own timing
information. This digital process was first commercialized by Philips [24] where
they only introduced one time-to-digital converter for a large number of SPAD. The
number of triggered SPADs represents the energy of the event.
Table 2.2 summarizes the principle properties of the photodetectors used in pre-
clinical PET systems to convert the optical photon flux to an electric signal.
The use of inorganic scintillating crystal coupled to photodetector is an indirect
approach to convert 511 keV to electric signal. Another solution is to use a semicon-
ductor material such as cadmium zinc telluride (CZT) to directly convert the
absorbed energy of the gamma photon into a significant electronic signal. CZT is a
wide band gap compound semiconductor with a relatively high effective atomic
number. The electric signal resulting from the gamma interaction is read out by a
network of metallic electrodes. Several parameters can be optimized to maximize
the performance of such detector for PET imaging such as the width of the cathode
strip, the steering strip, and the bias voltage [25]. Based on these developments,
design of preclinical PET scanner has been proposed targeting high performance
[26–29].

2.2.2 Event Positioning

In the block detector approach, the size of the individual crystal element and the
light readout scheme determine the intrinsic spatial resolution [30]. The crystal seg-
mentation is driven by several parameters to improve the spatial resolution of the
detection module. It could be driven either by the photodetector segmentation in a
one-to-one coupling scheme or by the limits of crystal identification when a light
sharing approach is preferred.
A complete PET system consists in a large number of block detectors positioned
around the animal to be imaged. The most common geometrical configuration of a
preclinical PET system follows the ring geometry. The number of PET modules in
the transverse plane defined the polygonal shape of the scanner, while the number
of rings determines the axial extent of the imaging system.
2 Instrumentation Challenges in (S)PE(C)T Systems 31

The detection efficiency depends on the geometrical solid angle acceptance and
the density and thickness of the detector material. To improve detection efficiency,
thicker crystal elements can be used. However, the trade-off is a reduction in spatial
resolution through parallax errors due to the lack of an accurate measurement of the
depth of interaction (DOI).
To overcome this issue, a possible solution is to employ a so-called phoswich
detector with two or more crystal layers sampling the interaction depth and thus
obtaining an estimate of the DOI [31]. The DOI is discretized based on the number
of layers used. The number of layers can be limited by the proper identification of
each individual crystal. Several approaches have been investigated dealing with the
offset of the layers in the flood map or with the intrinsic properties of the crystal
composing each layer such as the scintillation decay or the concentration of the dop-
ant [32, 33].
As an example, Ito et al. demonstrated the capability of identifying up to four
layers [34].
For a continuous sampling of the DOI and a single-ended readout of the detector,
DOI information can be extracted from the pulse shape. A dedicated coating is
applied on the crystal surface resulting in a modification of the decay time of the
detected pulse in a depth-dependent manner [35]. Several groups have used scintil-
lator array coupled at both sides to avalanche photodiode (APD) arrays [36, 37], to
position-sensitive APDs [38–41], or using hybrid solution [42–45]. In this configu-
ration, the DOI resolution is mainly driven by the light shared on both photodetector
sides.
Already in 1978, Ter-Pogossian et al. developed an original concept where
thallium-­doped sodium iodide (NaI(Tl)) crystals were positioned in the axial direc-
tion with a PMT at each end of the crystals [46]. The main advantage of this geom-
etry is that neither resolution nor sensitivity is compromised due to a simplification
in the depth encoding procedure. Figure 2.3 illustrates this geometrical approach.

Fig. 2.3 Diagram of a Photodetector

detector module where
crystals are positioned
according to the axial Scintillation
direction with a Axial position
photodetector at both ends direction
of each crystal element

LOR

Pixel element

Photodetector
32 D. Brasse and F. Boisson

Since this pioneer paper, several groups have used this geometrical concept
where the DOI information gives the position in the axial direction [47–49]. The hit
crystal gives the radial coordinates, and the axial coordinate is computed based on
the light sharing on both photodetectors. The axial resolution mainly depends on the
geometrical and intrinsic properties of the crystal, its coating, and the timing resolu-
tion between both sides. To achieve a reasonable axial resolution, the axial extent of
the crystals needs to be adjusted, possibly limiting the axial field of view (FOV) of
the PET system [50]. To overcome this issue, one solution is to add several modules
in the axial direction. The challenge is then to reduce the dead space introduced by
the photodetector thickness.
Another approach has been investigated to retrieve the z-coordinate of such axi-
ally oriented crystals: The AX-PET collaboration [51] developed a system with axi-
ally oriented LYSO crystal bars with wavelength shifter (WLS) strips placed
orthogonally in between the layers of crystals. Photodetectors located at the end of
the crystal bars and at the WLS strips are used to measure the energy deposited in
both the bars and the WLS strips. The radial coordinates of each hit LYSO crystal
are given by its relative position, and the z-coordinate is given by the light spread on
the WLS strips. In general, these proposed geometries make it possible to obtain 3D
information of the photon interactions inside the detection modules with the DOI as
a discrete measurement and the axial position as a continuous one. Introducing TOF
information and digital SiPM improves the resolution in the axial direction as
described in [52].
To push forward toward a 3D continuous sampling of the event interaction, the
use of monolithic crystals read by multichannel photodetectors has been proposed
and experimentally used. This solution can improve the filling fraction with a less
complex implementation of the scintillation stage. In addition, both events’ posi-
tioning characteristics and DOI can be obtained through different methods, such as
statistic-based positioning or maximum likelihood algorithms [53–55].
Figure 2.4 illustrates the detection strategies, going from the identification of the
crystal where the interaction occurred to the localization of the interaction within a
monolithic crystal.
The last decade has thus seen improvements in both hardware (photodetector)
and software (optical information processing) dedicated to the 3D positioning of

a b c
z z z

y y y

x x x

Fig. 2.4 Schematic representation of the event positioning. (a) illustrates the identification of the
crystal where the interaction occurred (2D), (b) the addition of the DOI information to the crystal
identification (2.5D), and (c) the determination of the interaction position within a monolithic
crystal (3D)
2 Instrumentation Challenges in (S)PE(C)T Systems 33

events in a monolithic crystal. In 2002, the University of Washington developed the

cMICE detector [56]. An expectation-maximization (EM) algorithm, associated to
a neural network, was used to read the signals. Both simulated and experimental
data reported a 1.4 mm FWHM spatial resolution for a 4-mm-thick LSO crystal
read by a PS-PMT. A few years later, the same team assessed the DOI resolution in
an 8-mm-thick LSO crystal coupled to a H8500 MA-PMT (Hamamatsu Photonics)
[57]. Both simulated and experimental data reported a 1.3 mm FWHM in the center
and 1.7 mm FWHM at the edge. A parametric approach enabled to estimate the
position of the scintillation leading to a spatial resolution of 1.06 mm (center) and
1.27 mm (edge) with a DOI resolution of 3.24 mm [58]. They also published the
work on modeling the light response that improves the spatial resolution [59]. The
latest results released have shown 1.33 ± 0.31 mm FWHM with 3.5 ± 0.31 mm DOI
FWHM for an 8-mm-thick LYSO crystal with beveled edges [60].
At the same time, the University of Delft investigated the reading of a monolithic
crystal using avalanche photodiodes [61]. The first results showed that the crystal
surface properties have little influence on the measured spatial resolution. However,
reading the optical distribution in front of the crystal allows improving the spatial
resolution. In addition, the reading of the optical signal on both sides of the crystal
enables to increase its thickness without deteriorating the spatial resolution. In
2007, they showed that 1.7 and 2.0 mm FWHM spatial resolution may be obtained
using LSO crystals with thicknesses of 10 and 20 mm, respectively [62]. In 2009,
the same team obtained results with a monolithic LYSO crystal read on each side by
two APD matrices [53, 61], [63–65]. This double readout provided a spatial resolu-
tion of 1.05 mm in the center of the detector with an energy resolution of 11% and
a time resolution of 2.8 ns. To improve temporal resolution, SiPM matrices were
preferred [64], which led to a spatial resolution of 1.6 mm with an energy resolution
of 14% and a time resolution of 960 ps.
Two preliminary papers introduced the impact of digital SiPM matrices on
intrinsic detector performance. A coincidence timing resolution (CTR) of 157 ps
FWHM was obtained with a 10-mm-thick LSO crystal [66] by considering the time
of arrival of optical photons in the estimation algorithm. Equivalent timing results
have been obtained with a 24 mm x 24 mm x 10-mm-thick LSO crystal while keep-
ing a spatial resolution of 1 mm and an energy resolution of 12.8% [67].
The use of monolithic scintillator detectors (32 × 32 × 22 mm3) with dual-sided
digital SiPM readout is under investigation for high-resolution pediatric PET. The
detector presents a FWHM spatial resolution of 1.1 mm, a DOI resolution of
2.4 mm, an energy resolution of 10.2%, and a coincidence resolving time of
147 ps [68].
In 2007, Benlloch et al. published the results of a small-animal PET based on
LYSO continuous crystals [69]. The scintillating detector with a trapezoidal shape
and a thickness of 10 mm is coupled to a H8500 MA-PMT. The 64 PMT anodes are
associated to two resistive networks leading to five output signals to determine the
scintillation properties and one additional signal for the trigger. This experimental
setup led to a 2 mm spatial resolution at the center with an energy resolution of 19%
at 511 keV.
34 D. Brasse and F. Boisson

In 2010, the same university in collaboration with the University of Pisa, the
Bruno Kessler Foundation (FBK, Trento, Italy), and the Linear Accelerator
Laboratory (LAL, Orsay, France) published results obtained using SiPM matrix
[70]. A 12 mm × 12 mm × 5-mm-thick LYSO crystal, with a white and black coat-
ing, was coupled to 64 channels SiPM 1.5 × 1.5 mm2 read by MAROC2 ASIC-­
based readout board, resulting in a 0.9 mm FWHM spatial resolution and 16%
energy resolution. The Valencia team designed a small-animal PET based on eight
modules of MPPCs and 10-mm-thick monolithic LYSO crystals. Each MPPC array
is coupled to a resistive readout circuit providing signal outputs for each row and
column of the array. This configuration achieved an intrinsic detector resolution of
1.1 mm with an energy resolution of 14% [71].
In 2011, a Japanese team obtained 3 mm spatial resolution with a 15-mm-thick
LYSO crystal read by a H8500 MA-PMT [72].
After almost 15 years of research programs, the use of monolithic crystals
enables to reach spatial resolutions of about 1 mm comparable to or better than
traditional crystal matrix solutions. Commercial preclinical PET/CT systems
already offer both detection configurations.
Table 2.3 presents the data of five commercially available preclinical PET/CT
systems. Additional performance on other preclinical PET systems evaluated fol-
lowing the NEMA NU 4–2008 can be found in [9].

Table 2.3 Performance characteristics according to NEMA NU-4 protocol for five commercially
available preclinical PET/CT scanners
Company Bruker Mediso Inviscan Trifoil Sedecal
Scanners Albira NanoPET IRIS LabPET12 Argus
Axial FOV (mm) 148 94.8 95 114 48
Radial FOV (mm) 80 123 80 100 67
Scintillator
 Size (mm3) 50 × 50 × 10 1.12 × 1.6 × 1.6 × 2×2× 1.45 × 1.45 ×
1.12 × 13 12 (14 + 12) (8 + 7)
 
Type LYSO LYSO LYSO LGSO+LYSO GSO + LYSO
 
Shape Monolithic Matrix Matrix Matrix, dual Matrix, dual
layers layers
Photodetector MA-PMT MA-PMT MA-PMT APD MA-PMT
Sensitivity at center 6.3% 7.7% 8% 4.3% 4%
(358– (250–750 (250– (250– (250–700 keV)
664 keV) keV) 750 keV) 650 keV)
Resolution at 5 mm ML-EM 2D FBP ML-EM 2D-FBP 2D FBP
 Radial (mm) 1.5 1.7 1.0 1.5 1.6
 Tangential (mm) 1.5 1.5 1.0 1.6 1.6
 Axial (mm) 1.5 1.4 1.2 1.4 1.5
Data from [73] [74] [75] [76] [77]
2 Instrumentation Challenges in (S)PE(C)T Systems 35

2.2.3 Readout Strategies

In order to extend the intrinsic performance achieved at the level of one detection
module to the entire PET system, dedicated electronics is required. The quality of
signal processing for nuclear coincidence detection directly affects the ability of
PET scanners to extract relevant information such as energy, time stamp, and crystal
identification. The data acquisition (DAQ) system therefore is one of the key com-
ponents of such an instrument. The major function of the DAQ is to process and
digitize the analog signals provided by the photodetectors. Most DAQ systems are
built around analog subsystems to extract basic information from the detected
events [78–81]. Since the current is generated from a detector, two main parameters
can be extracted from the analog signal. The first one is a value directly correlated
to the energy deposited in the crystal. It is often the charge given by the peak value
of the integrated/shaped output signal. The second is the arrival time of the detected
photon. To date, readout electronics achieve coincidence time resolutions of 120 to
335 ps [82–86]. A discriminator together with time measurement circuits such as
time-to-digital converters is commonly used. The analog approach is often complex
and developed for a specific and optimized detector solution but requires low power
consumption [78].
With the development of Very-Large-Scale Integration (VLSI) and computer sci-
ence, front-end electronics enter the era of “go digital as soon as possible.” One
option to fulfill this requirement is digitizing in real time the entire signal using free-
running ADCs to retrieve the required timing and energy information [87]. This
method requires sampling rates in the range of GHz and then could become expen-
sive, especially if a large number of acquisition channels are needed. Another
approach can be to digitize the signal already at the photodetector level using, for
example, digital SiPM.
With the quest to improve as much as possible the spatial resolution, the number
of electrical available signals and the complexity of the DAQ increase. One way to
reduce the number of readout channels of a PET system is the use of a resistive
network directly attached to the pixelated photodetector. In 1996, Siegel et al.
described three different position encoding readout circuits to reduce the number of
processed signals from 64 to 4 [88]. The first one was based on the scintillation
camera readout approach developed by Hal Anger [13], the second was a discretized
single-wire, position-sensitive proportional counter readout (DPC), and the third
one was a hybrid circuit of the Anger and the DPC. These 2D readout schemes
allowed reducing the number of channels being read out while preserving the detec-
tor identification accuracy. In 2003, Popov et al. proposed a novel resistive network
approach called symmetric charge division (SCD) circuit [89]. This scheme has the
particularity that each network input is connected by a resistance to a line and by
another resistance to a column. In this case, the network makes it possible to reduce
the number of channels from n2 to 2n readout channels. Using this resistive network
approach combined to a dedicated reconstruction algorithm, the spatial distribution
of the charges on the photodetector surface can be reconstructed. The reconstructed
36 D. Brasse and F. Boisson

a b c

Fig. 2.5 3D rendering of three readout strategies. Illustrations of the charge distribution without
light sharing (a), with light sharing (b), and the corresponding charge projections using a SCD
readout approach (c)

charge distribution and the network properties provide new opportunities in terms
of photodetector gain correction and depth of interaction estimation when using
monolithic crystal [90]. Figure 2.5 illustrates the different strategies to extract the
event positioning when using a crystal matrix.

2.2.4 Focus on MRI-Driven Developments

While the above section presents the recent technological advances in improving
3D location of the photon interaction and its timing, the following section shall
highlight the instrumentation developments aiming at combining dedicated PET
with MRI. The challenge to integrate the PET component inside a MRI system is
responsible for almost all current research in instrumentation dedicated to preclini-
cal PET imaging.
The authors would like to point readers interested in PET/MRI technology to the
review recently published by Vandenberghe and Marsden, presenting a comprehen-
sive review of this dual-modality imaging technique [91]. In this review, the authors
report on all the technical challenges that have to be solved when designing (S)
PE(C)T/MRI systems using SiPMs, such as temperature dependence, noise, power
consumption, or cooling systems.
In 1997, a collaboration between the Crump Institute, UCLA, and Guy’s and St
Thomas’ Clinical PET Centre in London obtained the first simultaneous PET/MRI
experimental results [92, 93]. The first prototypes were based on 2 × 2 × 5(10) mm3
LSO crystals coupled to PMTs via optical fibers. The performances measured in a
0.2 T magnetic field were 2 mm spatial resolution, 41–45% energy resolution, and
a CTR of 20 ns FWHM. In 1998, Pichler et al. proposed another approach using
LSO crystals coupled to APDs to test their performance in a tomograph prototype.
3.7 × 3.7 × 12 mm3 crystals were one-to-one coupled to a two-dimensional APD
array. They obtained an energy resolution of 14.7%, an intrinsic spatial resolution of
2.2 mm, and a CTR of 3.2 ns FWHM [94]. Few years later, the collaboration con-
tinued the work using the same detection module and presented a spatial resolution
2 Instrumentation Challenges in (S)PE(C)T Systems 37

of 2.3 mm within the whole reconstructed field of view, with a system sensitivity of
350 cps/MBq [95].
In 2006, Grazioso et al. presented an APD-based PET detector for simultaneous
PET/MRI acquisition [96]. Each detector block consisted in array of APDs coupled
to 2 × 2 × 20 mm3 LSO crystals. Results showed an average crystal time resolution
of 1.8 ns, while the average crystal energy resolution was 17%. Although no spatial
resolution measurements were performed, PET and MR images of a micro-Derenzo
phantom were acquired simultaneously. The 2 mm rods were the smallest holes
clearly visible in the PET image.
For this period, all these research works highlighted the benefits of using APDs
to design and build PET inserts compatible with MRI. The APDs’ compactness as
semiconductor detectors made them very suitable photodetectors and attracted the
interest of the scientific community [14, 97–102].
With the introduction of SiPM, the use and interest for APD photodetector were
reduced. In 2011, Kwon et al. developed a small-animal PET prototype using SiPMs
[103]. The detection modules consisted of 4 × 13 arrays of 1.5 × 1.5 × 7 mm3 LGSO
crystals and 2 × 6 arrays of SiPMs (MPSS by Photonics SA). Performances obtained
were 1.0 to 1.4 mm spatial resolution depending on the radial offset, 0.085% sensi-
tivity for a 250–750 keV energy window, and an energy resolution of 25.8%. In
2012, Yoon et al. created a second version of the PET system developed by Kwon
et al. using a similar setup with 8 × 8 SiPM arrays (Hamamatsu Photonics) coupled
to 20 × 18 LGSO of 1.5 × 1.5 × 7 mm3 [104]. This setup improved the sensitivity up
to 0.195% for a 250–750 keV energy window. They obtained radial, tangential, and
axial resolutions of 1.0, 1.2, and 1.5 mm, respectively.
In 2013, Ko et al. reported their developments of SiPM-based preclinical PET
system to be inserted inside a small bore diameter and high magnetic field (>7 T)
[105]. The detection modules consisted of 9 × 9 arrays of 1.2 × 1.2 × 10 mm3 LYSO
crystals and monolithic 4 × 4 arrays of SiPMs (Hamamatsu Photonics). Intrinsic
performances obtained for the module were 0.88 to 1.09 mm spatial resolution and
an energy resolution of 14.2%. At the system level, the absolute peak sensitivity was
measured 3.4% for an energy window of 250–750 keV, an axial FOV of 55 mm, and
a ring diameter of 64 mm [106, 107].
In 2011, Yamamoto et al. developed a SiPM-based PET system for small animal
using 16 detector blocks consisting of 4 × 4 SiPM arrays (Hamamatsu Photonics)
[108–110]. Two types of LGSO scintillators with different percentages of cerium
doping and different lengths were used to form a DOI-encoding detector. The crys-
tal sizes were 1.1 × 1.2 × 5 mm3 and 1.1 × 1.2 × 6 mm3, respectively. The measured
system performances were 1.6 mm spatial resolution and 0.6% sensitivity with an
energy resolution ranging from 14 to 55%. In 2014, Weissler et al. proposed a PET
ring that consists of ten RF shielded detection modules [111]. Each module contains
1.3 × 1.3 × 10 mm3 LYSO crystals. A 1.5-mm-thick glass plate was used to spread
the light over 16 monolithic mounted and bonded SiPM arrays (FBK, Trento, Italy),
resulting in a fill factor close to 92%. Performances reported were 29.7% energy
resolution, 2.5 ns timing resolution, and a volumetric spatial resolution lower than
1.8 mm3 in the center of the FOV.
38 D. Brasse and F. Boisson

The challenge in developing PET/MRI systems is to ensure that the overall per-
formance is not hampered compared to stand-alone PET or MRI modalities.
Eventually the PET components should be completely inserted into the open bore of
the MRI gantry, between the gradient system and the MRI coil. In that configura-
tion, restrictions regarding the PET detector size and the associated DAQ must be
considered.
With the introduction of the digital version of the SiPM and the entire digitiza-
tion process taken place inside the MR bore, the questions of digital electromag-
netic noise patterns in the MR image were highlighted, and proper PET design
shielding has been investigated [112]. The Hyperion-IID PET system was inserted in
a 3 T clinical MRI system. The authors demonstrated a clear influence on the MRI
environment such as a distortion of the B0 field. Concerning the PET performance,
the values of energy and timing resolutions are increased up to 14% [113]. A full
description of this preclinical PET system and the associated performance can be
found in [114].
To achieve a high level of integration, one solution is to integrate the MRI coil
into the PET detector system [115]. In this published approach, the MRI coil is
positioned between the scintillator and the photosensor. In that case, the material
used must have optimal protection for the electromagnetic interferences, should
limit the noise introduced by the PET component, and have to be transparent to the
scintillation light.
The use of PET systems side-by-side with standard MRI systems already
allows for a dual-modal PET/MRI approach to be performed sequentially.
Moreover, several systems of PET inserts derived from academic developments
(Table 2.4) and dedicated simultaneous PET/MRI systems are already available
on the market.

Table 2.4 Summary of preclinical PET detector developments compatible with MRI
PET module Performance
Crystal Spatial
Section Thickness resolution Energy
Consortium (mm) (mm) TypePhotodetector (mm) resolution (%)
Shao et al. (1997) 2 5 LSO PMT + optical 2 41–45
fibers
Pichler et al. (1998) 3.7 12 LSO APD 2.2 14.7
Ziegler et al. (2001) 3.7 12 LSO APD 2.3 22
Grazioso et al. (2006) 2 20 LSO APD ~2.0 17
Yamamoto et al. 1.1 5+6 LGSO SiPM 1.6 27
(2010)
Kwon et al. (2011) 1.5 7 LGSO SiPM 1–1.4 25.8
Yoon et al. (2012) 1.5 7 LGSO SiPM 1–1.5 13.9
Weissler et al. (2014) 1.3 10 LYSO SiPM 1.8 29.7
Ko et al. (2016) 1.2 10 LYSO SiPM 0.7–1.3 14.2
Schug et al. (2016) 0.93 12 LYSO Digital SiPM 0.9 12.7
2 Instrumentation Challenges in (S)PE(C)T Systems 39

2.2.5 Performance Evaluation for Preclinical PET

The outline of the possible applications for small-animal in vivo imaging is mainly
driven by the performance of preclinical PET systems.
In the previous sections, we have summarized the variety of design of PET
modules to address high intrinsic spatial resolution and high detection effi-
ciency. The overall image quality also depends on the scanner geometry, the
animal to be imaged, and the software suite used to reconstruct and correct the
raw data.
In order to fairly compare the different existing preclinical PET systems, the
National Electrical Manufacturers Association (NEMA) set out a standardized
methodology for evaluating the performance of preclinical PET scanners (NEMA
NU-4). The challenge for such standardized methodology is to be applicable to a
large range of scanner designs with the objective to establish a baseline of system
performance in typical imaging condition.
Some conditions are mandatory to use the proposed standard to evaluate the
PET component: to have access to transverse sinograms and axial slices recon-
structed with filtered back projection algorithm and to have a scanner transverse
field of view higher than 33.5 mm in diameter to acquire data from an image qual-
ity phantom. The resulting standardized measurements can be used, for example,
by manufacturers to promote their equipment and by potential customers to com-
pare the different PET systems or to use them as gold standards for acceptance
testing.
Spatial resolution, scatter fraction, noise equivalent count rate, random coinci-
dence measurements, sensitivity, image quality, and accuracy of corrections are the
main figure of merits used to evaluate the preclinical PET imaging systems.
Table 2.5 summarizes the range of performance obtained for selected commer-
cial preclinical PET systems using the NEMA NU-4 standards. The purpose of this
table is not to list, in an exhaustive manner, all the performance of the entire set of
preclinical PET systems but to give to the reader a range of performance achievable
with existing commercial systems.

2.3 Instrumentation in SPECT

Like PET imaging, single-photon emission computed tomography (SPECT) is

based on the tracer principle. Although using a similar detector technology to that
of PET, the use of radioactive isotopes emitting simple photons requires the use of
a collimator to determine the origin of gamma photons. Thus, developments relating
to SPECT systems are essentially based on the collimation stage that is most likely
to fulfill the specifications of the instrument. The following sections present the
recent instrumental developments dedicated to preclinical SPECT imaging, with an
emphasis on SPECT systems dedicated to combined SPECT/MRI. We will also
discuss the main collimation profiles and the geometric parameters responsible for
40 D. Brasse and F. Boisson

Table 2.5 Non-exhaustive range of performance obtained for selected commercial preclinical
PET systems based on the NEMA NU-4 evaluation procedure
System characteristic
Ring diameter 110–260 mm
Axial FOV 45–130 mm
Crystal section 1.1–2.2 mm
Crystal thickness 10–25 mm
System performance
Spatial resolution (FWHM) At 5 mm At 25 mm
 Transverse 1–2.3 mm 1.4–2.9 mm
 Axial 1–2.3 mm 1.4–3.3 mm
Peak detection efficiency 1.2–8%
 Scatter fraction
 Mouse 5–30%
 Rat 12–35%
Image quality
 Uniformity 4.5–15.5%
 Recovery coefficient for 1 mm 3–27%
 Recovery coefficient for 5 mm 75–100%

the performance of the system, before presenting alternate collimation approaches,

which open new perspectives.

2.3.1 Overview of Instrumentation Advances

In 2006, Ronald J. Jaszczak published a wonderful paper describing the origins of

SPECT imaging, which took place about 60 years ago [116]. This historical review
mentions the beginnings of this imaging modality through stories from pioneers
who shaped the SPECT instrumentation. Although based on concepts common to
those offered by these pioneers, SPECT instrumentation remains a research field of
interest. Many SPECT systems, especially preclinical systems, have emerged over
the last 20 years. Technological advances have enabled and constantly help to
improve the overall performance of these systems. Table 2.6 presents the basic per-
formance of noncommercial preclinical SPECT scanners developed over the past
decade [117–123].
These academic developments aimed to explore all the possibilities offered
by the different modes of detection and collimation profiles. In parallel with
these developments, manufacturers have also proposed preclinical systems. It is
interesting to note that these commercial systems are all based on a crystal-PMT
approach associated with a (multi-) pinhole collimation. The use of large detec-
tion FOVs and strong magnification factors allow them in particular to increase
the detection efficiency with very high spatial resolutions. Table 2.7 presents
some of the latest commercial SPECT/CT systems for preclinical
investigations.
2 Instrumentation Challenges in (S)PE(C)T Systems 41

Table 2.6 Performance characteristics of selected noncommercial preclinical SPECT scanners

System Spatial resolution (best) Sensitivity (best) Detector Collimation
HiRe-SPECT 1.35 mm 0.0035% CsI(Na) + PMTs Slit
Fast-SPECT II 0.65 mm 0.0400% NaI(Tl) + PMTs Pinholes
LOrA-SPECT 0.75 mm 0.1100% CsI(Na) + PMTs Slit
CoALA-­SPECT 1.20 mm 0.0150% NaI(Tl) + PMTs Parallel-hole
A-SPECT 0.74 mm 0.0180% NaI(Tl) + PMTs Pinholes
Semi-SPECT 1.45 mm 0.0077% CdZnTe + Asic Pinholes
Mouse-­SPECT 1.70 mm 0.0029% NaI(Tl) + PMTs Pinholes

Table 2.7 Performance characteristics of preclinical SPECT/CT scanners from three major
manufacturers
Siemens Healthcare Mediso MILabs
Company scanners Inveon SPECT NanoSPECT U-SPECT+
Maximum FOV 110 × 130 mm 200 × 250 mm NC
Scintillator NaI (2 × 2 × 10) NaI (monolithic) NaI (monolithic)
Sensitivity (max) 0.04% 1.30% 1.00%
Resolution at CFOV (best) 0.60 mm 0.30 mm 0.25 mm

2.3.2 Scintillation Versus Direct Conversion

Over the past decades, thallium-doped Sodium Iodide has been the scintillating
crystal of choice when building SPECT imaging systems. Based on the Anger cam-
era, which consists in a monolithic crystal coupled to photomultiplier tubes, SPECT
detector heads have presented an intrinsic spatial resolution of about 3 mm. The
reliability, robustness, and relatively low cost of these detectors are the reasons why
they are still used today in preclinical systems. However, their intrinsic spatial reso-
lution requires the use of collimators presenting a magnification factor in order to
obtain spatial resolutions that are suitable for imaging the small animal.
The quest for an ever better intrinsic spatial resolution has led to the use of crys-
tal matrices. Pixelated crystal arrays comprising small tightly packed and optically
isolated crystals are now used in preclinical systems to ensure high intrinsic spatial
resolution. The intrinsic spatial resolution of the detector is approximately assume
to be the same as the crystal pitch, provided it is coupled to a high-resolution detec-
tor capable of resolving individual crystal elements, such as a PS-PMT or
SiPM. While a small crystal pitch is desirable for high spatial resolution, the thick-
ness must be sufficient to absorb most photons for good detection efficiency.
However, as the crystal length-to-width ratio increases, the light output decreases
due to internal reflections, which leads to an energy resolution loss [124–126].
Several inorganic scintillators were used in preclinical systems. Table 2.8 lists
the principal properties of thallium-doped Sodium Iodide (NaI[Ce] or NaI), thal-
lium-doped Cesium Iodide (CsI[Ce] or CsI), cerium-doped Yttrium Aluminium
Perovskite (YAlO3[Ce] or YAP), and cerium-doped Lanthanum Bromide (LaBr3[Ce]
or LaBr3), cerium-doped Lanthanum Chloride (LaCl3[Ce] or LaCl3) crystals.
42 D. Brasse and F. Boisson

Table 2.8 Properties of scintillating crystals commonly used in single-photon imaging

Scintillation crystal NaI CsI YAP LaBr3 LaCl3
Density (g cm−1) 3.67 4.51 5.55 5.29 3.79
Index of refraction 1.85 1.79 1.94 1.90 1.90
Luminescence (ph keV−1) 38 54 18 63 49
Peak emission wavelength (nm) 415 550 350 380 350
Emission decay time (ns) 250 1000 27 26 28
Attenuation length at 140 keV (cm) 0.41 0.28 0.58 0.29 0.37
Hygroscopicity Yes Poor No Yes Yes

Although the use of a crystal-photodetector pair remains the main choice for
detecting gamma rays, semiconductor detectors based on CdTe or CdZnTe were
used to directly convert gamma rays into electrical charges. With respect to scintil-
lation devices, solid-state detectors provide a larger conversion yield: typically
30,000 charges for a 140 keV energy release. As a consequence, the energy resolu-
tion is not limited by charge generation statistics but by other phenomena like elec-
trical noise or material uniformity. To date, the best devices achieve an energy
resolution below 2% at 140 keV, while standard systems are close to 5% at 140 keV
[127–129]. This opens a new path to properly investigate the development of dual-­
isotope imaging protocols [130].
Another advantage of semiconductor detectors is their extremely good spatial
resolution, which is not limited by light spreading and photon statistics but rather by
the readout circuitry. A typical device resolution is of the order of 2.5 mm, but the
use of high density readout [129] or sub-pixel positioning electronics [131] allows
to obtain an intrinsic resolution of few hundreds of micrometers (200–400 μm).
Additionally, CdTe- or CdZnTe-based detectors are nowadays integrated in small
modules that couple the semiconductor crystals and the readout electronics on the
same substrate. This module compactness presents an interesting way to build inno-
vative SPECT systems that enhance sensitivity and resolution.

2.3.3 Focus on Collimation

For over 50 years, several collimators have been proposed. The most known and
used are the parallel-hole collimator [132–134], the (multi-)pinhole collimator
[135–139], and the fan-beam/cone-beam collimator [140–143]. Although less used
or still the subject of current research, other collimators have been proposed, such
as coded aperture [120, 144] or slit-slat collimators [145–147]. All these collimators
present different resolution-sensitivity trade-offs, which are directly related to their
geometries. Mainly governed by collimation parameters, the resolution-sensitivity
trade-off is one of the main factors determining the collimator the most suitable for
an intended study. It is therefore essential to optimize these parameters to get the
best system performance. Several approaches have been proposed to optimize the
collimation parameters. The most commonly used method remains the use of Monte
Carlo methods. For instance, in 2004, Song et al. investigated the optimal pinhole
2 Instrumentation Challenges in (S)PE(C)T Systems 43

diameter and channel height using Monte Carlo simulations. Trade-off curves of
spatial resolution and sensitivity were obtained for pinhole diameters varying from
0.25 mm to 2 mm and channel heights varying from 0 to 1 mm (knife-edge) [148].
Similar works have been conducted on parallel-hole collimators. Monte Carlo simu-
lations have been used to optimize the hole diameter, the collimation height, the
septal thickness, and the collimation material [141, 148, 149].
Although these collimators enable a large number of studies, a low sensitivity
can limit the use of single-photon imaging and SPECT systems. One alternative to
the parallel-hole or the pinhole approaches is the slit-slat collimation [145–147].
The slit-slat profile combines fan-beam and parallel-beam properties and allows a
better detection efficiency than the multi-pinhole collimation [147]. Similar to the
pinhole and cone-beam collimation, the slit-slat approach presents an improved
transverse resolution provided by a transaxial magnification. Furthermore, the axial
septa lead to an extended axial FOV compared to pinhole imaging [146]. Figure 2.6
presents the common collimation profiles and their corresponding parameters.
A second alternative is the rotating slat collimator (RSC). Originally, the
design of a linear detector was proposed independently by Keyes (1975) and
Tosswill (1977) [150, 151]. In 1979, Entine et al. combined a CdTe detector with
a parallel plate collimator [152]. In 2001, Gagnon et al. used RSC in combination
with solid-­state detectors in the SOLSTICE (SOLid STate Imager with Compact
Electronics) system [153]. This collimation profile provides a better trade-off
between spatial resolution and detection efficiency than a parallel-hole collimator
[153, 154]. In addition, Webb (1993) previously showed that the use of a RSC
combined with a conventional SPECT detector leads to a sensitivity enhancement
of 40 times [155]. Although the pinhole approach provides a good trade-off

a b
h t
θ
e f

c d
w
h
θ
f a
f

Fig. 2.6 Schematics of four different collimation profiles. (a) Parallel-hole collimator. e corre-
sponds to the aperture dimension, h the collimation height, and t the septal thickness. (b) (Single)
pinhole collimator. f corresponds to the focal length and θ the opening angle. (c) Fan-beam colli-
mator and (d) slit-slat profile. d corresponds to the gap between two consecutive slats, a is the slat
height, and W the slit width
44 D. Brasse and F. Boisson

between spatial resolution and sensitivity, the price in most cases is a reduction in
the reconstructed field of view where RSC allows for a similar FOV to be main-
tained [136]. RSCs are fundamentally different from other collimators. RSCs do
not measure line integrals but plane integrals acquired over an extra rotation of the
detector around its own axis, called spin rotation. This rotation allows the com-
plete set of one-dimensional (1D) projections to be obtained [145, 146, 156],
which is then used to reconstruct two-­dimensional (2D) projections. However,
this necessary reconstruction step makes it difficult to directly compare RSC to
other collimators in terms of sensitivity. For parallel-hole collimators, sensitivity
is directly related to the collimator transparency, a relation biased by the recon-
struction step in the case of RSC.
More recently, Boisson et al. proposed a dedicated detection system presenting
both high sensitivity and submillimetric spatial resolution [157]. The innovative
aspect of this system was the use of a YAP:Ce crystal segmented into 32 slices of
0.570 mm wide, associated with a RSC. This 1D geometry however imposes the
rotation of the entire detection module (including the collimator), thus reducing the
reconstructed field of view. They demonstrated later the value of using a system
matrix to reconstruct 2D projections for a RSC system [158]. This work also aimed
at studying the resolution-sensitivity trade-offs obtained by varying different colli-
mation parameters: the slats height (H) and the gap between two consecutive slats
(g). The GATE (Geant4 Application for Tomographic Emission, [159]) simulation
platform was used to generate probability matrices corresponding to (H, g) couples
offering the best sensitivity-spatial resolution trade-off for specific applications.
Based on these preliminary simulation results, they showed that both a submillime-
ter spatial resolution and sensitivity greater than 0.2% could be obtained using opti-
mized collimation parameters considering a submillimetric intrinsic spatial
resolution [160]. In the context of a full SPECT system, such a preclinical prototype
opens new perspectives. A SPECT system consisting in four detection heads would
provide a sensitivity of about 0.8% while keeping (1) a submillimeter spatial resolu-
tion and (2) the initial 50 × 50 mm2 FOV.

2.3.4 Rotating Versus Stationary SPECT Systems

Most clinical and small-animal SPECT systems consist in detector-collimator pairs

mounted on a gantry that rotates around the subject considering either a precise
step-and-shoot or a continuous motion. Multiple detector systems allow sensitivity
improvement and reduce the need for a system to acquire projection data over a full
180° or 360°. However, small-animal SPECT systems based on a pinhole collima-
tion approach often suffer from a limited FOV due to high magnification or small
compact detectors. One option to overcome this limitation is to increase the radius
of rotation, but this results, in most cases, in the reduction of both the system’s spa-
tial resolution and sensitivity. Despite these possible limitations, currently available
commercial systems achieve submillimeter spatial resolution and absolute sensitiv-
ity greater than 1%.
2 Instrumentation Challenges in (S)PE(C)T Systems 45

Several systems have been developed based on another approach: stationary

detectors that surround the animal and multiple stationary or rotating pinholes [118,
123, 136, 161, 162]. Based on multi-pinholes, these systems present a high sensitiv-
ity, while avoiding mechanical misalignments due to gantry rotation. But the most
important advantage of stationary SPECT systems is their ability to perform
dynamic SPECT of tracers with fast kinetics [118, 162] due to the simultaneous
acquisition of the entire set of projections, which eliminates potential reconstruction
errors due to redistribution of the radiopharmaceutical.

2.3.5 Multimodality Systems

It is becoming increasingly important in research applications to accurately localize

radiopharmaceutical biodistribution relative to known anatomical structures. Both
X-ray CT and MR provide detailed anatomical information, which is highly com-
plementary to the SPECT study. Thus, it is common in commercial systems for PET
or SPECT to be one component of a dual- or trimodality imaging system. There are
several possible combinations and approaches to multimodality imaging, including
PET/CT, SPECT/CT, PET/MR, and SPECT/MR. In the following section, we will
focus on the latest multimodal developments: SPECT/MR imaging systems.
Like instrumental developments dedicated to PET/MRI, recent years have seen
several projects to develop SPECT inserts that can be integrated into an
MRI. Although SPECT/CT systems remain the reference in a multimodal approach
for single-photon imaging, combining SPECT and MRI information opens new
perspectives.
Some technical challenges inherent in the development of SPECT and PET
inserts are similar, namely, the constraints of size or the susceptibility of detectors
in the magnetic field. However, in the case of SPECT, an essential element adds to
the complexity of development: the collimator. The need to use a collimator adds to
the constraint of space and thereby reduces the useful field of view of the insert. It
is also essential to design a collimator in a material, which does not disturb the
magnetic field of the MRI, but with a geometry that would provide acceptable per-
formance. Typically, a SPECT insert should present a spatial resolution of one mil-
limeter or less and the greatest efficiency possible. Here we only present recent
instrumental developments of SPECT inserts dedicated to simultaneous SPECT/
MRI. We invite readers to read a full review published by Van Audenhaege et al.
(2015) concerning the selection, optimization, and manufacturing methods of col-
limators within the framework of clinical and preclinical imaging and mentioning
the SPECT/MRI approach [163]. Although SPECT inserts are still in early stages of
development, the following paragraph presents the latest research developments in
SPECT instrumentation dedicated to simultaneous SPECT/MRI and their
performance.
The benefits of combining SPECT and MRI information are undeniable.
However, the concept of simultaneous SPECT/MRI is relatively recent, with the
first prototypes appearing in the late 2000s. In 2009, Wagenaar et al. developed a
46 D. Brasse and F. Boisson

stationary SPECT system for simultaneous preclinical SPECT/MRI based on

CZT. The system consists in four rings of eight CZT modules, 16 × 16 pixels of
1.6 mm each. They considered a cylindrical arrangement of pinholes with a mag-
nification of approximately 1.0. This stationary tomographic SPECT system is
capable of obtaining 24 views simultaneously. Two different arrangements of pin-
holes are used to perform whole-body or organ-specific acquisitions. A spatial
resolution of about 1.5 mm FWHM with the 1.6 mm pixels and a pinhole collima-
tor of 1.0 mm diameter was reported. Limiting factors such as a limited volume
within the magnet bore and a magnification of 1.0 are mostly responsible for this
loss in terms of spatial resolution. In 2011, Meier et al. continued this work using
the same CZT detector modules [164]. They reported an average energy resolution
of 5.4 keV FWHM at 122 keV. Sensitivity and spatial resolution measurements
were performed considering a magnification of 0.85. They obtained a sensitivity of
0.062% using a 2 mm diameter pinhole and 2.2 mm FWHM spatial resolution for
a 0.5 mm aperture. These collaborative works presented preliminary results of the
first SPECT insert dedicated to simultaneous SPECT/MRI. In 2014, Cai et al. pro-
posed an alternative design of a stationary SPECT insert based on ten energy-
resolved photon-counting (ERPC) detectors assembled as a compact ring [165].
The distance between the opposite detectors is 15.6 cm, and the detection area of
each detector is 22.5 × 45 mm2. Each detector proposes four 300 or 500 μm pin-
holes, with a magnification factor of 1.2. The object-to-pinhole distance is designed
to be around 36 mm. Pinhole inserts are made of cast platinum (90%)-iridium
(10%) alloy, which provides the maximum stopping power and are compatible
with MR scanners. A total of 40 pinholes were used in the 10-detector ring system
leading to sensitivity up to 0.07%. In terms of spatial resolution, the system was
able to clearly resolve the 500 mm features of a homemade Jaszczak phantom
using the 300 mm pinholes.
In 2013, the INSERT research project was funded by the Seventh Framework
Programme of the European Commission. The objective of the INSERT project is
to develop an innovative system combining SPECT and MRI for simultaneous
imaging. The system will be validated at both the preclinical and clinical levels
thanks to the creation of animal models and to a pilot study, which will involve
patients with glioma. In 2014, Busca et al. published the simulation results of the
expected performance of INSERT for both preclinical and clinical imaging [166].
The fundamental unit is a 5 × 5 cm gamma camera, based on the Anger architecture
with a continuous scintillator (e.g., CsI(Tl), NaI(Tl)), read out by an array of silicon
photodetectors. Two possible solutions have been taken into account for the scintil-
lator readout: silicon drift detectors (SDDs) and SiPMs. The basic SDD photodetec-
tor unit consists of a matrix composed of nine square SDDs, each one of 8 × 8 mm
active area, in a final 3 × 3 format, for an overall size of 26.08 × 26.08 mm2. In the
simulation, a 8-mm-thick CsI(Tl) scintillator with parallelepiped edges was consid-
ered, providing a final UFOV of about 4 × 4 cm2. The intrinsic spatial resolution
(FWHM) was estimated by a Gaussian fitting of the reconstructed points with val-
ues between 0.85 mm (center) and 1.05 mm (borders). The energy resolution was
found between 9.6 and 14.6% when the system is cooled down to −20 °C. The
2 Instrumentation Challenges in (S)PE(C)T Systems 47

second option is based on a detection plane composed by 10 × 10 single SiPMs. To

shortcut the output terminal, they considered a channel merging with a single
merged unit composed by 2 × 2 SiPMs. Also in this simulation, a CsI(Tl) scintillator
wrapped with Teflon and with parallelepiped edges is considered. The resulting
UFOV is roughly 4 × 4 cm2, and the intrinsic spatial resolution was estimated
between 1.3 mm (center) and 1.5 mm (borders). The energy resolution has been
evaluated with both analytical formulation and Monte Carlo simulations between
11.6 and 15.9% also in this case. The preclinical SPECT system will be composed
by one circular ring of detection modules designed for MRI (7 T and 9.4 T) with
20 cm aperture. The necessary UFOV of the SPECT system for mouse brain would
be 16 mm transaxially x 11 mm axially, whereas for rat brain 20 × 20 mm.
Considering a FOV of the detectors larger in both transaxial and axial directions
than the animal size, a multi-pinhole technology with divergent projections was
chosen for the preclinical imaging system. While a 1 mm spatial resolution was
measured for both configurations, effective sensitivity top values of 0.22% and
0.094% were obtained for the mouse and rat brain configurations, respectively. The
aim of the clinical system is to image a whole human brain (FOV 20 × 15 cm) to
avoid the need for prior knowledge of tumor location from other modalities. The
SPECT system should be composed of at least two adjacent circular rings of detec-
tion modules to be inserted in a 59 cm aperture 3 T PET/MRI system (Siemens).
The clinical design is oriented to maximize sensitivity with an overall resolution
maintained at a level similar to that achieved with conventional gamma camera-­
based SPECT (8–10 mm). Higher sensitivity can be achieved with the use of slit-­
slat collimator. Simulations of a Derenzo-type phantom were performed considering
a slit-slat collimator and a 42.4 mm projection UFOV, which led to a targeted reso-
lution of 8 mm.

2.4 Summary

Whether based on academic or commercial developments, preclinical SPECT sys-

tems now offer a spatial resolution of less than 0.3 mm and a sensitivity of up to
1.3%, where preclinical PET systems’ sensitivity is on the order of 10% and a spa-
tial resolution down to 0.6 mm. Although it is fundamentally not possible to com-
pare the performance of PET and SPECT systems, the latter are nevertheless
governed by a set of intrinsic trade-offs. Preclinical PET systems are intrinsically
not limited in terms of sensitivity, unlike SPECT systems whose sensitivity is
strongly impacted by the collimation stage. Conversely, PET systems resolution is
limited by various factors such as the positron range among others, unlike SPECT
systems, which are intrinsically not limited in terms of spatial resolution. However,
it should be noted that the improvement of the spatial resolution of SPECT systems
almost inevitably leads to the reduction of the system’s field of view.
Unlike PET systems dedicated to small-animal imaging, for which NEMA has
defined the standard NU 4–2008, preclinical SPECT systems are, to date, subject to
no standards. A task force has recently been assigned to establish what could be the
48 D. Brasse and F. Boisson

new NEMA NU-5, also including a new phantom, adapted to small-animal SPECT
systems and to define the IQ parameters.
PET and SPECT systems show constantly improving performance. These
modalities should not be opposed but rather associated in a multimodal approach to
open new perspectives in the field of molecular imaging.

References
1. Massoud TF, Gambhir SS. Molecular imaging in living subjects: seeing fundamental biologi-
cal processes in a new light. Genes Dev. 2003;17:545.
2. James ML, Gambhir SS. A molecular imaging primer: modalities, imaging agents, and appli-
cations. Physiol Rev. 2012;92:897.
3. Flohr TG, et al. Multi-detector row CT systems and image-reconstruction techniques.
Radiology. 2005;235:756.
4. Brasse D, et al. Towards an inline reconstruction architecture for micro-CT systems. Phys
Med Biol. 2005;50:5799.
5. O’Connor MK, Kemp BJ. Single-photon emission computed tomography/computed tomog-
raphy: basic instrumentation and innovations. Semin Nucl Med. 2006;36:258.
6. Madsen MT. Recent advances in SPECT imaging. J Nucl Med. 2007;48:661.
7. Khalil MM, et al. Molecular SPECT imaging: an overview. Int J Mol Imaging.
2011;2011:796025.
8. Levin CS, Zaidi H. Current trends in preclinical PET system design. PET Clinics.
2007;2:125.
9. Goertzen AL, et al. NEMA NU 4-2008 comparison of preclinical PET imaging systems. J
Nucl Med. 2012;53:1300.
10. Hollingworth W, et al. The diagnostic and therapeutic impact of MRI: an observational multi-­
centre study. Clin Radiol. 2000;55:825.
11. Pichler BJ, et al. PET/MRI: paving the way for the next generation of Clinical multimodality
imaging applications. J Nucl Med. 2010;51:333.
12. Ebbini ES, Ter Haar G. Ultrasound-guided therapeutic focused ultrasound: current status and
future directions. Int J Hyperth. 2015;23:1.
13. Anger HO. Scintillation camera. Rev Sci Instr. 1958;29:27.
14. Pichler BJ, et al. Performance test of an LSO-APD detector in a 7-T MRI scanner for simul-
taneous PET/MRI. J Nucl Med. 2006;47:639.
15. Keereman V, et al. Temperature dependence of APD-based PET scanners. Med Phys.
2013;40:092506.
16. Chatziioannou AF. PET scanners dedicated to molecular imaging of small animal models.
Mol Imaging Biol. 2002;4:47–63.
17. Casey ME, Nutt R. A multicrystal two dimensional BGO detector system for positron emis-
sion tomography. IEEE Trans Nucl Sci. 1986;33:460–3.
18. Conti M, et al. Characterization of 176Lu background in LSO-based PET scanners. Phys Med
Biol. 2017;62:3700–11.
19. Lecoq P. Development of new scintillators for medical applications. Nucl Instrum Methods
Phys Res A. 2016;809:130–9.
20. Derenzo SE, et al. The quest for the ideal inorganic scintillator. Nucl Instrum Methods Phys
Res A. 2003;505:111.
21. Otte AN, et al. A test of silicon photomultipliers as readout for PET. Nucl Instrum Methods
Phys Res A. 2005;545:705.
22. Ginzburg D, et al. Optimizing the design of a silicon photomultiplier-based radiation
detector. Nucl Instrum Methods Phys Res A. 2011;652:474. https://doi.org/10.1016/j.
nima.2011.01.022.
2 Instrumentation Challenges in (S)PE(C)T Systems 49

23. Thiessen JD, et al. Performance evaluation of SensL SiPM arrays for high-resolution
PET. IEEE Nucl Sci Symp Conf Rec. 2013:M21.
24. Frach T, et al. The digital silicon photomultiplier - principle of operation and intrinsic detec-
tor performance. In: 2009 IEEE Nucl. Sci. Symp. Conf. Rec. (Orlando, FL, USA, 24 Oct.–1
Nov.); 2009, p. 1959–1965.
25. Gu Y, Levin CS. Study of electrode pattern design for a CZT-based detector. Phys Med Biol.
2014;59:2599–621.
26. Yoo, et al. Simulation for CZT Compton PET (Maximization of the efficiency for PET using
Compton event). Nucl Instrum Methods Phys Res A. 2011;652:713–6.
27. Gu Y, et al. Study of a high-resolution, 3D positioning cadmium zinc telluride detector for
PET. Phys Med Biol. 2011;56:1563–2011.
28. Levin CS. Promising new photon detection concepts for high-resolution clinical and preclini-
cal PET. J Nucl Med. 2012;53:167–70.
29. Abbaszadeh S, et al. Positioning true coincidences that undergo inter- and intra-crystal scat-
ter for a sub-mm resolution cadmium zinc telluride-based PET system. Phys Med Biol.
2018;63(2):025012. https://doi.org/10.1088/1361-6560/aa9a2b.
30. Thompson C. The effects of detector material and structure on PET spatial resolution and
efficiency. IEEE Trans Nucl Sci. 1990;37:718–24.
31. Chung YH, et al. Characterization of dual layer phoswich detector performance for small
animal PET using Monte Carlo simulation. Phys Med Biol. 2004;49:2881.
32. Jung JH, et al. Optimization of LSO/LuYAP phoswich detector for small animal PET. Nucl
Instrum Meth Phys Res A. 2007;571:669–75.
33. Eriksson L, et al. Design Considerations of phoswich detectors for high resolution positron
emission tomography. IEEE Trans Nucl Sci. 2009;56:182–8.
34. Ito M, et al. A Four-Layer DOI detector with a relative offset for use in an animal PET sys-
tem. IEEE Trans Nucl Sci. 2010;57:976–81.
35. Roncali, et al. Design considerations for DOI-encoding PET detectors using phosphor-coated
crystals. IEEE Trans Nucl Sci. 2014;61:67–73.
36. Shao Y, et al. Design studies of a high resolution PET detector using APD arrays. IEEE Trans
Nucl Sci. 2000;47:1051–7.
37. Shao Y, et al. Dual APD array readout of LSO crystals: Optimization of crystal surface treat-
ment. IEEE Trans Nucl Sci. 2002;49:649–54.
38. Dokhale PA, et al. Performance measurements of a depth-encoding PET detector-
module based on position-sensitive avalanche photodiode readout. Phys Med Biol.
2004;49:4293.
39. Dokhale PA, et al. Intrinsic spatial resolution and parallax correction using depth-­encoding
PET detector modules based on position-sensitive APD readout. IEEE Trans Nucl Sci.
2006;53:2666–70.
40. James SS, et al. Experimental characterization and system simulations of depth of interaction
PET detectors using 0.5 mm and 0.7 mm LSO arrays. Phys Med Biol. 2009;54:4605–19.
41. Yang Y, et al. Depth of interaction resolution measurements for a high resolution PET detec-
tor using position sensitive avalanche photodiodes. Phys Med Biol. 2006;51:2131–42.
42. Moses WW, et al. A room temperature LSO/PIN photodiode PET detector module that mea-
sures depth of interaction. IEEE Trans Nucl Sci. 1995;42:1085–9.
43. Chaudhari AJ, et al. PSPMT/APD hybrid DOI detectors for the PET component of a dedi-
cated breast PET/CT system: A feasibility study. IEEE Trans Nucl Sci. 2008;55:853–61.
44. Godinez F, et al. Characterization of a high-resolution hybrid DOI detector for a dedicated
breast PET/CT scanner. Phys Med Biol. 2012;57:3435–49.
45. Godinez F, et al. Development of an ultra high resolution PET scanner for imaging rodent
paws: PawPET. IEEE Trans Rad Plas Med Sci. 2018;2:7–16.
46. Ter-Pogossian MM, et al. A multislice positron emission computed tomograph (PETT IV)
yielding transverse and longitudinal images. Radiology. 1978;128:477–84.
47. Shimizu K, et al. Development of 3-D detector system for positron CT. IEEE Trans Nucl Sci.
1988;35:717–20.
50 D. Brasse and F. Boisson

48. Braem A, et al. Feasibility of a novel design of high resolution parallax-free Compton
enhanced PET scanner dedicated to brain research. Phys Med Biol. 2004;49:2547.
49. Jan S, et al. GePEToS: A Geant4 Monte Carlo simulation package for positron emission
tomography. IEEE Trans Nucl Sci. 2005;52:102–6.
50. Salvador S, et al. Design of a high performances small animal PET system with axial oriented
crystals and DOI capability. IEEE Trans Nucl Sci. 2009;56:17–23.
51. Beltrame P, et al. The AX-PET demonstrator-Design, construction and characterization. Nucl
Instrum Methods Phys Res A. 2011;654:546–59.
52. Casella C, et al. A high resolution TOF-PET concept with axial geometry and digital SiPM
readout. Nucl Instrum Methods Phys Res A. 2014;736:161–8.
53. Maas MC, et al. Monolithic scintillator PET detectors with intrinsic depth of interaction cor-
rection. Phys Med Biol. 2009;54:1893–908.
54. van Dam HT, et al. Improved nearest neighbor methods for gamma photon interaction
position determination in monolithic scintillator PET detectors. IEEE Trans Nucl Sci.
2011;58:2139–47.
55. van Dam HT, et al. A practical method for determining depth of interaction in monolithic
PET scintillator detectors. Phys Med Biol. 2011;56:4135–45.
56. Joung J, et al. cMICE: a high resolution animal PET using continuous LSO with a statistics
based positioning scheme. Nucl Instrum Methods Phys Res A. 2002;489:584–98.
57. Ling T, et al. Depth of interaction decoding of a continuous crystal detector module. Phys
Med Biol. 2007;52:2213–28.
58. Ling T, et al. Parametric positioning of a continuous crystal PET detector with depth of inter-
action decoding. Phys Med Biol. 2008;53:1843–63.
59. Miyaoka RS, et al. Detector response modeling light for a thick slab continuous detector. J
Nucl Sci Technol. 2008;45:634–8.
60. Miyaoka RS, et al. New miniature continuous crystal element (cMICE) detector geometries.
IEEE Nucl Sci Symp Conf Rec. 2009;2009:3639.
61. Maas MC, et al. Experimental characterization of monolithic crystal small-animal PET
detectors read out by APD arrays. IEEE Trans Nucl Sci. 2006;53:1071–7.
62. Bruyndonckx P, et al. Towards a continuous crystal APD-based PET detector design. Nucl
Instrum Methods Phys Res A. 2007;571:182–6.
63. Ling T, et al. Performance comparisons of continuous Miniature Crystal elemento (cMICE)
detectors. IEEE Trans Nucl Sci. 2006;53:2513–8.
64. Schaart DR, et al. A novel, SiPM-array-based monolithic scintillator detector for PET. Phys
Med Biol. 2009;54:3501–12.
65. van der Lann DJ, et al. Optical simulation of monolithic scintillator detectors using GATE /
Geant4. Phys Med Biol. 2010;55:1659–75.
66. van Dam HT, et al. Sub-200 ps CRT monolithic scintillator in PET detectors using digi-
tal SiPM arrays and maximum likelihood estimation time interaction. Phys Med Biol.
2013;58:3243–57.
67. Seifert S, et al. First characterization of a digital SiPM based time-of-flight PET detector with
1 mm spatial resolution. Phys Med Biol. 2013;58:3061–74.
68. Borghi G, et al. A 32 mm × 32 mm × 22 mm monolithic LYSO:Ce detector with dual-sided
digital photon counter readout for ultra-performance TOF-PET and TOF-PET/MRI. Phys
Med Biol. 2016;61:4929–49.
69. Benlloch JM, et al. Scanner calibration of a small animal PET Camera based on continuous
LSO crystals and flat panel PSPMTs. Nucl Instrum Methods Phys Res A. 2007;571:26–9.
70. Llosa G, et al. Characterization of a PET detector head based on continuous LYSO
crystals and monolithic, 64-pixel silicon photomultiplier arrays. Phys Med Biol.
2010;55:72997315.
71. Gonzales A, et al. Small animal PET based on 16x16 TSV-MPPCs and monolithic crystals.
Eur J Nucl Med Mol Imaging Physics. 2015;2:A16.
72. Yoshida E, et al. Basic performance of a wide area PET detector with a monolithic scintilla-
tor. Phys Radiol Technol. 2011;4:4134–9.
2 Instrumentation Challenges in (S)PE(C)T Systems 51

73. Spinks TJ, et al. Quantitative PET and SPECT performance characteristics of the Albira
Trimodal pre-clinical tomograph. Phys Med Biol. 2014;59:715–31.
74. Szanda I, et al. National Electrical Manufacturers Association NU-4 performance evalua-
tion of the PET component of the NanoPET/CT preclinical PET/CT scanner. J Nucl Med.
2011;52:1741–7.
75. Belcari N, et al. NEMA NU-4 performance evaluation of the IRIS PET/CT preclinical scan-
ner. IEEE Trans Rad Plas Med Sci. 2017;1:301–9.
76. Bergeron M, et al. Imaging performance of LabPET APD-based digital PET scanners for
pre-clinical research. Phys Med Biol. 2014;59:661.
77. Wang, et al. Performance evaluation of the GE Healthcare eXplore VISTA dual-ring small-­
animal PET scanner. J Nucl Med. 2006;47:1891–900.
78. Habte F, Levin CS. Investigation of low noise, low cost readout electronics for high sensitiv-
ity PET systems based on Avalanche Photodiode arrays. IEEE Nucl Sci Symp Conf Rec.
2002;2:661.
79. Monzo JM, et al. Evaluation of a timing integrated circuit architecture for continuous crystal
and SiPM based PET systems. JINST. 2012;8:C03017.
80. Espana S, et al. Performance evaluation of SiPM photodetectors for PET imaging in the pres-
ence of magnetic fields. Nucl Instrum Methods Phys Res A. 2010;613:308.
81. Zeng H, et al. Design of a novel front-end readout ASIC for pet imaging system. J Signal
Inform Proc. 2013;4:129.
82. Dey S, et al. Impact of analog IC impairments in SiPM interface electronics. IEEE Trans Nucl
Sci. 2012;2012:3572.
83. Jarron P, et al. Time based readout of a silicon photomultiplier (SiPM) for Time Of Flight
Positron Emission Tomography (TOF-PET). IEEE Nucl Sci Symp Conf Rec. 2009.
84. Del Guerra A, et al. Silicon Photomultipliers (SiPM) as novel photodetectors for PET. Nucl
Instrum Methods Phys Res A. 2011;648:S232.
85. Choong WS, et al. High-performance electronics for time-of-flight PET systems. JINST.
2013;8:T01006.
86. Roloa MD, et al. TOFPET ASIC for PET applications. JINST. 2013;8:C02050.
87. Streun M, et al. Coincidence detection by digital processing of free-running sampled pulses.
Nucl Instrum Methods Phys Res A. 2002;487:530.
88. Siegel S, et al. Simple charge division readouts for imaging scintillator arrays using a multi-­
channel PMT. IEEE Trans Nucl Sci. 1996;43:1634–996.
89. Popov V, et al. Readout electronics for multianode photomultiplier tubes with pad matrix
anode layout. IEEE Nucl Sci Symp Conf Rec. 2003.
90. Boisson F, et al. Description and properties of a resistive network applied to emission tomog-
raphy detector readouts. Nucl Instrum Methods Phys Res A. 2017;872:100–6.
91. Vandenberghe S, Marsden PK. PET-MRI: a review of challenges and solutions in the devel-
opment of integrated multimodality imaging. Phys Med Biol. 2015;60:R115.
92. Shao Y, et al. Simultaneous PET and MR imaging. Phys Med Biol. 1997;42:1965.
93. Shao Y, et al. Development of a PET detector system compatible with MRI/NMR systems.
IEEE Trans Nucl Sci. 1997;44.
94. Pichler B, et al. Studies with a prototype high resolution PET scanner based on LSO-APD
modules. IEEE Trans Nucl Sci. 1998;45:1298.
95. Ziegler SI, et al. A prototype high-resolution animal positron tomograph with avalanche pho-
todiode arrays and LSO crystals. Eur J Nucl Med Mol Imaging. 2001;28:136.
96. Grazioso R, et al. APD-based PET detector for simultaneous PET/MR imaging. Nucl Instrum
Methods Phys Res A. 2006;569:301.
97. Delso G, et al. Performance measurements of the siemens mMR integrated wholebody PET/
MR scanner. J Nucl Med. 2011;52:1914.
98. Maramraju SH, et al. Small animal simultaneous PET/MRI: initial experiences in a 9.4 T
microMRI. Phys Med Biol. 2011;56:2459.
99. Peng H, et al. Investigation of a clinical PET detector module design that employs large-area
avalanche photodetectors. Phys Med Biol. 2011;56:3603.
52 D. Brasse and F. Boisson

100. Kolb A, et al. Technical performance evaluation of a human brain PET/MRI system. Eur
Radiol. 2012;22:1776.
101. Judenhofer MS, et al. Simultaneous PET-MRI: a new approach for functional and morpho-
logical imaging. Nat Med. 2006;14:459.
102. Catana C, et al. Simultaneous in vivo positron emission tomography and magnetic resonance
imaging. PNAS. 2008;105:3705.
103. Kwon SI, et al. Development of small-animal PET prototype using silicon photomultiplier
(SiPM): initial results of phantom and animal imaging studies. J Nucl Med. 2011;52:572.
104. Yoon HS, et al. Initial results of simultaneous PET/MRI experiments with an MRI-compatible
silicon photomultiplier PET scanner. J Nucl Med. 2012;53:608.
105. Ko, et al. New high performance SiPM PET insert to 9.4-T MR scanner for simultaneous
PET/MRI studies. J Nucl Med. 2013:54.
106. Ko, et al. Evaluation of a silicon photomultiplier PET insert for simultaneous PET and MR
imaging. Med Phys. 2016;43:72–83.
107. Ko, et al. Simultaneous multiparametric PET/MRI with silicon photomultiplier PET and
ultra-high-field MRI for small-animal imaging. J Nucl Med. 2016;57:1309–15.
108. Yamamoto S, et al. Development of a Si-PM-based high-resolution PET system for small
animals. Phys Med Biol. 2010;55:5817–31.
109. Yamamoto S, et al. Interference between PET and MRI sub-systems in a siliconphotomultiplier-­
based PET/MRI system. Phys Med Biol. 2011;56:4147.
110. Yamamoto S, et al. Simultaneous imaging using Si-PM-based PET and MRI for development
of an integrated PET/MRI system. Phys Med Biol. 2012;57:N1.
111. Weissler B, et al. MR compatibility aspects of a silicon photomultiplier-based PET/RF insert
with integrated digitisation. Phys Med Biol. 2014;59:5119.
112. Wehner J, et al. PET/MRI insert using digital SiPMs: investigation of MR-compatibility.
Nucl Instrum Methods Phys Res A. 2014;734:116–21.
113. Wehner J, et al. MR-compatibility assessment of the first preclinical PET-MRI insert equipped
with digital silicon photomultipliers. Phys Med Biol. 2015;60:2231–55.
114. Schug D, et al. Initial PET performance evaluation of a preclinical insert for PET/MRI with
digital SiPM technology. Phys Med Biol. 2016;61:2851–78.
115. Parl C, et al. A novel optically transparent RF shielding for fully integrated PET/MRI sys-
tems. Phys Med Biol. 2017;62:7357–78.
116. Jaszczak RJ. The early years of single photon emission computed tomography (SPECT): an
anthology of selected reminiscences. Phys Med Biol. 2006;51:R99–R115.
117. Sajedia S, et al. Design and development of a high resolution animal SPECT scanner dedi-
cated for rat and mouse imaging. Nucl Instrum Methods Phys Res A. 2014;741:169–76.
118. Furenlid LR, et al. FastSPECT II: A Second-Generation High-Resolution Dynamic SPECT
Imager. IEEE Trans Nucl Sci. 2004;51:631–5.
119. Walrand S, et al. Evaluation of novel whole-body high-resolution rodent SPECT (Linoview)
based on direct acquisition of linogram projections. J Nucl Med. 2005;46:1872–80.
120. Meikle SR, et al. A prototype coded aperture detector for small animal SPECT. IEEE Trans
Nucl Sci. 2002;49:2167–71.
121. McElroy DP, et al. Performance evaluation of A-SPECT: a high resolution desktop pinhole
SPECT system for imaging small animals. IEEE Trans Nucl Sci. 2002;49:2139–47.
122. Kim H, et al. SemiSPECT: A small-animal single-photon emission computed tomography
(SPECT) imager based on eight cadmium zinc telluride (CZT) detector arrays. Med Phys.
2006;33:465–74.
123. Goertzen AL, et al. First results from the high-resolution mouseSPECT annular scintillation
camera. Trans Med Imaging. 2005;24:863–7.
124. Barrett HH, Hunter WCJ. Detectors for small-animal SPECT I. Overview of technologies.
In: Kupinski MA, Barrett HH, editors. Small animal SPECT imaging. New York: Springer;
2005.
125. Wirrwar A, et al. The optimal crystal geometry for small-field-of-view gamma cameras:
arrays or disks? IEEE Nucl Sci Symp Conf Rec. 2000;3:91–3.
2 Instrumentation Challenges in (S)PE(C)T Systems 53

126. Wirrwar A, et al. Influence of crystal geometry and wall reflectivity on scintillation photon
yield and energy resolution. IEEE Nucl Sci Symp Conf Rec. 1999;3:1443–5.
127. Eisen Y, et al. CdTe and CdZnTe X-ray and gamma-ray detectors for imaging systems. IEEE
Trans Nucl Sci. 2004;51:1191.
128. Verger L, et al. “Performance and perspectives of a CdZnTe-based gamma camera for medi-
cal imaging,” Nuclear Science. IEEE Trans Nucl Sci. 2004;51:3111–7.
129. Meng LJ, et al. Preliminary evaluation of a novel energy-resolved photon-counting gamma
ray detector. Nucl Instrum Methods Phys Res A. 2009;604:548–54.
130. Ben-Haim S, et al. Simultaneous dual-radionuclide myocardial perfusion imaging with a
solid-state dedicated cardiac camera. Eur J Nucl Med Mol Imaging. 2010;37:1710–21.
131. Montemont G, et al. Evaluation of a CZT gamma-ray detection module concept for SPECT,
IEEE Nucl Sci Symp Conf Rec. 2012;4091–4097.
132. Del Guerra A, et al. An integrated PET-SPECT small animal imager: preliminary results.
IEEE Trans Nucl Sci. 2000;47:4.
133. Loudos GK, et al. A 3D high-resolution gamma camera for radiopharmaceutical studies with
small animals. Appl Radiat Isot. 2003;58:501–8.
134. R. Pani, et al. A novel parallel hole collimator for high resolution SPET imaging with a com-
pact LaBr3 gamma camera. IEEE Nucl Sci Symp Conf Rec. 2008;3824–8.
135. Schramm NU, et al. High-resolution SPECT using multi-pinhole collimation. IEEE Trans
Nucl Sci. 2003;50:315–20.
136. Beekman FJ, Vastenhouw B. Design and Simulation of a high-resolution stationary SPECT
system for small animals. Phys Med Biol. 2004;49:4579–92.
137. Cao Z, et al. Optimal number of pinhole in multi-pinhole SPECT for mouse brain imaging-a
simulation study. Phys Med Biol. 2005;50:4609–24.
138. Funk T, et al. A novel approach to multi-pinhole SPECT for myocardial perfusion imaging. J
Nucl Med. 2006;47:595–602.
139. Vanhove C, et al. Three-pinhole collimator to improve axial spatial resolution and sensitivity
in pinhole SPECT. Eur J Nucl Med Mol Imaging. 2008;35:407–15.
140. Gullberg GT, et al. Estimation of geometrical parameters for cone beam tomography. Med
Phys. 1989;17:264–72.
141. Wang H, et al. Determination of collimator and acquisition parameters for astigmatic SPECT
imaging. IEEE Nucl Sci Symp Conf Rec. 1995;2:1116–20.
142. Lewis DP and Tsui BMW Estimation of half fan-beam collimator parameters for brain spect
with an l-shaped dual camera spect system. IEEE Nucl Sci Symp Conf Rec. 1998;1052–6.
143. Qi JY, Huesman RH. Effect of errors in the system matrix on maximum a posteriori image
reconstruction. Phys Med Biol. 2005;50:3297–312.
144. Mu Z, et al. Recent progress on SPECT imaging with near-field coded aperture collimation:
a small animal study. IEEE Nucl Sci Symp Conf Rec. 2010;3450–3.
145. Zeng GL, Gagnon D. Image reconstruction algorithm for a SPECT system with a convergent
rotating slat collimator. IEEE Trans Nucl Sci. 2004;51:1.
146. Metzler SD, et al. On-axis sensitivity and resolution of a slit-slat collimator. J Nucl Med.
2006;47:1884–90.
147. Novak JR, et al. Verification of the sensitivity and resolution dependence on the incidence
angle for slit-slat collimation. Phys Med Biol. 2008;53:953–66.
148. Song TY, et al. Performance amelioration for small animal SPECT using optimized
pinhole collimator and image correction technique. IEEE Nucl Sci Symp Conf Rec.
2005;52:1396–400.
149. Rajaee A, et al. Simulation study of the influence of collimator material on image quality
improvement for high energy photons in nuclear medicine using MCNP code. J Theor Appl
Phys. 2011;4:13–8.
150. Keyes W. The fan beam camera. Phys Med Biol. 1975;20:489–91.
151. Tosswill C. Computerized rotating laminar collimation imaging system US Patent Application
646 (granted December 1977), p. 917–67.
152. Entine G. Cadmium telluride gamma camera. IEEE Trans Nucl Sci. 1979;26:552–8.
54 D. Brasse and F. Boisson

153. Gagnon D, et al. Design considerations for a new solid-state gamma-camera: SOLSTICE
Proc. IEEE Nucl Sci Symp Conf Rec. 2001;2:1156–60.
154. Van Holen R, et al. Comparing planar image quality of rotating slat and parallel hole collima-
tion: influence of system modeling. Phys Med Biol. 2008;53:19892002.
155. Webb S, et al. Geometric efficiency of a rotating slit collimator for improved planar gamma
camera imaging. Phys Med Biol. 1993;38:627–38.
156. Lodge MA, et al. A prototype rotating slat collimator for single photon emission computed
tomography. Trans Med Imaging. 1996;15:500–11.
157. Boisson F, et al. Characterization of a rotating slat collimator system dedicated to small ani-
mal imaging. Phys Med Biol. 2011;56:1471–85.
158. Boisson F, et al. Assessment of a fast generated analytical matrix for rotating slat collimation
iterative reconstruction: a possible method to optimize the collimation profile. Phys Med
Biol. 2015;60:2403–19.
159. Jan S, et al. GATE: a simulation toolkit for PET and SPECT. Phys Med Biol. 2004;49:4543–61.
160. Boisson F, et al. Determination of optimal collimation parameters for a rotating slat collima-
tor system: a system matrix method using ML-EM. Phys Med Biol. 2016;61:2302–18.
161. Van der Have F, et al. U-SPECT-II: An ultra-high-resolution device for molecular small-­
animal imaging. J Nucl Med. 2009;50:599–605.
162. Fresneau N, et al. Design of a serotonin 4 receptor radiotracer with decreased lipophilicity for
single photon emission computed tomography. Eur J Med Chem. 2015;94:386–96. https://
doi.org/10.1016/j.ejmech.2015.03.017.
163. Van Audenhaege K, et al. Collimator design for a multipinhole brain SPECT insert for
MRI. Med Phys. 2015;42:6679.
164. Meier D, et al. A SPECT camera for combined MRI and SPECT for small animals. Nucl
Instrum Methods Phys Res A. 2011;652:731–4.
165. Cai L, et al. MRC-SPECT: A sub-500 mm resolution MR-compatible SPECT system for
simultaneous dual-modality study of small animals. Nucl Instrum Methods Phys Res A.
2014;734:147–51.
166. Busca P, et al. Simulation of the expected performance of INSERT: A new multimodality
SPECT/MRI system for preclinical and clinical imaging. Nucl Instrum Methods Phys Res A.
2014;734:141–6.
Influence of Animal Handling
and Housing on Multimodality Imaging 3
David Stout

3.1 Introduction: History and Hardware

Medical imaging began first in humans, most famously with Roentgen’s X-ray
images of his wife’s hand and wedding ring [1]. Since then, many types of imaging
modalities have been discovered and developed, usually first in humans or cells. The
application of imaging methods to preclinical animal research often came later, in
part because the immediate application of the technology under development was
aimed at improving human health [2, 3]. The more challenging reason for the lag in
preclinical use has been that animals used in research are almost always smaller
than humans; thus the resolution and sensitivity of these imaging systems need to be
correspondingly greater. For example, the difference between the size of a human
and a mouse is ~4000-fold, so to study the same concentration of an imaging agent
in mice as used in humans requires three orders of magnitude improvement for
preclinical versus clinical imaging. There is a trade-off between resolution and sen-
sitivity, and it becomes expensive and difficult to improve both. The manufacturing
tolerances required for accurate and precise high-resolution measurements and the
signal capturing capability required for high sensitivity and low noise are challeng-
ing. Data processing requirements have historically been limiting as well, where the
file sizes challenged computer memory and processing capabilities. Improvements
in technology and manufacturing have led to the development of imaging systems
dedicated specifically to preclinical work in mice and rats [4, 5]. There are now
many different commercially available imaging systems with excellent performance
and fast data processing capabilities to provide image data quickly and accurately.
With proper quality control [6], the equipment is usually very reliable and
reproducible.
After initial development of separate PET, SPECT, CT, MR, and optical biolumi-
nescence and fluorescence systems, these modalities began to be combined into

D. Stout (*)
D&D Design, Culver City, CA, USA

© Springer Nature Switzerland AG 2019 55

C. Kuntner-Hannes, Y. Haemisch (eds.), Image Fusion in Preclinical
Applications, https://doi.org/10.1007/978-3-030-02973-9_3
56 D. Stout

single gantry devices [7]. This was driven in part by the need for anatomical infor-
mation to help understand the location of metabolic signals and to improve the
accuracy of molecular imaging data (scatter and attenuation corrections). When
clinical PET/CT systems were developed, they quickly became the standard within
just a few years, and now only dual-modality PET/CT and SPECT/CT clinical sys-
tems are commercially available. It took longer for this trend to find its way into the
preclinical field. Similar convergence has occurred with optical and MR or CT sys-
tems [8]. Clinical PET/MR has been well established [9]; however, preclinical mul-
timodality MR systems are only slowly coming to market, with either inserts for
PET or SPECT devices into MR magnets [10] or using interchangeable imaging
chambers that dock to other separate modality gantries [11].
Initially it was possible to conduct multimodality imaging using a removable
imaging chamber to hold the animal while being imaged in separate imaging devices
[12]. This design was sometimes prone to movement and positioning errors and
required that one chamber work in two or more potentially incompatible gantries
from different companies. Consolidation of preclinical hardware companies has led
to more compatible gantry designs, since several vendors now offer a wide range of
imaging devices and a common animal handling platform has become essential.
Ever since animals were imaged multiple times and in different systems, there
has been frequent need for software image co-registration and image fusion. This
task is made more complicated because different systems have different hardware
and software designs, resulting in variable factors such as fields of view, resolution,
and voxel sizes. Often there are software setting choices when configuring the
image acquisition and processing that determine the final image parameters. When
using separate systems for a combined study, careful selection of the image protocol
and using the same setting each time becomes essential. With the development of
multimodality systems, the co-registration and fusion process has been taken into
account and is often easier or automated. These devices usually provide reproduc-
ible positioning and image fusion through hardware co-registration of an animal
located inside a chamber or fixture of some kind, usually with a computer-controlled
bed movement that enables accurate positioning.
Image fusion can be accomplished using known positioning of the animals or
through software co-registration, where known features or signals are aligned.
Multimodality gantries make hardware-based fusion easy, provided the imaging
parameters are optimized. Using multiple separate gantries, a chamber can still be
used in a reproducible manner to make fusion trivial [12]. Where things get tricky is
if there is a need to co-register data sets that might have animals in different orienta-
tions or attempting to line up multiple data sets to use a common region or analysis
tool. Hardware-based options do not require any knowledge of the image content;
however, software-based methods require the image data and some method to opti-
mize the match between data sets. This can be done using fiducial markers, though
the markers become a distraction in the final image and may require being cropped
out. Markers also have to provide a signal in all the modalities being used, which is
a challenge with nuclear medicine methods due to half-life loss of signal over time.
Another challenge is that functional metabolic data can be quite different than
3 Influence of Animal Handling and Housing on Multimodality Imaging 57

anatomical information; thus it can be quite challenging to match up PET, SPECT,

or optical data with MR or CT data. Many groups have worked on this challenging
process, with varying degrees of success [13]. There are also several mouse and rat
atlases that have been developed with varying degrees of warping and deformation
possible to help match up the data sets [14, 15]. Atlases have the advantage of being
able to create a uniform orientation of the animals and have predefined regions for
determining imaging signal properties. While useful, atlas data are often only suit-
able for use with normal or near normal anatomy and physiology.
The key to making image fusion work rests with a good reproducible positioning
system, thus a well-designed imaging chamber. In addition to positioning, the
chamber also helps with reproducible physiological support. Physics, electronics,
and the imaging systems are all known, inanimate objects that behave in predictable
ways; however what gets imaged within these systems is another story. Animals are
not always predicable or behave in the same way each time, despite our best efforts
to minimize experimental variables. Some parts of an experiment may have obvious
factors that can alter imaging results, such as animal temperature, anesthesia, posi-
tioning, injection method, and any chemical or surgical intervention [16]. Other
factors are much more subtle and easily overlooked, such as housing conditions
[17], diet, circadian effects, cage conditions, and changing frequency. These factors
can be divided into two types, acute conditions associated with the imaging experi-
ment that take place within a few hours and chronic conditions that exist in the
housing location.

3.2 Imaging Chambers: Early Development

The idea of needing a chamber to hold animals only became important when imag-
ing systems were developed for relatively small animals, such as mice and rats.
Previously, imaging of larger animals was done by aligning the animal by eye, per-
haps using laser alignment lights used in patient scanners. With small-animal sys-
tems, there were often few if any tools or options for animal placement. From the
beginning of their development, microPET systems had a motorized computer-con-
trolled bed that would move animals in and out of the field of view along the axial
direction. The bed was permanently centered in the left/right orientation. Height
control motion was possible using a screw or motor that enabled different vertical
positions for accommodating rats or mice. As seen in Fig. 3.1, positioning of the
mouse or rat in early generation PET scanners was by eye, with fixation using paper
tape on cardboard, and there was no temperature control (Fig. 3.1). Image fusion or
co-registration mainly depended on either good manual placement or software reg-
istration and perhaps image warping.
The metabolic information in PET scans was sometimes difficult to interpret,
especially with very specific PET signals where little if any anatomical informa-
tion was present. It became immediately obvious that there was a need to pair up
PET with an anatomical imaging modality, and the most readily available and sim-
plest to integrate was X-ray computed tomography, CT. Fortunately, shortly after
58 D. Stout

Fig. 3.1 PET imaging of rats and mice prior to development of imaging chambers. Positioning
was done by eye and laser light, with no temperature control or ability to transfer animals to other
imaging systems

the development of microPET, small-animal CT systems became available. Since

microCT and microPET imaging systems were developed independently, they
were initially separate devices, which created a need for a common platform to
hold the animals that could facilitate image co-registration. Figure 3.2 shows a
chamber designed specifically for mice and rats for use in small-animal PET and
CT systems, which provided heating, positioning, gas anesthesia, and pathogen
control [12].
One might be tempted to think that the rodent chamber development was aimed
at imaging animals in two imaging systems, creating multimodality capability, but
that was actually a secondary benefit. The initial reason chambers were developed
was to create a controlled gas anesthesia environment and pathogen barrier for the
immune-compromised mice that were quickly becoming the most commonly
imaged animals, due to their useful ability to grow human tumor cell lines used in
oncology research. These animals needed to be protected from the environment and
us, since they were missing part or nearly all of their immune system. Preserving the
health of the animals through pathogen control, stable anesthesia, and reproducible
positioning is essential to maintain the validity of the animal disease model and
relevancy of the results to translate into clinical value and also helps with data anal-
ysis and image presentation.
Chamber development started with the goals set forth in Table 3.1 and resulted in
the chamber shown in Fig. 3.2. This chamber enabled the animal to be positioned
within about 1 mm reproducibly, with temperature control provided by a resistive
wire heating element [12]. The 30 μm heating wire was so small that there were no
artifacts in the CT image. By docking the chamber to a mounting plate fixed to the
imaging system gantry platform, the chamber could be reproducibly positioned
within the limits of the imaging system resolution (~100 μm for CT).
3 Influence of Animal Handling and Housing on Multimodality Imaging 59

a b

d c

Fig. 3.2 First-generation imaging chamber (a), designed to fit in small-animal PET (b) and CT (c)
gantries. Using a fixed known bed position enabled a simple software co-registration alignment to
create fused PET/CT images (d)

Table 3.1 Design elements • Pathogen barrier (isolation chamber)

required for multimodality • Gas anesthesia support
imaging chambers
• Temperature monitoring and control
• Materials compatible with multiple modalities
• Fixed reproducible positioning in multiple gantries
• Reproducible animal positioning for longitudinal studies
• Dynamic injection capability
• Allow for animal monitoring (visual and sensors)
• Easy to use

The combination of chamber and animal reproducibility coupled with position-

ing the chamber in the exact same location for each scan enabled a fixed co-regis-
tration alignment to be applied to co-register the image data sets (Fig. 3.2d). This
was also made possible by using very specific voxel and image matrix size dimen-
sions for PET (0.4 × 0.4 × 0.8 mm3 voxel size, 128 × 128 matrix size) and CT
(0.2 × 0.2 × 0.4 mm3 voxel size, 256 × 256 matrix size). These image acquisition
and reconstruction parameters meant that the PET and CT images could be overlaid
with matching fields of view without the need for resolution matching and software
interpolation. Only a fixed offset value was needed to consistently match the image
volumes, which was hardware dependent and easy to measure and did not rely on
60 D. Stout

the image content for co-registration. The creation of 3D co-registered fused images
became a trivial part of the image assembly, which was straightforward to automate
for every set of images. Prior to this hardware/software co-registration, separate
image data sets had to be co-registered using software techniques that might be
time-consuming and labor intensive. It may also have been necessary to use fiducial
markers that could be used as guides to register the data sets. Fiducial markers are
not always reproducible in terms of their position if placed on animals. Marker abil-
ity to give a consistent signal is challenging due to half-life decay, and they may
need to be removed from images for analysis and display.
The chambers shown in Fig. 3.2 were tested and refined over many years and
100,000+ animal studies. They proved to be extremely reliable, providing auto-
mated co-registered PET-CT images, ready for analysis without requiring any soft-
ware co-registration. The fused data sets were automatically scaled to give a
reasonable contrast level for both the CT and PET images. There were some prob-
lems however, where animals could move while being transported between imaging
systems and if the chamber was not properly secured to the mounting plate or the
electrical and anesthesia connections were not manually connected. These short-
comings led to the development of the next generation of chambers.

3.3 Imaging Chambers: Second-Generation Development

The chamber design and development discussed above was all done by eye and
hand, using skilled machinists. This approach was very time-consuming and expen-
sive, often taking 6 months or more to get a new design completed. With the advent
of computer-aided design tools (CAD), the next generation of chambers was created
using computers. The goals remained the same as before in Table 3.1; however cer-
tain improvements were needed. The original chamber required that the heating
wire and anesthesia supply and exhaust lines be connected by hand. Sometimes this
was overlooked, leading to the animals waking up and the data lost or unnecessary
exposure to personnel to the anesthetic gas. The chambers also required manual
connection to the mounting plate on the scanner bed. If not mounted correctly, the
chamber could be tilted at an angle, requiring manual image co-registration. There
was also a problem with animals slipping out of position, most frequently with the
nose dropping out of the anesthesia delivery cone. Having the head in two different
positions for the two imaging modalities was not something that could be corrected
by realignment of the images. A better solution was needed, and it turns out that
experience was the best guide with respect to ideal animal placement.
The second generation of imaging chambers uses plug-and-play connection
technology (Fig. 3.3). A series of chamber designs were tested, resulting in the final
version on the bottom of the left side image. The chamber can be slid into a docking
cradle, with the heating and anesthesia connections made automatically. This
ensures better positioning and consistent connections, in part since there is both a
feel of the chamber clicking into place and a light turns on when the chamber is
docked correctly. The chamber is actually just one part of a complete system, which
3 Influence of Animal Handling and Housing on Multimodality Imaging 61

Fig. 3.3 Second-generation (2009–2015) imaging chamber design evolution (left). Right image
shows prototype anesthesia vaporizer, induction boxes, and an early version of a docking station
and imaging chamber

Fig. 3.4 Second-generation chamber reproducibility testing. Each panel has five overlaid CT
images, where the chamber and rat were removed and replaced between each scan. Chamber
reproducibility (left image) was <0.21 mm using 0.5 mm voxels. Rat body part positioning, left to
right, of sternum, sagittal head view, coronal skull view, and left arm were within 0.28 mm

also includes the anesthesia induction box, docking station for working with the
animals in the chamber within a biosafety cabinet and the gantry dock located inside
the imaging system. The right side of Fig. 3.3 shows an early design for an anesthe-
sia vaporizer, induction boxes, and docking station with a prototype chamber.
The other improvement to animal positioning and image co-registration was
made by removing the nose cone and using a ∩-shaped cowling. The animal can be
laid down on the chamber with its head in a natural position. Reproducible trans-
axial positioning is provided by the curved bed surface, and axial positioning is
provided by where the animal forehead touches the cowling (Fig. 3.4). This design
turned out to have much better reproducible positioning than the earlier design and
far fewer movement problems when moving between imaging systems and imaging
positions within a single gantry. An examination of 100 consecutive studies by mul-
tiple different investigators showed an amazingly reproducible positioning capabil-
ity, without using any tape or restraint devices. All that was needed was the right
curvature and place for the nose to touch, plus a little training to have the mouse
posture consistently positioned. This improved design resulted in fewer misaligned
data sets, improving the image fusion results.
62 D. Stout

These improvements in chamber technology have been incorporated by nearly

every small-animal imaging manufacturer into all in vivo imaging systems. Bruker,
Somni, Perkin-Elmer, Sofie, Cubresa, and Mediso now include some kind of animal
anesthesia, support, and positioning solution that utilizes some or all of these cham-
ber principles. Chamber designs are matched to the physics and requirements of
each imaging modality. For example, a resistive wire heating strip similar to a car
window defrost wire is compatible and used for PET, SPECT, and CT systems,
whereas this would interfere with MR signals, so warm air or water is used for those
systems. Optical imaging often uses a simple box with anesthesia nose ports that is
placed in a heated enclosure (e.g., IVIS systems); however other optical systems
such as fluorescence tomography (FMT) require the ability to see through the ani-
mal, so the chambers have the mouse sandwiched between two panels.
Imaging chambers may also include the ability to monitor animal physiology for
monitoring heart rate, respiration, and temperature. Gating for cardiac and respira-
tion allows different parts of these cycles to be separated out for investigation or to
remove motion artifacts. The ability to keep animals warm for proper physiology is
crucial, as is the need to monitor and maintain an ideal depth of anesthesia. The
ability to monitor may be built into the chamber systems from the vendor, such as
Bruker, or may be a third-party monitoring systems such as the MouseOx system
from Starr Life Sciences.

3.4 Imaging Chambers: Future Development

The ability to design parts on a computer coupled with automated machinery to cre-
ate these parts brought down the cost and time needed to make chambers. The next
technology to impact chamber development has been 3D printing, which is ideally
suited to preclinical applications. Most 3D printers use plastic-like materials, which
are melted and built up layer by layer to create any shape desired. These materials
are low density and well suited to PET, CT, SPECT, and MR imaging methods.
They can even be used to create custom beds or disposable parts for individual
experiments, since the cost to purchase and use these printers is very low and free
software to design and print is readily available online. Figure 3.5 shows examples
of both a computer-designed and machine-milled imaging chamber assembly (CAD
part, left side) and a 3D printed PET/MR bed system for mice. The development and
manufacturing time has now shrunk to a matter of days or even hours to draw, print,
and use a new design.
Three-dimensional printed parts can be useful for prototyping and customized
parts designed for specific studies or even specific animals. For long-term routine
use however, these parts can in some cases become brittle and difficult to clean. The
infill, layer thickness, and other settings can influence the strength, ability to clean,
and density of the chamber parts, all of which may or may not matter with respect
to the imaging modality being used. At present, it is not possible to 3D print a clear
part, so this method is not yet useful for coverings if it is necessary to see the ani-
mals, either for positioning or monitoring purposes.
3 Influence of Animal Handling and Housing on Multimodality Imaging 63

Fig. 3.5 Left: computer-aided designed (CAD) third-generation PET/CT imaging chamber (left
image) made using traditional manufacturing techniques. Right: 3D printed mouse PET/MR
chamber. PET/CT chamber uses resistive wire heating, and PET/MR system uses warm air

3.5 Image Fusion and Analysis

The use of a chamber enables automated creation of co-registered fusion images

that helps tremendously with image analysis. Regions of interest can be determined
using any modality and applied to all images. This makes quantitative measure-
ments easier to derive from any image source, whether different modalities or dif-
ferent imaging sessions. One advantage to the reproducible positioning is that often
animals are imaged multiple times as part of an experiment, to look at treatment
effects or progression of a disease model. With accurate positioning, the images
acquired from multiple sessions can be uniformly acquired, enabling the use of
predefined regions to be applied to all the image data. The positioning capability of
these chambers is far better than the nuclear medicine-based image resolutions, so
the co-registration is extremely good. It is also aesthetically desirable to have the
animals displayed in the same orientation when creating figures with multiple time
point images for publication and presentation.
For image analysis of animals pre- and posttreatment, the regions drawn may
need to grow with a changing signal, such as with tumor growth, or they may need
to stay the same size when the object is not changing in size over time and treat-
ment. An example of using the same region size would be evaluation of dopami-
nergic function in a Parkinson’s disease model, where the dopamine producing
neurons may be treated to reduce the imaging agent uptake to mimic the disease.
The reduction of image signal after treatment will appear to be a shrinking of the
area, when in reality only a small number of cells will be lost. In this case, the
baseline region size and location would be most appropriate to use for a posttreat-
ment evaluation of the signal measurement. Co-registration of the pre and post
images thus becomes essential. This can be accomplished or at least greatly aided
by using good reproducible positioning hardware, or it may also be necessary to
use software for a fixed three-axis alignment (x, y, z), or it may be necessary to use
rotational alignment that requires a six-axis alignment or perhaps image warping
and interpolation. The more complicated the alignment requirements and
64 D. Stout

processing, the more labor and resource intensive and possibly error prone the
process will become.

3.6 Handling and Housing Factors

The use of chambers helps position animals reproducibly, enables gas anesthesia,
provides heating and perhaps pathogen control, and allows for automated image
co-registration and fusion. These are all useful benefits for the imaging process;
however, there are other factors that also can significantly alter the physiology and
thus metabolic signals being measured using molecular and anatomical imaging
techniques. These factors can be broadly divided into two classes, acute factors
related to the imaging session and chronic conditions related to the housing
conditions.

3.6.1 Acute Imaging Factors

From the time animals leave the vivarium until their return after imaging work is
completed, the imaging process ideally follows standard operating protocols
(SOPs). Nearly all imaging-related work follows the same steps each time, from
anesthesia and injections to imaging and recovery. Procedures are followed to
ensure consistent conditions and data reproducibility. In the United States, Europe,
and many other countries, laws regulating animal research require SOPs that cover
all the various procedures and conditions required for any research, including all
noninvasive imaging work. These include thermal warming of animals, injection
routes and methods, blood and tissue sampling, anesthetics and both biosafety and
radiation controls. Monitoring the animal physiology is usually required and in
many cases essential to ensure reproducible experimental procedures. Monitoring
can be accomplished by visual inspection, remote-sensing devices such as EKG or
rectal thermocouple, or remote visual sensing of heart or respiratory rates using a
camera.
Temperature control for animals, especially mice with small body mass, is an
essential part of keeping the animals healthy and also physiologically stable. Mice
quickly adapt the temperature of the surface they are on once anesthetized, so keep-
ing them in a heated induction box, imaging chamber, and recovery area is very
important. Mice thermoregulate their body temperature by activating brown adipose
tissue (BAT), which can consume up to 50% of all energy used within the body [18].
Preheating of animals for 20–30 min can dramatically reduce the BAT signal com-
monly seen with energy use imaging agents such as FDG (Fig. 3.6). Mice also
control body temperature through controlling blood flow in the tail, which is a com-
mon injection site for imaging agents. Warm animals will have more blood flow in
the tail, so it will be easier to find the vein, inject, and have the imaging agent deliv-
ered to the bloodstream in warm animals with greater blood flow. Because mice
rapidly equilibrate to the environment when under anesthesia, the surface the
3 Influence of Animal Handling and Housing on Multimodality Imaging 65

Fig. 3.6 PET/CT images of FDG in a mouse without warming (left and center images), showing
high brown fat (BAT) uptake in the neck and shoulder regions. Right image of a different mouse
was acquired with warming, shutting off BAT uptake, allowing visualization of neck tumor, and
draining lymph nodes

animal rests upon can be monitored for temperature rather than requiring the use of
a more invasive rectal thermocouple. This approach is both safer for the animal and
for pathogen control.
There are many other factors that can alter metabolism, including time of day
[19], sex of the person handling the animals, fasting, stress, noise, smells, presence
of other animals, and other factors beyond the scope of this chapter. There are also
factors related to the imaging agent, such as the volume injected, specific activity,
pH, presence of alcohol to help dissolve nonpolar molecules, and injection route.
These can be trivial to major in terms of their effect, depending on the nature of the
experiment. Acclimating the animals and keeping conditions consistent, along with
investigating the ideal conditions for your disease model, is an essential part of the
experimental design. Reporting these conditions as part of any publication is imper-
ative as well, so that the findings can be replicated [20]. The DICOM committee has
expanded the information that can be captured in image header information in an
effort to provide a method to record and track this essential information.
66 D. Stout

The handling conditions, environmental factors, imaging system settings, and hous-
ing conditions are all important factors that can influence experimental results.
There may be specific parameters necessary to monitor as well, depending on the
experimental design. These could include diet, blood sample measurements, metab-
olite analysis, or behavior. Monitoring may be required over the course of the imag-
ing session or may expend to the length of the entire experiment. For example,
longer-lived isotopes may be injected once and followed over days or weeks and
may require metabolite analysis to determine the signal information being mea-
sured. Good experimental design and protocol adherence becomes key to a success-
ful project.

3.6.2 Chronic Imaging-Related Factors

Animals used for experiments spend nearly all their time living in cages within a
vivarium, where the veterinary staff typically determines the housing conditions.
Vivarium conditions include caging type, room temperature and humidity, lighting,
feed, water, cage changing, maximum number of animals per cage, bedding mate-
rial and amount, and so forth. The choices made can have a huge impact on animal
physiology and are known to alter metabolism, histology results, and anatomy [18].
Choices are often made based on cost and the ability to conduct daily heath checks;
however, these conditions may make the animal models invalid or variable.
Investigators using imaging systems do not normally have control over or pay much
attention to the housing conditions, since those are often set by the institution and
perhaps thought of as beyond their control or not a source of variability. The varia-
tions between vivarium conditions, either within or between institutions, may be
one of the biggest causes of irreproducibility of preclinical research results. Even
within a single institution, there may be different caging types in use at any one
time; thus results may not be replicable simply due to housing location.
When designing an experiment, it is important to consider the housing condi-
tions and their potential impact on the research results. At a minimum, these condi-
tions need to be reported as part of the methods in any publication. In 2015, the
National Electrical Manufacturers Association (NEMA) revised the Digital Imaging
and Communications in Medicine (DICOM) standards that define all preclinical
imaging file formats to expand the header fields to include animal housing and han-
dling conditions, in recognition that these factors play a critical role in experimental
results [21].

3.7 Summary

The independent development of small-animal imaging systems, coupled with the

need to image both metabolism and anatomy while maintaining pathogen control,
anesthesia, heating, and reproducibly positioning led to the development of imaging
chambers. These chambers enhanced the ability to co-register data sets, use
3 Influence of Animal Handling and Housing on Multimodality Imaging 67

standardized regions of interest, and improved the data analysis process. They
enabled hardware-based image co-registration and fusion, which is faster and easier
when creating fused images. Succeeding generations of chambers have been inte-
grated into the imaging system design, making it easier to work with animals out-
side the imaging gantry and simpler to plug the chambers into the gantry with
automated connections for anesthesia and heating. The idea of caring for the animal
and making physiological conditions suitable and reproducible has now become
part of the imaging system design and equipment.
We have also noted how both acute and chronic animal handling and care condi-
tions can drastically alter metabolic imaging data. Animals are used to conduct
experiments that would otherwise be unfeasible or impossible in humans, with the
implied understanding that the metabolic states of these animals are surrogate rep-
resentations of human conditions. Animal models of disease states are well known
and validated; however, the exact conditions of the validation studies are often not
reported in enough detail, nor are they likely to match the conditions in your own
facilities. This in part is due to the progressive discovery of new factors that turned
out to be important and to the development of new equipment and procedures for
housing and care of animals.
Image physics and detailed equipment knowledge have generally been associ-
ated with physicists, while animal housing and care procedures are the domain of
the veterinary care providers. A researcher using animals to investigate a medical
question may not have much shared knowledge with either of these two groups;
however, it is the combination of all three of these fields that is essential to acquiring
and analyzing meaningful data from living animals. The use of a particular cage
type or room temperature decided by the veterinary staff may have a huge unfore-
seen impact on metabolic data and can lead to irreproducible study results. Likewise,
equipment and software changes can alter both qualitative and quantitative imaging
results. It is imperative for people working in preclinical research to validate and
ensure that the animals are accurately representing an appropriate metabolic state
for research work, just as it is to ensure appropriate quality control measurements
are consistently acquired to ensure system performance.

References
1. Glasser O. Wilhelm Röntgen and the early history of the Roentgen rays. San Francisco, CA:
Norman Publishing; 1993. p. 25, ISBN: 0930405226.
2. Jaszczak RJ. The early years of single photon emission computed tomography (SPECT):
an anthology of selected reminiscences. Phys Med Biol. 2006;51:R99–115. https://doi.
org/10.1088/0031-9155/51/13/R07.
3. Portnow LH, Vaillancourt DE, Okun MS. The history of cerebral PET scanning. Neurology.
2013;80:952. https://doi.org/10.1212/WNL.0b013e318285c135.
4. Cherry SR, Gambhir SS. Use of positron emission tomography in animal research. ILAR J.
2001;42(3):219–32. https://doi.org/10.1093/ilar.42.3.219.
5. Zinn KR, Chaudhuri TR, Szafran AA, O’Quinn D, Weaver C, Dugger K, Lamar D, Kesterson
RA, Wang X, Frank SJ. Noninvasive bioluminescence imaging in small animals. ILAR J.
2008;49(1):103–15. https://doi.org/10.1093/ilar.49.1.103.
68 D. Stout

6. Osborne DR, Kuntner C, Berr S, Stout DB. Guidance of efficient small animal imaging quality
control. Mol Imaging Biol. 2017;19:485–98. https://doi.org/10.1007/s11307-016-1012-3.
7. Magota K, Kudo N, Kuge Y, Nishijima K, Zhao S, Tamaki N. Performance characterization of
the Inveon preclinical small-animal PET/SPECT/CT system for multimodality imaging. Eur J
Nucl Med Mol Imaging. 2011;38:742–52. https://doi.org/10.1007/s00259-010-1683-y.
8. Chehade M, Srivastava AK, Bulte JWM. Co-registration of bioluminescence tomography,
computed tomography, and magnetic resonance imaging for multimodal in vivo stem cell
tracking. Tomography. 2016;2:158–65. https://doi.org/10.18383/j.tom.2016.00160.
9. Torigian DA, Zaidi H, Kwee TC, Saboury B, Udupa JK, Cho ZHC, Alavi A. PET/MR imaging:
technical aspects and potential clinical applications. Radiology. 2013;267:26–44. https://doi.
org/10.1148/radiol.13121038.
10. Judenhofer MS, Wehrl HF, Newport DF, et al. Simultaneous PET-MRI: a new approach for
functional and morphological imaging. Nat Med. 2008;14:459–65. https://doi.org/10.1038/
nm1700.
11. Sasser TA, Chapman SE, Sanders I, Liepert L, Leevy WM. Cross-platform MRI/PET or MRI/
SPECT imaging, and co-registration. 2016. Available at: https://www.bruker.com/filead-
min/user_upload/8-PDF-Docs/PreclinicalImaging/Brochures/T156737_App_Note_Cross-
Platform_edit_bur.pdf. Accessed 1 Nov 2016.
12. Suckow CE, Kuntner C, Chow PL, Silverman RW, Chatziioannou AF, Stout DB. Multimodality
rodent imaging chambers for use under barrier conditions with gas anesthesia. Mol Imaging
Biol. 2009;11:100–6. https://doi.org/10.1007/s11307-008-0165-0.
13. Oliveira FP, Tavares JM. Medical image registration: a review. Comput Methods Biomech
Biomed Eng. 2014;17:73–93. https://doi.org/10.1080/10255842.2012.670855.
14. Wang H, Stout D, Chatziioannou A. A deformable atlas of the laboratory mouse. Mol Imaging
Biol. 2015;17:18–28. https://doi.org/10.1007/s11307-014-0767-7.
15. Segars WP, Tsui BMW, Frey EC, Johnson GA, Berr SS. Development of a 4D digital mouse
phantom for molecular imaging research. Mol Imaging Biol. 2004;6:149–59. https://doi.
org/10.1016/j.mibio.2004.03.002.
16. Fueger BJ, Czernin J, Hildebrandt I, Tran C, Halpern BS, Stout DB, Phelps ME, Weber
WA. Impact of animal handling on the results of FDG-PET studies in mice. J Nucl Med.
2006;47(6):999–1006.
17. David J, Knowles S, Lamkin D, Stout D. Individually ventilated cages impose cold stress on
laboratory mice: a source of systemic experimental variability. J Am Assoc Lab Anim Sci.
2013;52:738–44.
18. David J, Chatziioannou A, Taschereau R, Wang H, Stout D. The hidden cost of housing prac-
tices: quantifying the metabolic demands of chronic cold-stress of laboratory mice with non-
invasive imaging. Comp Med. 2013;63(5):386–91.
19. Jilge B, Kunz E. The laboratory mouse, Chapter 20. London: Elsevier Academic Press; 2004.
https://doi.org/10.1016/B978-012336425-8/50073-X.
20. Stout D, Berr S, LeBlanc A, Kalen J, Osborne D, Price J, Schiffer W, Kuntner C, Wall
J. Guidance for methods descriptions used in preclinical imaging papers. Mol Imaging Biol.
2013;12:1–15.
21. NEMA DICOM WG30 Supplemental 187. https://medical.nema.org/medical/dicom/final/
sup187_ft_preclinicalanimalacquistitioncontext.pdf.
Multimodal Optoacoustic Imaging
4
Murad Omar, Dominik Soliman, and Vasilis Ntziachristos

4.1 Introduction

Imaging is an extremely important tool in modern biomedical research. As the old

saying states, “A picture is worth a thousand words.” As in many aspects of life,
observing events while they occur makes it easier to understand them, thus imaging
was and will always be on the forefront of biomedical sciences. By means of imag-
ing, we can diagnose and distinguish between healthy and malignant tissue or make
and test hypothesis related to certain biological questions, such as if a certain drug
will be able to work as expected or if it doesn’t even target the intended organ in the
first place.
In the imaging field, four parameters play a major and competing role when
selecting the technology of choice: spatial resolution, temporal resolution, sensitiv-
ity, and penetration depth (see Fig. 4.1 for a graphical representation). Although
imaging techniques have developed tremendously over the last decades, there is still
not a single super-technique capable of performing well in all of these parameters.
Generally speaking, fully optimizing one of them comes at the cost of the other
parameters.
More particularly, either one of these parameters needs to be optimized for a
specific top-performance application, or a balance between several parameters
needs to be established to cover a broader range of applications. In the last two
decades, a technique called optoacoustic (photoacoustic) imaging has been intro-
duced, which is coming close to pushing the boundary of the possible along all

M. Omar · D. Soliman · V. Ntziachristos (*)

Chair of Biological Imaging, Technische Universität München, Munich, Germany
Institute of Biological and Medical Imaging, Helmholtz Zentrum München,
Neuherberg, Germany
e-mail: [email protected]

© Springer Nature Switzerland AG 2019 69

C. Kuntner-Hannes, Y. Haemisch (eds.), Image Fusion in Preclinical
Applications, https://doi.org/10.1007/978-3-030-02973-9_4
70 M. Omar et al.

Fig. 4.1 Imaging pyramid Imaging speed

showing the different
parameters that can be
optimized when designing
an imaging system.
Generally, if one parameter
is maximally optimized,
compromises in the other Spatial
dimensions need to be resolution
made
Penetration
depth

Sensitivity

four dimensions in the parameters space. This technique is derived from optical
imaging. Thus, we will first discuss the case for optical imaging in the biomedical
sciences.

4.1.1 Optical Imaging

Optical imaging is a very powerful and widely applied technique. It has been used
for several centuries to gain new biological insights, to study and understand dis-
ease, and even to diagnose patients. The strength of optical imaging stems from two
facts. Firstly, we as human beings tend to better understand what we see, and, thus,
looking at an optical image facilitates the relation of the observed to a certain func-
tion or disease. Secondly, biological tissue commonly changes its color based on the
underlying organ or the disease of interest. Therefore, it is only natural to expect
that such a contrast will deliver valuable information about biological processes,
such as disease developments, function, and metabolism, and generate images of
the bio-distribution of a certain molecular agent.
Optical imaging methods can be divided into two main categories: microscopic
and tomographic modalities. In the microscopic case, usually a focused beam of
light is used to collect information from the sample of interest. This can be achieved
either point by point or line by line, depending on the specific configuration. Such a
method, as can be already understood from its name, delivers images with spatial
resolutions high enough to observe single cell or even subcellular features. On the
other hand, these techniques can only collect information from within a thin layer
(around the optical focus) of the examined sample and can thus only image a small
volume within a reasonable amount of time. The major limiting factor in the case of
penetration depth is the strong scattering of visible light that is experienced in bio-
logical tissue [1]. Because of this optical scattering, it becomes impossible to focus
a beam of light within a biological sample beyond a few hundred micrometers of
depth. Consequently, to use a microscopic method for preclinical experiments, it is
4 Multimodal Optoacoustic Imaging 71

necessary to (1) image only the superficial layers in an animal (e.g., the skin); (2)
replace the part of the animal between the organ of interest and the microscope with
a window chamber, which is widely used in neuroimaging or cancer research; (3)
use cell cultures if possible, which can also deliver multiple insights in many pro-
cesses and models; or (4) clear the organ of interest, which, however, limits the
applicability to ex vivo studies.
The other category of optical imaging techniques is so-called tomographic meth-
ods. Examples of such techniques include diffuse optical tomography (DOT) and
fluorescence molecular tomography (FMT). In the tomographic case, instead of
using a focused beam, a broad illumination is employed. In order to obtain a recon-
struction of the underlying structure or function of the examined sample, light prop-
agation is modeled using the optical diffusion or the radiative transfer equation [2].
Based on these models, an inverse optimization problem is solved to reconstruct the
final images. In other words, mathematical techniques are used for image recon-
struction. As the models are based on optical diffusion, the imaging depth in tissue
is no longer limited to the first few hundreds of micrometers, rather it becomes pos-
sible to image several millimeters or even centimeters of depth in tissue. Additionally,
optical tomography allows to image the function of a living tissue by using an exci-
tation at appropriate wavelengths and thus to deduce information about the hemo-
dynamics and metabolism [3]. Finally, it is possible also to inject molecular
fluorescence agents, which, instead of merely visualizing structure or function,
enable the imaging of the bio-distribution of such an agent, which provides high
specificity at centimeters of depth [2]. Nonetheless, the cost to pay for this extended
imaging depth is spatial resolution. For example, in the case of FMT, a spatial reso-
lution of 1–2 mm is achieved in the best-case scenario [2].

4.1.2  lexander Graham Bell and the Discovery

A
of the Optoacoustic (Photoacoustic) Effect

Alexander Graham Bell, who is more famous for his invention of the phone, has
discovered the optoacoustic (photoacoustic) effect in 1883. The optoacoustic effect
is based on the conservation of energy and momentum, where optical energy in
particular and electromagnetic energy in general are absorbed by certain molecules
that absorb light at specific wavelengths (chromophores) in the specimen.
Consequently, this absorption results in an instantaneous raise in the local tempera-
ture around the chromophores, leading to a thermoelastic expansion and a genera-
tion of acoustic waves. Thus, energy is transformed from electromagnetic waves to
heat and finally to sound (see Fig. 4.2 for a graphical illustration of this process).
Bell noticed that acoustic waves are generated when a thin membrane is shined
on with modulated light. To capitalize on this newly discovered effect, he suggested
building a wireless link between two persons: one is listening to the generated
acoustic waves, while the other one is modulating the shining light. This was effec-
tively the first wireless communication link in history, called photophone. Although
wireless communication was a brilliant idea, it was not practical as it only worked
72 M. Omar et al.

Sample Acoustic wave

Laser Optoacoustic
source signal

Ultrasound
Optical absorber detector

Fig. 4.2 Optoacoustic imaging principle. The sample is illuminated by a pulsed laser source.
Chromophores in the sample absorb optical energy, which is converted into heat via non-radiative
processes. This transient temperature rise causes a local pressure build-up, which propagates
through the sample as an acoustic wave and which is finally measured by an ultrasound detector
(e.g., transducer)

at short distances and needed a line of sight between the sender and receiver. After
the invention of the photophone, the optoacoustic effect was exploited in the 1930s
to characterize and spectrally measure certain gases. In fact, this has been the major
use of the optoacoustic effect until the early 1980s, when Theodore Bowen pro-
posed the applicability of this effect for biological imaging [4]. Finally, optoacous-
tic imaging started to pick up traction in the 1990s with the introduction of
sufficiently strong laser sources and sensitive ultrasonic detectors.

4.1.3 The Case for Optoacoustic Imaging: Listening to Light

As we have seen in the previous discussion, optical imaging yields strong and rich
contrast, but the spatial resolution starts degrading dramatically beyond the first few
hundred micrometers of depth. To overcome this limitation, optoacoustic imaging
has been introduced. In optoacoustics, instead of relying on an optical focus to get
the desired information from a pixel or voxel, the information about the location is
derived from the acoustic propagation time. In such a case, ultrasonic waves, which
are 2–3 orders of magnitude less susceptible to scattering than light, are used to
generate an image. Because of this usage of ultrasonic waves, it becomes possible
to image several centimeters into tissue with an adequate spatial resolution in the
range of 100 μm, which is at least an order of magnitude better than what could be
achieved with pure optical imaging techniques deep inside the tissue. Moreover,
because ultrasonic detectors can be responsive to all kinds of frequency ranges
depending on the material and fabrication, it is possible to trade spatial resolution
with imaging depth. More specifically, it is possible to partially sacrifice the pene-
tration depth and to get images with a resolution of 20 μm in return, which again is
much better than what pure optical techniques can deliver from deeper layers of
tissue.
Essentially, optoacoustics delivers high optical contrast side by side with ultra-
sonic resolution. Consequently, it takes the best features of the two worlds and com-
bines them into a single imaging modality. Additionally, because it facilitates the
4 Multimodal Optoacoustic Imaging 73

parallelization of acoustic detection, it is possible to acquire optoacoustic images at

video rates or higher. Thus, the optical imaging capabilities have been pushed by the
optoacoustic modality along three of the dimensions of the parameter space: pene-
tration depth, spatial resolution, and imaging speed.
In the next sections, we will introduce the basic principles of optoacoustic imag-
ing, where we derive the governing equations and describe the limitations on spatial
resolution and imaging depth. Subsequently, we dive into the different scales achiev-
able with optoacoustic imaging; those are the macroscopic, mesoscopic, and micro-
scopic scales. Additionally, we describe the different possibilities for multimodal
imaging in the macroscopic, mesoscopic, and microscopic cases based on selected
example applications. Finally, we conclude the chapter with a summary and an
outlook.

4.2 Basic Principles

4.2.1 Initial Pressure Generation

The origin of optoacoustic signals is the absorption of laser light by tissue chromo-
phores, followed by the transient heating of a local tissue region that leads to a
fractional volume expansion, which can be expressed as [5]

dV  
= -k p ( r ) + b T ( r ) . (4.1)
V
 
Here, p ( r ) denotes the initial pressure change (Pa), T ( r ) is the temperature
change (K, typically below 0.1 K), κ represents the isothermal compressibility
(~5 × 10−10 Pa−1), and β denotes the isobaric thermal expansion coefficient
(~4 × 10−4 K−1) [5]. For efficient optoacoustic pressure generation, the laser pulses
have to be sufficiently short. More precisely, the pulse width τ has to be shorter than
both the thermal relaxation time tth = lh2 / a th and the stress relaxation time tst = lh/c,
which represent the time that it takes the heat and built-up stress, respectively, to
propagate out of the heated volume. The length of the heated volume is represented
by lh (m), whereas αth denotes the thermal diffusivity (m2/s), and c is the speed of
sound (~1500 m/s in tissue). Usually, tst is much smaller than tth (e.g., for an absorber
diameter of 20 μm, tth ≈ 2.6 ms and tst ≈ 13 ns).
Under thermal and stress confinement, the fractional volume change can be
neglected, and the initial optoacoustic pressure rise can be written as
  
p0 ( r ) = Gh h ma ( r ) f ( r ) , (4.2)

where Γ represents the dimensionless and tissue-dependent Grüneisen parameter

Γ = β/κ ρCV, ρ is the mass density (kg/m3), CV denotes the specific heat capacity (J/

(kg K)), ηh is the heat conversion efficiency, ma ( r ) denotes the optical absorption

coefficient (1/m), and f ( r ) represents the optical fluence (J/cm2). Because
74 M. Omar et al.

optoacoustic signal generation originates from non-radiative electronic relaxation,


ηh and thus p0 ( r ) are increased for chromophores with a low fluorescence quantum
yield (i.e., the ratio of the number of emitted to absorbed photons).

4.2.2 Optoacoustic Wave Equation and Forward Solution

The propagation of the generated pressure as a bipolar acoustic wave under the
condition of thermal confinement and in an acoustically homogeneous medium is
described by the optoacoustic wave equation [5–7]:

æ 2 1 ¶2 ö  G ¶ 
ç Ñ - 2 2 ÷ p ( r ,t ) = - 2 H ( r , t ) . (4.3)
è c ¶t ø c ¶t

Here, H ( r ,t ) denotes the heating function (J/(cm3 s)), which is defined as the
thermal energy deposited in the tissue through optical absorption per unit time and
unit volume [5]. It is defined as
  
H ( r ,t ) = h h m a ( r ) F ( r ,t ) , (4.4)
  
where F ( r ,t ) is the optical fluence rate F ( r ,t ) = ¶f ( r ,t ) / ¶t (J/(cm2 s)), which
corresponds to an average value of the optical intensity in a turbid diffusive medium
[8]. Due to the time derivative in (4.3), only time-variant heating can result in the
generation of optoacoustic signals, i.e., pulsed or intensity-modulated laser sources.

The forward solution p ( r ,t ) to the optoacoustic wave equation, i.e., the total

pressure measured at position r and time instant t, can be found by using the
Green’s function approach to be

 G ¶ é d 3 r¢  ù
p ( r ,t ) = òòò ê   ¢ H ( r ¢ ,t ¢ ) ú . (4.5)
4p c ¶t V ê r - r
2
úû t¢=t - r -r¢ / c
ë

The heating function can be decomposed into a position- and time-dependent

term, whereas the latter can be approximated as a delta function in the case of suf-
 
ficiently short laser pulses: H ( r ,t ) = H r ( r ) × d ( t ) . Then, the forward solution sim-
plifies to

 ¶ ét  ù
pd t ( r ,t ) = ê òò d W p0 ( r ¢ ) ú . (4.6)
¶t ë c S û r -r¢ =ct

According to (4.6), the total pressure signal pd t ( r ,t ) is obtained by integrating
 
the contributions of all sources located at spherical shells with radius r - r ¢ cen-
tered at c ⋅ t (spherical Radon transform).
In the special case of a spherical absorber with the same material properties as
the surrounding medium, an analytical expression of the forward solution can be
obtained as
4 Multimodal Optoacoustic Imaging 75

  æd ö Dr - ct
pd t ,sph ( r ,t ) = p0 ( r ) Q ç - Dr - ct ÷ , (4.7)
è2 ø 2Dr
  
where Dr = r - ra , ra is the position of the absorber center and d is the absorber
diameter [6]. Θ(x) is the Heaviside function, which is defined as

ì0, x < 0
Q ( x) = í . (4.8)
î1, x ³ 0
The corresponding optoacoustic signals yield a characteristic bipolar N-shape, as
illustrated for absorber diameters of 10, 20, and 40 μm in Fig. 4.3a. Under stress
confinement, the amplitude and the duration of the signals are proportional to the
source diameter. For solid absorbers with a higher speed of sound and mass density,
the optoacoustic signals deviate from the perfect N-shape, yielding a reduced width
and several trailing amplitude oscillations [9, 10].
Finally, the solution for a finite laser pulse duration is calculated by convolving

pd t ( r ,t ) with the temporal pulse profile Ht(t):
¥
 
p ( r ,t ) = ò dt H t ( t - t ) pd t ( r ,t ) . (4.9)
-¥

If the absorber is sufficiently small compared to the resolution of the used ultra-
 
sound detector, it can be approximated as a point source (i.e., H r ( r ) = d ( r ) ), and
the forward solution further simplifies to

G ¶ æ rö
pd t ,d r ( r ,t ) = d ç t - ÷. (4.10)
4p c r ¶t è c ø
2

This is especially relevant for optoacoustic microscopy applications, which will

be discussed later.

a 40 µm b 1 40 µm
1
20 µm 20 µm
0.8
Amplitude (a.u.)

Amplitude (a.u.)

0.5 10 µm 10 µm
0.6
0
0.4
-0.5
0.2

-1
0
0 10 20 30 0 100 200 300
t (ns) f (MHz)

Fig. 4.3 Simulation of optoacoustic signals from 10, 20, and 40 μm absorbers based on (4.7). (a)
Time courses of the simulated optoacoustic signals yielding the characteristic N-shape. (b)
Corresponding frequency spectra
76 M. Omar et al.

4.2.3 Optoacoustic Frequencies

Optoacoustic signals are generally composed of a continuous set of frequencies in

the MHz range, the bandwidth of which being inversely proportional to the duration
of the time signals and thus to the size of the optical absorber. Figure 4.3b presents
the frequency spectra of the simulated N-shaped time signals from 10, 20, and
40 μm absorbers shown in Fig. 4.3a. As can be seen, the smaller absorbers generate
broader spectra, whereas all frequency spectra show a main lobe and several side
lobes with intermediate minima.
The frequency content of optoacoustic signals is essential for the respective
imaging application, as the highest detectable frequencies determine the highest
achievable spatial resolution (an exception is the lateral resolution in optoacoustic
microscopy, which is governed by the optical focusing capabilities, as will be
discussed later). Consequently, the differentiation between macroscopy (tomogra-
phy), mesoscopy, and microscopy based on the achieved spatial resolutions can be
related to the respective ranges of optoacoustic frequencies that are typically
detected.

4.2.4 Acoustic Attenuation

The detection of different optoacoustic frequency ranges has not only implications
on the achievable spatial resolution but also on the maximum imaging depth. More
specifically, the lower the highest detectable optoacoustic frequencies, the larger the
maximum penetration depth and vice versa. The reason for this relationship is a
process known as acoustic attenuation, which refers to a partial energy loss of
acoustic waves as they propagate through a medium. This effect originates from
frictional losses through the molecules of the medium and from heat diffusion
between adjacent volumes of differing temperature [11]. The acoustic attenuation
effect is generally stronger for higher than for lower frequencies and can be
described by the following power law:

p ( f ,r ) = p0 e
n
-a 0 f r
. (4.11)

Here, p(f, r) represents the component of the pressure amplitude with frequency
f after propagating a distance r in the medium, p0 is the initial pressure amplitude,
α0 denotes the attenuation constant in units of nepers, and n is a constant that
depends on the medium properties [12]. In tissue, the attenuation constant has an
average value of α0,t = 0.5 dB/MHz and an exponent of nt ≈ 1. The respective values
for water are α0,w = 0.000217 dB/MHz2 and nw ≈ 2 [13].
Consequently, acoustic attenuation in tissue is the limiting factor regarding the
highest detectable optoacoustic frequencies and thus the achievable spatial resolu-
tion with a particular optoacoustic imaging modality.
4 Multimodal Optoacoustic Imaging 77

4.2.5 Tomographic Reconstruction Techniques

The goal of tomographic reconstruction in optoacoustic imaging is to find the spa-


tial distribution of the initial pressure amplitudes p0 ( r ) and thus the optical absorp-

tion ma ( r ) (assuming a constant optical fluence rate, see (4.2)) in two or three
 
dimensions from the pressure waves p ( ri ,t ) measured at different positions ri (so-
called inverse problem) [14]. The available reconstruction techniques can be divided
into two main categories: analytical and numerical methods.
The most commonly used analytical reconstruction technique is the backprojec-
tion method, which is easy to implement, fast, and memory efficient. It is based on

the projection of the optoacoustic signals measured at positions ri onto spherical

shells (in the 3D case) centered at ri by exploiting the time-of-flight (TOF) infor-
mation of the time-domain signals and the known or assumed speed of sound c of
the medium. Consequently, for each time instant ti, the respective part of the
recorded signals is projected back to all positions from which they could have
emerged (i.e., spheres with radius c ⋅ ti, according to (4.6)), and all contributions are
summed up in the reconstruction volume. For the most common detection geome-
tries, i.e., planar, spherical, and cylindrical, the following backprojection formula

[15] provides an analytical expression for the initial pressure p0 at position r :

 dW é  ¶  ù
p0 ( r ) = 2 òò i ê p ( ri ,t ) - t ¶t p ( ri ,t ) ú   . (4.12)
W
W ë û t = r - ri / c

The factor dΩi/Ω is a solid angle weighting factor corresponding to detection

 
position i, which is constant in the far-field approximation (i.e., r - ri is much
larger than the imaged objects), and Ω is the total solid angle (2π for perfect planar
and 4π for closed spherical and cylindrical geometries).
Even though the backprojection method is fast and simple to implement, it does
not easily allow for other parameters, such as the detector properties, to be incorpo-
rated. On the other hand, numerical (also called model-based) reconstruction tech-
niques facilitate the modeling of detector properties and other physical effects by
using a system-dependent forward model and by discretizing the forward problem
spatially and temporally. The discretized forward problem can be written as a matrix
vector equation of the form
 
p = M × H, (4.13)

where
 p is the vector of measured pressure at different positions and time instants,
H is the vector containing the values of deposited energy (i.e., optical absorption)
in the reconstruction volume, and M represents the model matrix [14]. The goal of
all model-based reconstruction algorithms is to invert (4.13) in orderto solve for H ,
e.g., by iteratively minimizing the following expression to obtain H sol :
   
H sol = arg min

p - M × H + d reg . (4.14)
H
78 M. Omar et al.


Here, ‖⋅‖ refers to the l2 norm, and d reg is a regularization term that is necessary
in the case of ill-posed inverse problems [14]. The disadvantage of model-based
reconstruction methods is their high computational demand in terms of time and
memory. Therefore, model-based approaches are typically used in optoacoustic
tomography applications, where a mere hundreds of measurements are performed
per image. On the other hand, high-resolution (especially mesoscopic) optoacoustic
modalities that record thousands or even hundreds of thousands of projections per
scan usually rely on the backprojection technique for reconstruction.

4.3 Optoacoustic Macroscopy (Tomography)

4.3.1 Introduction

The word “tomography” originates from the ancient Greek words “tomos” (slice,
section) and “graphō” (to write) and refers to imaging by sections or sectioning
through the use of any kind of penetrating waves or mechanical method [16]. Hence,
optoacoustic tomography refers to optoacoustic sectional imaging of biological
organisms. Applications of optoacoustic tomography include the imaging of vascu-
lar structures, physiological readings, bio-distributions of targeted optical contrast
agents, kidney function, cerebrovascular activity, and tumor hypoxia [17].

4.3.2 Workings of Optoacoustic Tomography

Optoacoustic tomography (OAT)1 is, similar to all optoacoustic methods, based on

the optoacoustic effect, where a laser pulse illuminates the whole animal or imaged
section at once, followed by a collection of the generated optoacoustic signals, usu-
ally by an array of ultrasonic detectors (e.g., transducers). Finally, mathematical
methods or models are used to generate a useful image from the recorded signals.
To collect signals from whole animals, e.g., mice, low megahertz ultrasonic fre-
quencies, typically 3–10 MHz, are used in tandem with near-infrared illumination
to penetrate through 1–5 cm of tissue. Such a deep penetration allows for the nonin-
vasive acquisition of, e.g., cross-sectional slices of mice.
The type of collected information can be divided into three categories: anatomi-
cal, functional, and molecular. Anatomical information visualizes, for example, the
different organs of a mouse, such as the kidney and the liver, and gives information
about biological structures. This anatomical information requires only a single

1
Although the word “tomography” implies the use of mathematical methods for image reconstruc-
tion, in the context of imaging, it also generally refers to whole-body imaging at a macroscopic
scale. Hence, we will follow this terminological tradition here and use “macroscopy” and “tomog-
raphy” synonymously.
4 Multimodal Optoacoustic Imaging 79

illumination wavelength. If multiple wavelengths are available, it is possible to per-

form functional and molecular imaging based on the distinct absorption coefficients
of chromophores at different wavelengths. Functional imaging shows the activity
and the metabolism of an organ or tissue and is facilitated through imaging the oxy-
genation state of tissue rather than its shape. As previously mentioned, this is
achieved through imaging the tissue at multiple wavelengths in what is called mul-
tispectral imaging. Similarly, molecular imaging is enabled by imaging the distribu-
tion of a contrast agent inside the tissue or the bio-distribution of the molecular
agent. This kind of visualization is achieved by injecting a molecule that has a
defined absorption spectrum and mathematically unmixing it from the images
obtained at different wavelengths.
Finally, in order to perform sectional imaging, two approaches can be used.
Firstly, it is possible to have the ultrasonic detectors collect signals from all direc-
tions, to reconstruct a three-dimensional (3D) image, and to take a cross section
from the final reconstruction. Alternatively, it is possible and generally faster to use
cylindrically focused arrays, i.e., arrays that collect signals only from a single plane.
In such a way, only signals originating from the focal plane are detected, and a
tomographic image is generated. If multiple planes are required, the array or the
animal can be translated parallel to the axis of the array, and the reconstructed
images can be stacked in order to form a 3D volume.

4.3.3 Example Setups

Among the multitude of optoacoustic tomography systems available worldwide, the

setup developed at the Technical University of Munich (TU Munich) is one of the
most suitable and versatile for the imaging of small animals. The system (see
Fig. 4.4a) is based on a technology termed multispectral optoacoustic tomography

a b c
S
S
S FPI
B

UTA UTA CL
OF
OF OF IL BS

Fig. 4.4 Optoacoustic macroscopy (tomography) implementations. (a) Multispectral optoacous-

tic tomography (MSOT) based on a transducer ring array and ring illumination [30]. (b) Spherical
transducer array with a central hole for illumination [18]. (c) Optical raster-scanning of an inter-
rogation laser beam within a Fabry-Pérot resonator [29]. Abbreviations: B backing stub, BS x–y
beam scanner, CL convex lens, FPI Fabry-Pérot interferometer, IL interrogation laser, OF optical
fiber (bundle), S sample, UTA ultrasound transducer array. Green color: laser illumination
80 M. Omar et al.

(MSOT). The central part is a cylindrically focused transducer ring array, which
acquires the signals from a single plane, and a fast tunable near-infrared laser, which
can be tuned in the spectral range of 700–1000 nm. The fast tunability of the laser
allows the system to image rapid dynamic processes within a living animal, such as
metabolism and contrast agent diffusion. In addition to the large number of possible
applications, some of which will be discussed in the next subsection, the high wave-
length-tuning speed enables the imaging of dynamic processes without the need for
co-registration of the acquired images. Other examples of optoacoustic tomographic
geometries include spherical transducer arrays [18] and optical raster-scanning of
an interrogation laser beam within a Fabry-Pérot resonator [19], as illustrated in
Fig. 4.4b, c.

4.3.4 Example Applications

Optoacoustic tomography offers a broad range of noninvasive preclinical and bio-

medical applications. In this section, we will discuss three exemplary applications:
neuroimaging, cancer imaging, and the imaging of the bio-distribution of contrast
agents. Other applications include the imaging of tumor heterogeneity [20], inflam-
mation [21], and cardiovascular diseases [22, 23].

1. Neuroimaging: Because optoacoustic tomography is capable of imaging oxy-

genation states of tissue, it is ideally suited for the monitoring of activity and
metabolism, which are directly linked to oxygenation states inside the organs.
An interesting related application is the imaging of neuro-activity inside a mouse
brain, which is similar to the blood-oxygen-level dependent (BOLD) imaging
technique used in magnetic resonance imaging (MRI) but faster. Hence, it is pos-
sible to image not only the resting state of the brain but also faster neuro-activi-
ties [24–26]. An example of imaging the perfusion of a molecular agent inside a
mouse brain is shown in Fig. 4.5a [18].
2. Cancer imaging: Optoacoustic tomography in general and MSOT in particu-
lar allow for the imaging of cancer in mouse models. This technology can be
used for visualizing cancer heterogeneity [20], cancer hypoxia [27], and the
reaction of cancer to different kinds of therapy, such as chemotherapy [28].
The capabilities of the system enable the imaging of whole tumors and not
only the superficial regions as commonly done in microscopy. Additionally,
because MSOT is capable of noninvasive imaging, there is no need for a win-
dow chamber, which is less stressful for the animal. Figure 4.5b shows an
example application of optoacoustic macroscopy in the monitoring of tumor
development [29].
3. Imaging of bio-distribution: Because of its unique capability at distinguishing
between the spectra of different molecules, it is possible to image the bio-distri-
bution of a certain molecule with MSOT and to distinguish it from the surround-
ing tissue (see Fig. 4.5c) [30].
4 Multimodal Optoacoustic Imaging 81

CICG (a.u.)
b 0 c

Unmixed signal (a.u.)

Depth (mm)

5.2

Fig. 4.5 Examples of optoacoustic macroscopy (tomography) applications. (a) Monitoring of the
diffusion of a molecular agent (ICG) in a mouse brain over time using a spherical transducer array.
Reprinted with permission from [18]. (b) Imaging of a solid tumor together with the vascular bed
in a mouse model of cancer in vivo using a Fabry-Pérot resonator. Reprinted with permission from
[29]. (c) Cross section from a mouse body acquired using MSOT, showing the unmixed signal
from ICG (color) over a single wavelength image (gray). Reprinted with permission from [30]

4.3.5 Multimodal Tomographic Imaging

To increase the accessible range of contrast and to collect more comprehensive

information from the examined samples, optoacoustics offers the possibility for
multimodal imaging. Since in optoacoustics the sample is excited by light and ultra-
sound is detected, the most obvious and simplest multimodal combination would be
either to have an optical or an acoustical add-on to the OAT modality. Generally
speaking, an optical add-on could be in the form of diffuse optical tomography [31]
or fluorescence molecular tomography. Such a hybrid device would either enable
better reconstruction by collecting more information about the optical fluence in the
sample or measure fluorescence light and hence combine optical absorption contrast
with fluorescence contrast. Although optoacoustics can principally measure any
absorbing molecules including fluorescent ones, FMT is much more sensitive, espe-
cially for low concentrations of fluorescent agents.
On the other hand, an acoustic add-on can provide an additional dimension to
optoacoustic tomography, where ultrasound imaging visualizes the anatomical
82 M. Omar et al.

a b

Max

Fig. 4.6 Multimodal optoacoustic macroscopy (tomography) applications. (a) Hybrid optoacous-
tic (colors) and ultrasound (gray) imaging of a mouse brain. Courtesy of Ivan Olefir [24]. (b)
Hybrid optoacoustic (green) and MRI (gray) imaging of glioblastoma cancer in a mouse brain.
Reprinted with permission from [33]

structure of the sample, and optoacoustics gives more information about the molec-
ular and the functional dimensions of a disease or a specific activity [24] (see
Fig. 4.6a). Additionally, it is possible to use the ultrasound data for correcting the
optoacoustic images by deducing the correct speed of sound from the ultrasound
images, which can be later used in the optoacoustic image reconstruction [24, 32].
The beauty of such a hybrid imaging combination is that the ultrasonic detectors (or
detector arrays) are readily available and shared by both modalities. Thus, only an
ultrasound pulser is needed for the extra ultrasound imaging modality.
Additionally, it is possible to combine the images generated from an MRI modal-
ity with those from optoacoustics (see Fig. 4.6b) [33].

4.3.6 Conclusion

Optoacoustic tomography is a powerful modality that is capable of whole-body

small animal imaging, noninvasively and in vivo. By building upon multiwave-
length excitation as enabled by fast and tunable nanosecond lasers, it is possible to
add the functional and the molecular dimensions to the captured images and thus to
follow not only the anatomy but also the physiology of a disease. So far, this imag-
ing modality has been mostly applied in cancer research and neuroimaging. A
strong addition to this modality is the combination of optoacoustics with conven-
tional ultrasound imaging, which is facilitated by commonly employed hardware.

4.4 Optoacoustic Mesoscopy

4.4.1 Introduction

While optoacoustic macroscopy (tomography) has enabled whole animal or clinical

imaging of chromophores at centimeter penetration depths, many biomedical appli-
cations, such as imaging microvasculature, early-stage model organisms, or the
4 Multimodal Optoacoustic Imaging 83

small-scale spatial heterogeneity of diseases, require higher spatial resolutions than

offered by conventional OAT approaches. This demand has led to the development
of various high-resolution optoacoustic imaging methods, owing to the inherent
scalability of spatial resolution and penetration depth of the optoacoustic modality
[1]. More specifically, the detection of optoacoustic frequencies higher than
100 MHz enables spatial resolutions below 30 μm, while the hardware has to be
only slightly modified, i.e., high-frequency transducers and laser pulses of only a
few nanoseconds have to be used. On the other hand, the penetration depth of high-
resolution optoacoustic imaging implementations is fundamentally limited by
acoustic attenuation, which acts as a low-pass filter for optoacoustic frequencies
[12]. As will be discussed later, in optoacoustic microscopy, spatial resolutions
below 1 μm can be realized by using focused illumination, which further reduces
the penetration depth to <1 mm.
Optoacoustic mesoscopy seeks to fill the gap between macroscopic and micro-
scopic implementations of the optoacoustic modality by providing high spatial reso-
lutions of several tens of micrometers at intermediate imaging depths of a few
millimeters.

4.4.2 Scanning Schemes

Most optoacoustic mesoscopy systems are based on a single-element ultrasound

transducer that is either cylindrically or spherically focused and that is mechanically
scanned for tomographic imaging. As in macroscopy, tomographic image recon-
struction algorithms are employed to form three-dimensional volumes of optical
absorption in the sample. In the simplest case, a planar scanning geometry is used,
i.e., the transducer is scanned laterally within a plane above the sample. Even though
optoacoustic mesoscopy has been realized in transmission mode [34], the epi-illu-
mination design, where detection and illumination are located on the same side of
the sample, is preferable because it allows for arbitrary specimens to be imaged.
One of the first optoacoustic mesoscopy systems (also called acoustic resolution
photoacoustic microscopy (AR-PAM)) used a ring-shaped illumination focused into
the sample to yield a broad illuminated area, whereas the transducer was positioned
in the central dark spot above (see Fig. 4.7a) [35]. Another configuration used a
transducer with a central hole that accommodated an optical fiber providing the
illumination [36]. A recent implementation developed at the TU Munich is termed
raster-scan optoacoustic mesoscopy (RSOM) and relies on a conically shaped
spherically focused transducer and several optical fibers guiding the illumination
underneath the detector (see Fig. 4.7b) [37]. In all of the aforementioned systems,
transducer and illumination are scanned together by means of linear translation
stages.
Besides the planar geometry, other scanning schemes have been successfully
implemented in the past, such as combinations of cylindrical scanning geometries
and linear scans of single detectors (termed MORSOM) [38] or transducer arrays
[39, 40] together with a stationary illumination (see Fig. 4.7c). Another system used
84 M. Omar et al.

a b
UT

UT
M
OF

c d
CL UT
OF
UTA

Fig. 4.7 Optoacoustic mesoscopy implementations. (a) Epi-illumination mode dark-field config-
uration [35]. (b) Epi-illumination mode RSOM [37]. (c) Hybrid linear and circular scanning con-
figuration [39]. (d) Transmission mode light sheet illumination MSOM [41]. Abbreviations: CL
cylindrical lens, M mirror, OF optical fiber, UT ultrasound transducer, UTA ultrasound transducer
array. Green color: laser illumination

light sheet illumination and cylindrically focused detection in transmission mode

(see Fig. 4.7d) [41], in order to reduce out-of-plane signals leading to image
artifacts.

4.4.3 Spatial Resolution

The spatial resolution of all optoacoustic mesoscopy modalities is governed by the

properties of the used ultrasound transducer [42]. The lateral resolution δlat (m) is
determined by the width of the acoustic focus and is defined as

lac F
d lat = 0.71 » 1.4 lac , (4.15)
NA ac D

where λac denotes the wavelength of the detected acoustic waves; F and D are the
focal distance (m) and the diameter of the active element of the transducer (m),
respectively; and NAac represents the numerical aperture of the transducer (unitless)
[43]. On the other hand, the axial resolution δax (m) is given by

lac c
d ax » 0.88 , (4.16)
BW

where BW is the detection bandwidth of the transducer (Hz).

4 Multimodal Optoacoustic Imaging 85

4.4.4 Example Applications

Optoacoustic mesoscopy has been applied to preclinical imaging in numerous stud-

ies. The following section provides an overview of several example applications
(see Fig. 4.8).

a f

b c

d e

h 0.95

0.85

0.75

Fig. 4.8 Imaging applications of optoacoustic mesoscopy. (a–c) RSOM imaging of model organ-
isms. (a) Imaging of a zebrafish larva ex vivo. Red color: low-frequency reconstruction. White
color: high-frequency reconstruction. Reprinted with permission from [37]. (b) Ex vivo imaging
of a Drosophila pupa, visualizing the anatomical outline and the future wing location via tissue-
specific GFP expression. Reprinted with permission from [34]. (c) Broadband imaging of a mela-
noma tumor and the surrounding vascular network in a mouse in vivo. Red color: low-frequency
reconstruction showing larger blood vessels. Green color: high-frequency reconstruction showing
smaller features. Reprinted with permission from [44]. (d, e) Imaging of mCherry expression in
the brain of a juvenile zebrafish in vivo using light sheet illumination MSOM. (d) MSOM image
showing mCherry signals in color. (e) Corresponding histological slice showing fluorescence sig-
nals in color. Reprinted with permission from [41]. (f, g) Hybrid linear and circular scanning
MSOM imaging of an excised mouse kidney. (f) Optoacoustic image of the kidney. Gray color:
high-frequency signals originating from hemoglobin in the sample. Green color: low-frequency
unmixed signals visualizing the fluorescent dye IRDye 800CW in the renal pelvis (RP, red circle).
(g) Corresponding cryo-section photograph overlaid with the fluorescence signal of the dye (green
color). Abbreviations: AV arcuate vein, MP medullary pyramid, SV segmental vessel. Reprinted
with permission from [39]. (h) Dark-field optoacoustic mesoscopy (AR-PAM) imaging of the
oxygen saturation (sO2) in a mouse ear in vivo. Red and blue colors indicate sO2 levels according
to the shown color bar. Reprinted with permission from [35]
86 M. Omar et al.

1. Imaging of model organisms: Model organisms such as a zebrafish (Danio

rerio), Drosophila melanogaster, or C. elegans are interesting samples as they
are often used by developmental biologists and experimental geneticists to
study the development of certain organs (organogenesis), the neuronal system
(neurogenesis), or the development of the organisms itself from a single fertile
egg (­ morphogenesis). All these processes or their disruption can be easily moni-
tored in model organisms as they are relatively small, their maintenance is easy,
and their reproduction cycle is short. Hence, it is possible to modify certain
genes or crossbreed two different lines (e.g., zebrafish), resulting in new gen-
erations within a few months. With the RSOM modality, it was possible to
image the distribution of melanin-containing cells (melanophores) in an entire
zebrafish larva in a label-free manner (see Fig. 4.8a) [37] and to visualize the
location of the developing wings in an optically opaque Drosophila pupa
through labeling with GFP (see Fig. 4.8b) [34]. An improved label-free visual-
ization of zebrafish larvae with isotropic spatial resolution was enabled by
MORSOM [38].
2. Tumor monitoring: As already pointed out in the discussion of macroscopy
applications, optoacoustic imaging is ideally suited to investigate animal models
of cancer due to its multispectral capabilities. Beyond that, the higher spatial
resolution of optoacoustic mesoscopy allows for the label-free imaging of the
vascular network surrounding and supporting tumors. Additionally, the monitor-
ing of vascular growth around the tumor enables an assessment of cancer-related
angiogenesis, as has been successfully demonstrated with RSOM in a melanoma
mouse model in vivo (Fig. 4.8c) [44]. This capability is an important feature of
the optoacoustic modality as it is expected to lead to a more comprehensive
understanding of cancer development and related treatment designs.
3. High-resolution visualization of fluorescent markers in deep tissue: Multispectral
optoacoustic mesoscopy (MSOM) extends the capabilities of MSOT to high spa-
tial resolutions while still achieving millimeters of penetration depth. By using
selective plane illumination MSOM, the expression of the fluorescent protein
mCherry in the brain of a 6-month-old zebrafish was imaged in vivo (Fig. 4.8d,
e) [41]. Such a high-resolution identification of fluorescent reporter molecules in
deep tissues is a unique feature of MSOM, overcoming the limitations of fluores-
cence microscopy in terms of depth or of pure optical tomographic methods in
terms of resolution and in vivo applicability.
Furthermore, MSOM is particularly well suited for the multispectral imag-
ing of the whole organs, as its imaging depth is usually sufficient to penetrate
through the entire sample. By using a combination of linear and rotational scan-
ning of two different transducer arrays with distinct detection bandwidths, the
distribution of the fluorescent dye IRDye 800CW in the renal pelvis of an
excised mouse kidney was imaged together with the vasculature in the kidney
(Fig. 4.8f, g) [39].
4. Visualization of blood oxygen saturation: The multispectral feature of opto-
acoustic mesoscopy allows for the separation of oxy- and deoxyhemoglobin and
the label-free extraction of metabolic parameters in vivo, such as the blood
4 Multimodal Optoacoustic Imaging 87

o­ xygen saturation (sO2). One of the first successful demonstrations has been
shown on the vascular network in a mouse ear using dark-field optoacoustic
mesoscopy (AR-PAM) (Fig. 4.8h) [35].

4.4.5 Multimodal Mesoscopic Imaging

Similar to OAT approaches, ultrasound imaging capabilities can be readily inte-

grated in optoacoustic mesoscopy systems, as both modalities share the same trans-
ducer. A successful application of hybrid pulse-echo ultrasound and optoacoustic
mesoscopic imaging has been demonstrated by the visualization of laser irradiation-
induced changes in the cartilage and sclera [36]. The samples were impregnated
with bio-functional magnetite nanoparticles for optoacoustic contrast, which pene-
trated damaged areas of tissue via induced pores and crater-shaped structures. This
staining procedure led to a signal increase of damaged areas compared to healthy
tissue and allowed for the monitoring of the spatial distribution of laser-induced
changes. On the other hand, the ultrasound scans provided mechanical contrast-
based information about the changes of tissue thickness induced by the laser irradia-
tion. Figure 4.9 shows the hybrid optoacoustic (top view maximum amplitude
projections (MAP), top row) and ultrasound (side view cross sections, bottom row)
imaging of sclera tissue ex vivo before (left) and after (right) laser treatment. While
both optoacoustic and ultrasound images show structural changes in the sample, the
optoacoustic signals are indicative of the formation of pores in the tissue. Such a
hybrid imaging approach might lead to promising noninvasive monitoring strategies
for laser treatment procedures in ophthalmology and orthopedic applications.

4.4.6 Conclusion

Optoacoustic mesoscopy bridges the gap between optoacoustic macroscopy and

microscopy applications by providing high spatial resolution (tens of micrometers)
at decent imaging depths (several millimeters). It is a promising technique for the
noninvasive imaging of the whole organs or small animals. Similar to OAT, opto-
acoustic mesoscopy offers the opportunity to monitor metabolism or to visualize
external contrast agents by capitalizing on the multispectral imaging feature of
optoacoustics. So far, multimodal optoacoustic imaging at the mesoscopic scale has
been achieved in combination with conventional ultrasound imaging.

4.5 Optoacoustic Microscopy

4.5.1 Introduction

The development of optical microscopes has revolutionized biomedical research for

centuries, as it allows the monitoring of cellular and subcellular processes during
88 M. Omar et al.

Absorption (a.u.)
<0.01 0.05 0.09 1

a b

c d

<0.01 0.02 0.03 1 <0.008 0.01 0.03 1

Absorption (a.u.) US reflection (a.u.)

Fig. 4.9 Hybrid optoacoustic and pulse-echo ultrasound mesoscopic imaging of sclera tissue
ex vivo stained with bio-functional magnetite nanoparticles. (a, b) Optoacoustic top view MAPs
of the sclera before (a) and after (b) laser treatment of the area indicated by the dashed white
circle. A signal increase due to nanoparticle penetration into laser-induced pores can be observed
in (b). (c, d) Side view ultrasound cross sections (gray) along the blue lines in (a, b) overlaid with
optoacoustic signals (orange) before (c) and after (d) laser treatment. Reprinted with permission
from [77]

their occurrence. The contrast of most microscopic systems is based on light scat-
tering, phase differences, or fluorescence. The first two implementations only give
morphological information about the specimen, while fluorescence-based contrast
requires labeling or staining, which is a cumbersome and lengthy process. In addi-
tion to that, although labeled cells might show function as in the case of the fluores-
cent protein gCamp, the labels might be unstable, weak, and prone to bleaching, and
they might lead to phototoxic effects or even cell death.
Optoacoustics, as previously discussed, offers an alternative to the aforemen-
tioned optical modalities. By indirectly measuring the absorption of light rather
than scattering, phase changes, or fluorescence, it is possible to tap to functional and
physiological parameters. Optoacoustic microscopy (OAM, sometimes also termed
optical-resolution photoacoustic microscopy—OR-PAM) is based on focused light
4 Multimodal Optoacoustic Imaging 89

beams and thus achieves lateral resolutions comparable to traditional optical micros-
copy techniques, such as confocal or multiphoton microscopy. The generated opto-
acoustic signals are passively measured using an ultrasound detector. By scanning
the focused beam through the sample or by moving the sample itself, it is possible
to generate a high-resolution optoacoustic image of optical absorption around the
focal plane in the sample.

4.5.2 Theory

As previously mentioned, OAM relies on tightly focused optical beams. Therefore,

the focusing capability of the microscope defines the lateral resolution, while the
axial resolution is still defined through the bandwidth of the acoustic detector. To be
more precise, the lateral resolution δlat of OAM is defined as

lopt
d lat = 0.51 , (4.17)
NA opt

where λopt is the wavelength of the excitation light (m) and NAopt is the numerical
aperture of the focusing lens, which is a number that determines the focusing per-
formance of the optical system [43]. On the other hand, the axial resolution δax is
determined by (4.5) as in the case of optoacoustic mesoscopy. Since focused laser
beams are used, the energy requirements are lower in comparison to both opto-
acoustic mesoscopy and tomography. In typical applications, pulse energies below
100 nJ can be used.
Finally, as in all methods based on optical focusing, the imaging depth is limited
by optical scattering to a maximum depth of a few hundred micrometers.

4.5.3 Optoacoustic Microscopy Setups

Optoacoustic microscopy setups can be divided into two main categories, as illus-
trated by Fig. 4.10: transmission and epi-illumination mode implementations.
Similar to many optoacoustic mesoscopy applications, in transmission mode, the
excitation and the detection are performed on opposite sides of the sample, while in
epi-illumination mode, both are performed from the same side.

4.5.3.1 Epi-illumination Mode OAM

Epi-illumination implementations of OAM can be realized in many different ways.
One design uses a spherically focused detector with a central hole to accommodate
a GRIN lens-tipped fiber, which delivers the focused illumination [36]. Alternatively,
it is possible to use transparent detectors such as ring detectors [45, 46] and fiber
Bragg gratings (FBGs) [47] or systems similar to the Fabry-Pérot resonator setup
introduced in the mesoscopy section [19]. Another method is based on a special
kind of transparent acousto-optical beam combiner [48].
90 M. Omar et al.

a b OL

UT UT
CL

RAP
S

CS AW
RHP
OL
SO

Fig. 4.10 Optoacoustic microscopy implementations. (a) Typical transmission mode configura-
tion using sample scanning [50]. (b) Epi-illumination mode OAM based on a mechanically
scanned acousto-optical beam combiner (RAP and RHP) [78]. Abbreviations: AW acoustic waves,
CL correction lens, CS cover slip, OL objective lens, RAP right angle prism, RHP rhomboid prism,
S sample, SO silicone oil, UT ultrasound transducer. Green color: laser illumination

The key advantage of epi-illumination mode OAM over transmission mode

designs is the applicability to samples with arbitrary geometries. On the other hand,
many epi-illumination mode setups are limited in spatial resolution, because the
space demands for the acoustic and optical components typically impede the usage
of high NA objectives.

4.5.3.2 Transmission Mode OAM

In common transmission mode configurations, the acoustic detector and the focused
optical illumination are coaxially aligned. Furthermore, the detector and the excita-
tion can be either confocally aligned in order to increase the sensitivity, or the
acoustic detector can be slightly defocused in order to increase the scanning field of
view (FOV). The latter design leads to an increased imaging speed by rapidly scan-
ning the laser beam within the acoustic sensitivity field by means of scanning mir-
rors or other beam deflection components. In the system developed at the TU
Munich, the first approach has been pursued, where both the acoustic detector and
the optical focusing are aligned in a confocal manner. The sample is placed on a
fully motorized xyz stage, and a three-dimensional scan is performed. The scanning
head is mounted on top of an inverted multimodal microscope, which will be
described in more detail in the next section. Additionally, several detectors can be
used and are easily interchangeable. The advantage of a transmission mode design
over epi-illumination mode is the easiness of alignment and the possibility to use
any kind of ultrasound detectors [49, 50]. Another advantage over the epi-illumina-
tion mode design is the variety of compatible optics, where it is possible to use the
same high-quality and high NA objectives as employed in fluorescence microscopy.
Hence, it is possible to easily combine OAM with other optical microscopy tech-
niques, such as confocal or multiphoton microscopy, as we will see in the next
section.
4 Multimodal Optoacoustic Imaging 91

The system of the TU Munich is based on a commercial microscope where

instead of a CW laser commonly used in fluorescence microscopy, a pulsed nano-
second laser is coupled into the objective. The laser and ultrasound detector are
confocally aligned for maximum sensitivity, and an image is generated by scanning
the confocal spot through the sample.
In another implementation, the laser beam is rapidly raster-scanned in the sam-
ple by means of a set of galvanometric mirrors. For this purpose, the ultrasound
detector is lifted, and the beam is scanned in the positive defocus of the acoustic
sensitivity field. Such a configuration allows for FOVs of ~600 × 600 μm2 to be
scanned within a few minutes while achieving sufficient sensitivity. For even larger
fields of view, a combination of mechanical and galvanometric scanning could be
used [51, 52].

4.5.4 Example Applications

Figure 4.11 provides a summary of exemplary OAM applications. Aside from the
label-free imaging of single red blood cells, melanophores, or melanoma cells (see
Fig. 4.11a for the latter [53]), OAM has been applied in the following fields (for
more information, the reader is referred to the references mentioned in this
section):

1. Imaging of model organisms: As previously discussed, model organisms are

widely used in biology to study the development of organs or the organisms
themselves. In the case of the group of the TU Munich, melanophores forming
the pigmentation of zebrafish larvae were imaged with subcellular spatial resolu-
tion using a transmission mode OAM system and an optical sensor based on a
π-shifted fiber Bragg grating [47] (see Fig. 4.11b). For a more isotropic spatial
resolution and thus better visualization results regarding structures oriented par-
allel to the detection axis, multi-view methods similar to mesoscopic MORSOM
can be used [54].
2. Mouse ear vasculature visualization: Mouse ears are interesting organs as they
contain a dense network of vasculature and microvasculature and are addition-
ally thin in nature (typically around 300 μm). Thus, they are readily accessible
by transmission mode setups, where the ear is flattened between the illumination
and the ultrasonic detector. On the one hand, mouse ears are frequently used for
highlighting the capabilities of OAM systems in imaging a range of different
vascular sizes. On the other hand, similar to any part of the animal, it is possible
to study a multitude of different diseases in a mouse ear [55], such as cancer
growth [43].
3. Neuroimaging: The oxygenation state of a vascular network is linked to the
activity of the organ fed by these vessels. This is because blood carries both
nutrients and oxygen necessary for metabolic processes, which increase during
organ activity. BOLD-MRI is capitalizing on the same principle when moni-
toring the neuro-activity in a human brain. Similar to MRI, as we described in
92 M. Omar et al.

a b

Fractional change (%)

0 20

Fig. 4.11 Examples of optoacoustic microscopy applications. (a) Optoacoustic imaging of a sin-
gle melanoma cell. Upper left: bright-field image. Upper right, OAM top view MAP image; lower,
OAM side view MAP image, reprinted with permission from [79]. (b) Bright-field (left) and cor-
responding optoacoustic microscopy (right) imaging of melanophores in a zebrafish trunk ex vivo.
Courtesy of Georg Wissmeyer and Dominik Soliman. (c) Optoacoustic microscopy imaging of
neuro-activity in a mouse brain in vivo during the stimulation of the left (LHS) and right (RHS)
hind limb. Abbreviations: LH left hemisphere, RH right hemisphere. Reprinted with permission
from [51]

the section about optoacoustic macroscopy, optoacoustics is capable of moni-

toring the oxygen level of blood in real-time, noninvasively, and in a label-free
manner. In contrast to optoacoustic tomography, OAM is able to capture func-
tional and morphological changes in microvasculature. Hence, it is possible
not only to monitor the global behavior but also local changes on a micro-
scopic scale. As in OAT, this activity can be recorded by using multiple
excitation wavelengths, which are appropriately chosen to separate both the
oxy and the deoxy components of the blood. Additionally, because higher light
fluence rates are possible in microscopy applications, it is possible to take
advantage of nonlinear optoacoustic effects such as absorption saturation.
Based on the difference in the absorption relaxation rate, oxy- and deoxyhemo-
globin can be separated, allowing for the determination of oxygen saturation
levels. This method has been used to monitor neuronal activity in a mouse
brain in vivo (see Fig. 4.11c [51]).
4 Multimodal Optoacoustic Imaging 93

4.5.5 Multimodal Microscopic Imaging

As OAM uses similar components as other optical microscopy systems, it can be

readily combined with other high-resolution modalities within a common hybrid
imaging framework. In the case of the system at the TU Munich, the OAM system
is built around a commercial inverted microscopy stand and is combined with mul-
tiphoton microscopy, incorporating second harmonic generation (SHG), third har-
monic generation (THG), and two-photon excitation fluorescence (TPEF)
microscopy [49, 50]. Different laser sources are used for the different modalities,
which are coupled into the same microscope. For OAM, a nanosecond laser is
used, while multiphoton microscopy is based on a femtosecond laser to excite
nonlinear optical effects. The integration of all these modalities into a common
device enables the hybrid imaging of the same sample without the need of moving
it between different systems. At the same time, an enrichment of accessible label-
free contrast mechanisms is achieved, which enables the concurrent visualization
of different anatomical features in biological specimens and other samples.
Figure 4.12a shows an example of the label-free hybrid imaging of a zebrafish tail
ex vivo, where muscles, connective tissue, and melanophores are simultaneously
visualized through SHG (blue), THG (green), and OAM (red) imaging, respec-
tively [50].
In another application, the necrotic core region of a human carotid atheroma
slice was imaged with the hybrid microscope (see Fig. 4.12b). The SHG (green),
THG (blue), TPEF (yellow), and OAM (red) signals visualize collagen, tissue mor-
phology, elastin, and blood embeddings, respectively. This multimodal and stain-
ing-free imaging method is expected to play an important role in understanding
how plaques develop and thus devising the right strategies for treatment [52].
Similarly, such a technology could be used for better understanding of cancer
development in model animals and the mutual interactions of different molecules
inside the tumors.
Figure 4.12c shows the hybrid OAM and multiphoton imaging of a mouse ear
in vivo, visualizing blood vessels through OAM (red), cell boundaries and hair
follicles through THG (blue), as well as collagen via SHG (green) imaging in
3D [56].
In other applications, OAM has been successfully combined with fluorescence
[57] and confocal microscopy [58], ultrasound imaging [59], or optical coherence
tomography (OCT) [60].

4.5.6 Conclusion

Optoacoustic microscopy is a potent technology that allows for the visualization of

optical absorption, related to different functional processes and disease biomarkers,
at a microscopic scale. In this section, we discussed different implementations as
well as various applications of OAM and gave a more detailed discussion about the
system developed at the TU Munich. This setup, although based on a transmission
94 M. Omar et al.

a b

Fig. 4.12 Multimodal optoacoustic and multiphoton microscopy applications. (a) Hybrid imag-
ing of a zebrafish trunk ex vivo, visualizing the musculature (SHG, blue), connective tissue (THG,
green), and melanophores (OAM, red). Reprinted with permission from [50]. (b) Hybrid imaging
of the necrotic core region of a human carotid atheroma slice, visualizing collagen fibrils (SHG,
green), elastin (TPEF, yellow), overall tissue morphology (THG, blue), and blood embeddings
(OAM, red). Reprinted with permission from [52]. (c) Hybrid 3D imaging of a mouse ear in vivo,
visualizing collagen (SHG, green), hair follicles and cell boundaries (THG, blue), as well as blood
vessels (OAM, red). Reprinted with permission from [56]

mode configuration, allows for the imaging of a multitude of different samples,

ranging from model organisms, such as zebrafish and mouse ears, to tissue slices
and cell cultures. Limitations of OAM include its reliance on optical focusing,
which does not allow for penetration beyond a few hundred micrometers. On the
other hand, this technology enables the label-free imaging of disease biomarkers
and other bio-components that cannot be imaged otherwise or require potentially
harmful labeling or staining procedures. Because OAM uses similar laser sources
and optical components as other laser microscopy techniques, it is relatively simple
to combine OAM with other optical microscopic modalities, such as multiphoton
and confocal microscopy or OCT, in order to simultaneously gather comprehensive
anatomical information from complex biological samples. Furthermore, similar to
the macroscopic and mesoscopic case, OAM can be combined with pulse-echo
ultrasound imaging without the need for complicated additional hardware.
4 Multimodal Optoacoustic Imaging 95

4.6 Summary and Outlook

In this chapter, we introduced optical and optoacoustic imaging as well as the com-
bination of these techniques with other modalities, such as ultrasound imaging and
various optical imaging technologies. Because of advances in laser and ultrasonic
detection technology, optoacoustics has developed into a strong imaging modality
over the last two decades and has been applied in multiple applications, such as
neuroimaging [24, 51], cancer research [44, 61], imaging of the bio-distribution of
certain molecules, as well as cardiovascular imaging [22]. The versatility of opto-
acoustics is facilitated by its inherent scalability, enabling macroscopic, meso-
scopic, and microscopic imaging applications based on similar technology.
In the future, we expect an increasing number of cases where optoacoustics
becomes a mainstream imaging method, especially when combined with other
modalities (i.e., multimodal optoacoustic imaging). In such cases, optoacoustics is
complemented with modalities that either give a different view on the anatomy of
the examined specimen or a more sensitive measurement of fluorescently labeled
molecular agents. Additionally, by developing biomolecules that are engineered
toward yielding stronger optoacoustic signals [29, 62–64], optoacoustic imaging is
likely to become a more rounded technology on its own.
Another potential future mainstream direction is clinical imaging, where the
same technologies described in this chapter could be used in a clinical setting, either
for early and metastatic cancer diagnosis [65–67], endoscopic applications [68, 69],
tissue oxygenation [70], intravascular imaging [71], dermatology [65, 72, 73], or
other applications [74–76].

References
1. Ntziachristos V. Going deeper than microscopy: the optical imaging frontier in biology. Nat
Methods. 2010;7(8):603–14.
2. Ale A, Ermolayev V, Herzog E, Cohrs C, de Angelis MH, Ntziachristos V. FMT-XCT: in vivo
animal studies with hybrid fluorescence molecular tomography–X-ray computed tomography.
Nat Methods. 2012;9(6):615–20.
3. Ntziachristos V, Yodh AG, Schnall M, Chance B. Concurrent MRI and diffuse optical
tomography of breast after indocyanine green enhancement. Proc Natl Acad Sci U S A.
2000;97(6):2767–72.
4. Bowen T. Radiation-induced thermoacoustic soft tissue imaging. In: IEEE Ultrasonic
Symposium; 1981.
5. Wang LV, Wu H-I. Biomedical optics: principles and imaging. Hoboken, NJ: Wiley
Interscience; 2007.
6. Diebold GJ, Sun T, Khan MI. Photoacoustic monopole radiation in one, two, and three dimen-
sions. Phys Rev Lett. 1991;67(24):3384–7.
7. Westervelt PJ, Larson RS. Laser-excited broadside array. J Acoust Soc Am. 1973;54(1):121–2.
8. Bossy E, Gigan S. Photoacoustics with coherent light. Photoacoustics. 2016;4(1):22–35.
9. Khan MI, Diebold GJ. The photoacoustic effect generated by an isotropic solid sphere.
Ultrasonics. 1995;33(4):265–9.
10. Khan MI, Diebold GJ. The photoacoustic effect generated by laser irradiation of an isotropic
solid cylinder. Ultrasonics. 1996;34(1):19–24.
96 M. Omar et al.

11. Treeby BE, Cox BT, Zhang EZ, Patch SK, Beard PC. Measurement of broadband temperature-
dependent ultrasonic attenuation and dispersion using photoacoustics. IEEE Trans Ultrason
Ferroelectr Freq Control. 2009;56(8):1666–76.
12. Deán-Ben XL, Razansky D, Ntziachristos V. The effects of acoustic attenuation in optoacous-
tic signals. Phys Med Biol. 2011;56(18):6129–48.
13. Szabo TL. Diagnostic ultrasound imaging: inside out. Amsterdam: Academic Press; 2004.
14. Rosenthal A, Ntziachristos V, Razansky D. Acoustic inversion in optoacoustic tomography: a
review. Curr Med Imaging Rev. 2013;9:318–36.
15. Xu M, Wang LV. Universal back-projection algorithm for photoacoustic computed tomogra-
phy. Phys Rev E. 2005;71:016706.
16. Tomography—Wikipedia [Online]. Available: https://en.wikipedia.org/wiki/Tomography.
Accessed 4 Sep 2016.
17. Buehler A, Kacprowicz M, Taruttis A, Ntziachristos V. Real-time handheld multispectral opto-
acoustic imaging. Opt Lett. 2013;38(9):1404–6.
18. Deán-Ben XL, Razansky D. Adding fifth dimension to optoacoustic imaging: volumetric time-
resolved spectrally enriched tomography. Light Sci Appl. 2014;3(1):e137.
19. Laufer J, Johnson P, Zhang E, Treeby B, Cox B, Pedley B, Beard P. In vivo preclini-
cal photoacoustic imaging of tumor vasculature development and therapy. J Biomed Opt.
2012;17(5):56016.
20. Herzog E, Taruttis A, Beziere N, Lutich AA, Razansky D. Optical imaging of cancer heteroge-
neity with multispectral. Radiology. 2012;263(2):461–8.
21. Beziere N, Schacky C, Kosanke Y, Kimm M, Nunes A, Licha K, Aichler M, Walch A,
Rummeny EJ, Ntziachristos V, et al. Optoacoustic imaging and staging of inflammation in a
murine model of arthritis. Arthritis Rheumatol. 2014;66(8):2071–8.
22. Taruttis A, Wildgruber M, Kosanke K, Beziere N, Licha K, Haag R, Aichler M, Walch A,
Rummeny E, Ntziachristos V. Multispectral optoacoustic tomography of myocardial infarc-
tion. Photoacoustics. 2013;1(1):3–8.
23. Taruttis A, Ntziachristos V. Advances in real-time multispectral optoacoustic imaging and its
applications. Nat Photon. 2015;9(4):219–27.
24. Olefir I, Merčep E, Burton NC, Ovsepian SV, Ntziachristos V. Hybrid multispectral optoacous-
tic and ultrasound tomography for morphological and physiological brain imaging. J Biomed
Opt. 2016;21(8):86005.
25. Nasiriavanaki M, Xia J, Wan H, Bauer AQ, Culver JP, Wang LV. High-resolution photoacoustic
tomography of resting-state functional connectivity in the mouse brain. Proc Natl Acad Sci U
S A. 2013;8943:9–13.
26. Gottschalk S, Fehm TF, Deán-Ben XL, Tsytsarev V, Razansky D. Correlation between volu-
metric oxygenation responses and electrophysiology identifies deep thalamocortical activity
during epileptic seizures. Neurophotonics. 2017;4(1):11007.
27. Tzoumas S, Nunes A, Olefir I, Stangl S, Symvoulidis P, Glasl S, Bayer C, Multhoff G,
Ntziachristos V. Eigenspectra optoacoustic tomography achieves quantitative blood oxygen-
ation imaging deep in tissues. Nat Commun. 2016;7:12121.
28. Bohndiek SE, Sasportas LS, Machtaler S, Jokerst JV, Hori S, Gambhir SS. Photoacoustic
tomography detects early vessel regression and normalization during ovarian tumor response
to the antiangiogenic therapy trebananib. J Nucl Med. 2015;56(12):1942–7.
29. Jathoul AP, Laufer J, Ogunlade O, Treeby B, Cox B, Zhang E, Johnson P, Pizzey AR, Philip
B, Marafioti T, et al. Deep in vivo photoacoustic imaging of mammalian tissues using a tyros-
inase-based genetic reporter. Nat Photon. 2015;9:239–46.
30. Razansky D, Buehler A, Ntziachristos V. Volumetric real-time multispectral optoacoustic
tomography of biomarkers. Nat Protoc. 2011;6(8):1121–9.
31. Bauer AQ, Nothdurft RE, Erpelding TN, Wang LV, Culver JP. Quantitative photoacoustic
imaging: correcting for heterogeneous light fluence distributions using diffuse optical tomog-
raphy. J Biomed Opt. 2011;16(9):96016.
32. Mercep E, Burton NC, Claussen J, Razansky D. Whole-body live mouse imaging by hybrid
reflection-mode ultrasound and optoacoustic tomography. Opt Lett. 2015;40(20):4643–6.
4 Multimodal Optoacoustic Imaging 97

33. Kircher M, La Zerda AD, Jokerst J, Zavaleta C. A brain tumor molecular imaging strategy using
a new triple-modality MRI-photoacoustic-Raman nanoparticle. Nat Med. 2012;18:829–34.
34. Omar M, Gateau J, Ntziachristos V. Raster-scan optoacoustic mesoscopy in the 25–125 MHz
range. Opt Lett. 2013;38(14):2472–4.
35. Zhang HF, Maslov K, Stoica G, Wang LV. Functional photoacoustic microscopy for high-
resolution and noninvasive in vivo imaging. Nat Biotechnol. 2006;24(7):848–51.
36. Estrada H, Turner J, Kneipp M, Razansky D. Real-time optoacoustic brain microscopy with
hybrid optical and acoustic resolution. Laser Phys Lett. 2014;11(4):45601.
37. Omar M, Soliman D, Gateau J, Ntziachristos V. Ultrawideband reflection-mode optoacoustic
mesoscopy. Opt Lett. 2014;39(13):3911–4.
38. Omar M, Rebling J, Wicker K, Schmitt-Manderbach T, Schwarz M, Gateau J, López-Schier H,
Mappes T, Ntziachristos V. Optical imaging of post-embryonic zebrafish using multi orienta-
tion raster scan optoacoustic mesoscopy. Light Sci Appl. 2017;6:e16186.
39. Chekkoury A, Gateau J, Driessen W, Symvoulidis P, Bézière N, Feuchtinger A, Walch A,
Ntziachristos V. Optical mesoscopy without the scatter: broadband multispectral optoacoustic
mesoscopy. Biomed Opt Express. 2015;6(9):3134–48.
40. Chekkoury A, Nunes A, Gateau J, Symvoulidis P, Feuchtinger A, Beziere N, Ovsepian
SV, Walch A, Ntziachristos V. High-resolution multispectral optoacoustic tomogra-
phy of the vascularization and constitutive hypoxemia of cancerous tumors. Neoplasia.
2016;18(8):459–67.
41. Razansky D, Distel M, Vinegoni C, Ma R, Perrimon N, Köster RW, Ntziachristos
V. Multispectral opto-acoustic tomography of deep-seated fluorescent proteins in vivo. Nat
Photon. 2009;3(7):412–7.
42. Cobbold RSC. Foundations of biomedical ultrasound. New York: Oxford University Press;
2006.
43. Yao J, Wang LV. Photoacoustic microscopy. Laser Photon Rev. 2014;7(5):1–36.
44. Omar M, Schwarz M, Soliman D, Symvoulidis P, Ntziachristos V. Pushing the optical imag-
ing limits of cancer with multi-frequency-band raster-scan optoacoustic mesoscopy (RSOM).
Neoplasia. 2015;17(2):208–14.
45. Li H, Dong B, Zhang Z, Zhang HF, Sun C. A transparent broadband ultrasonic detector based
on an optical micro-ring resonator for photoacoustic microscopy. Sci Rep. 2014;4:4496.
46. Huang S-W, Chen S-L, Ling T, Maxwell A, O’Donnell M, Guo LJ, Ashkenazi S. Low-
noise wideband ultrasound detection using polymer microring resonators. Appl Phys Lett.
2008;92(19):193509.
47. Wissmeyer G, Soliman D, Shnaiderman R, Rosenthal A, Ntziachristos V. All-
optical optoacoustic microscope based on wideband pulse interferometry. Opt Lett.
2016;41(9):1953–6.
48. Maslov K, Zhang HF, Hu S, Wang LV. Optical-resolution photoacoustic microscopy for in vivo
imaging of single capillaries. Opt Lett. 2008;33(9):929–31.
49. Tserevelakis GJ, Soliman D, Omar M, Ntziachristos V. Hybrid multiphoton and optoacoustic
microscope. Opt Lett. 2014;39(7):1819–22.
50. Soliman D, Tserevelakis GJ, Omar M, Ntziachristos V. Combining microscopy with mesos-
copy using optical and optoacoustic label-free modes. Sci Rep. 2015;5:12902.
51. Yao J, Wang L, Yang J-M, Maslov KI, Wong TTW, Li L, Huang C-H, Zou J, Wang LV. High-
speed label-free functional photoacoustic microscopy of mouse brain in action. Nat Methods.
2015;12(5):407–10.
52. Seeger M, Karlas A, Soliman D, Pelisek J, Ntziachristos V. Multimodal optoacoustic and mul-
tiphoton microscopy of human carotid atheroma. Photoacoustics. 2016;4:102–11.
53. Shelton RL, Mattison SP, Applegate BE. Volumetric imaging of erythrocytes using label-free
multiphoton photoacoustic microscopy. J Biophoton. 2014;7(10):834–40.
54. Zhu L, Li L, Gao L, Wang LV. Multiview optical resolution photoacoustic microscopy. Optica.
2014;1(4):217–22.
55. Wang L, Maslov K, Yao J, Rao B, Wang LV. Fast voice-coil scanning optical-resolution photo-
acoustic microscopy. Opt Lett. 2011;36(2):139–41.
98 M. Omar et al.

56. Song W, Xu Q, Zhang Y, Zhan Y, Zheng W, Song L. Fully integrated reflection-mode pho-
toacoustic, two-photon, and second harmonic generation microscopy in vivo. Sci Rep.
2016;6:32240.
57. Tserevelakis GJ, Tsagkaraki M, Zacharakis G. Hybrid photoacoustic and optical imaging of
pigments in vegetative tissues. J Microsc. 2016;263:300–6.
58. Wang Y, Maslov K, Kim C, Hu S, Wang LV. Integrated photoacoustic and fluorescence confo-
cal microscopy. IEEE Trans Biomed Eng. 2010;57(10):2576–8.
59. Harrison T, Ranasinghesagara JC, Lu H, Mathewson K, Walsh A, Zemp RJ. Combined photo-
acoustic and ultrasound biomicroscopy. Opt. Express. 2009;17(24):22041–6.
60. Jiao S, Xie Z, Zhang HF, Puliafito CA. Simultaneous multimodal imaging with inte-
grated photoacoustic microscopy and optical coherence tomography. Opt Lett.
2009;34(19):2961–3.
61. Yao J, Shcherbakova DM, Li C, Krumholz A, Lorca RA, Reinl E, England SK, Verkhusha VV,
Wang LV. Reversibly switchable fluorescence microscopy with enhanced resolution and image
contrast. J Biomed Opt. 2014;19:86018.
62. Stiel AC, Deán-Ben XL, Jiang Y, Ntziachristos V, Razansky D, Westmeyer GG. High-contrast
imaging of reversibly switchable fluorescent proteins via temporally unmixed multispectral
optoacoustic tomography. Opt Lett. 2015;40(3):367–70.
63. Jiang Y, Sigmund F, Reber J, Deán-Ben XL, Glasl S, Kneipp M, Estrada H, Razansky D,
Ntziachristos V, Westmeyer GG. Violacein as a genetically-controlled, enzymatically ampli-
fied and photobleaching-resistant chromophore for optoacoustic bacterial imaging. Sci Rep.
2015;5:11048.
64. Stritzker J, Kirscher L, Scadeng M, Deliolanis NC, Morscher S, Symvoulidis P, Schaefer K,
Zhang Q, Buckel L, Hess M, et al. Vaccinia virus-mediated melanin production allows MR and
optoacoustic deep tissue imaging and laser-induced thermotherapy of cancer. Proc Natl Acad
Sci. 2013;110(9):3316–20.
65. Aguirre J, Schwarz M, Soliman D, Buehler A, Omar M, Ntziachristos V. Broadband meso-
scopic optoacoustic tomography reveals skin layers. Opt Lett. 2014;39(21):6297.
66. Schwarz M, Omar M, Buehler A, Aguirre J, Ntziachristos V. Implications of ultra-
sound frequency in optoacoustic mesoscopy of the skin. IEEE Trans Med Imaging.
2015;34(2):672–7.
67. Stoffels I, Morscher S, Helfrich I, Hillen U, Leyh J, Burton NC, Sardella TCP, Claussen
J, Poeppel TD, Bachmann HS, et al. Metastatic status of sentinel lymph nodes in mela-
noma determined noninvasively with multispectral optoacoustic imaging. Sci Transl Med.
2015;7(317):317ra199.
68. Yang J-M, Favazza C, Chen R, Yao J, Cai X, Maslov K, Zhou Q, Shung KK, Wang
LV. Simultaneous functional photoacoustic and ultrasonic endoscopy of internal organs
in vivo. Nat Med. 2012;18(8):1297–302.
69. He H, Buehler A, Ntziachristos V. Optoacoustic endoscopy with curved scanning. Opt Lett.
2015;40(20):4667–70.
70. Diot G, Dima A, Ntziachristos V. Multispectral opto-acoustic tomography of exercised muscle
oxygenation. Opt Lett. 2015;40(7):1496–9.
71. Sethuraman S, Amirian JH, Litovsky SH, Smalling RW, Emelianov SY. Spectroscopic
intravascular photoacoustic imaging to differentiate atherosclerotic plaques. Opt Express.
2008;16(5):3362–7.
72. Aguirre J, Schwarz M, Garzorz N, Omar M, Buehler A, Eyerich K, Ntziachristos V. Precision
assessment of label-free psoriasis biomarkers with ultra-broadband optoacoustic mesoscopy.
Nat Biomed Eng. 2017;1:0068.
73. Schwarz M, Soliman D, Omar M, Buehler A, Aguirre J, Ntziachristos V. Optoacoustic dermos-
copy of the human skin: tuning excitation energy for optimal detection bandwidth with fast and
deep imaging in vivo. IEEE Trans Med Imaging. 2017;36:1287–96.
74. Taruttis A, Timmermans AC, Wouters PC, Kacprowicz M, van Dam GM, Ntziachristos
V. Optoacoustic imaging of human vasculature: feasibility by using a handheld probe.
Radiology. 2016;281(1):256–63.
4 Multimodal Optoacoustic Imaging 99

75. Neuschmelting V, Burton NC, Lockau H, Urich A, Harmsen S, Ntziachristos V, Kircher

MF. Performance of a multispectral optoacoustic tomography (MSOT) system equipped with
2D vs. 3D handheld probes for potential clinical translation. Photoacoustics. 2016;4(1):1–10.
76. Waldner MJ, Knieling F, Egger C, Morscher S, Claussen J, Vetter M, Kielisch C, Fischer S,
Pfeifer L, Hagel A, et al. Multispectral optoacoustic tomography in Crohn’s disease: non-
invasive imaging of disease activity. Gastroenterology. 2016;151:238–40.
77. Estrada H, Sobol E, Baum O, Razansky D. Hybrid optoacoustic and ultrasound biomicroscopy
monitors’ laser-induced tissue modifications and magnetite nanoparticle impregnation. Laser
Phys Lett. 2014;11(12):125601.
78. Hu S, Maslov K, Wang LV. Second-generation optical-resolution photoacoustic microscopy
with improved sensitivity and speed. Opt Lett. 2011;36(7):1134–6.
79. Strohm EM, Berndl ESL, Kolios MC. High frequency label-free photoacoustic microscopy of
single cells. Photoacoustics. 2013;1(3):49–53.
Small Animal Imaging in Oncology Drug
Development 5
Joseph D. Kalen and James L. Tatum

5.1 Introduction

Major advances in small animal imaging have been made during the last two decades
encompassing a full array of platforms that image along the electromagnetic spec-
trum from MRI (100–101 m), optical (10−6 m), X-ray (10−9 m), to nuclear (10−11–
10−12 m). This in part has been facilitated by the National Cancer Institute (NCI),
National Institutes of Health (NIH) through the support of Small Animal Imaging
Research Programs (SAIRP), and other initiatives to increase the availability of
small animal imaging platforms and develop the expertise in the use of these meth-
ods. While the primary application of these new techniques has been research tools
to answer scientific questions especially related to the understanding of in vivo sys-
tems, another area of interest has been the introduction of imaging-based in vivo
assay systems for drug development in oncology. In fact, a major effort has been
undertaken to integrate in vivo imaging biomarker development with in vitro bio-
marker development in contrast to the historical scenario of applying imaging only
late in the development plan, leading to the conundrum of validation of imaging
while trying to employ imaging as a biomarker.
Drug development is a high-risk business in which late-stage failures are espe-
cially costly, with an average cost (capitalized and out of pocket) (2016) of

J. D. Kalen (*)
Small Animal Imaging Program, Laboratory Animal Sciences Program, Frederick National
Laboratory for Cancer Research Sponsored by the National Cancer Institute,
Frederick, MD, USA
e-mail: [email protected]
J. L. Tatum
Cancer Imaging Program, Division of Cancer Treatment and Diagnosis,
National Cancer Institute, NIH, Bethesda, MD, USA
e-mail: [email protected]

© Springer Nature Switzerland AG 2019 101

C. Kuntner-Hannes, Y. Haemisch (eds.), Image Fusion in Preclinical
Applications, https://doi.org/10.1007/978-3-030-02973-9_5
102 J. D. Kalen and J. L. Tatum

approximately $3.95B ($1.528B (preclinical) and $2.43B (clinical)) normalized to

2013 dollars [1]. Trends in capitalized costs since the 1970s (based on 2013 dollars)
have shown dramatic increases in preclinical and clinical costs, 1007% and 2085%,
respectively [1]. Although the monetary costs are obvious, the opportunity cost of
such failures, consuming valuable resources and time, may lead to more significant
health cost. The focus on targeted agents has further complicated the development
process. Unfortunately, late-phase failures with new targeted agents are not uncom-
mon and frequently the result of inadequate biomarker development. It has been
reported that robust biomarkers are essential to successful drug development and
can improve the success rate for phase I to drug approval as much as 25.9% (with
biomarkers) as compared to 8.4% (without biomarkers) [2]. In oncology, drugs and
especially new pathway-specific drugs non-context in vitro assays may not be ade-
quate when applied to clinical scenarios which are highly contextually based. Small
animal imaging with image fusion encompassing different modalities and molecu-
lar probes has the potential to enhance our understanding of drug candidates and
combinations in context by serving as an in vivo assay system allowing evaluation
of various biomarkers in the complex biologic system of cancer. While imaging of
small animals is now routinely performed daily in many labs, converting such imag-
ing to an in vivo assay for drug development is significantly more challenging.
Performing small animal imaging in the context of an in vivo assay for drug
development encompasses numerous aspects: standardization of equipment pro-
cesses (SOPs) including quality control and quality assurance, data acquisition,
analysis, and validation of output with respect to a gold standard (i.e., pathology)
[3–6] are all important requirements for developing an assay for drug development.
Furthermore, other aspects such as animal handling, anesthesia and understanding
its effect on the animals physiology (pulmonary and cardiac functions and stress
levels) [7–11], personnel safety for handling animals that contain toxic chemicals
[12], validation of animal model platforms [5], multi-animal throughput to obtain
statistical significance, and multi-imaging platforms to discern a drugs effect on the
various physiological [13], anatomical, and molecular pathways are all significant
aspects for developing a robust drug development in vivo assay. Numerous refer-
ences have been cited that describe the above factors for performing small animal
imaging.
The main aspect of this book is to provide an overview of multimodality preclini-
cal imaging, and this chapter is intended to demonstrate how these techniques can
be implemented for oncology drug development. Initial utilization of in vivo imag-
ing incorporated clinical scanners for simple analysis of anatomical tumor volumes
or a metabolic function using the clinically available radiopharmaceutical 2-deoxy-­
2-(18F)fluoro-d-glucose [18F]FDG. As equipment manufacturers developed and
modified scanners explicitly for small animals, incorporating higher spatial image
resolution and faster acquisitions to analyze the small animal’s rapid heart, pulmo-
nary, and biological rates, these scanners were providing in vivo techniques to allow
researchers the ability to question and comprehend specific processes pertinent to
developing an oncologic drug. Furthermore, due to the inherent physics of each
system’s attributes to acquire data along the electromagnetic spectrum and
5 Small Animal Imaging in Oncology Drug Development 103

Table 5.1 Comparison of preclinical in vivo imaging modalities

Image Probe Amount
resolution 3D capability sensitivity Intrinsic of probe Activatable Dynamic
Modality (μm) (tomography) (mol/L) contrast required probes studies
X-ray CT 5 Yes N/A Yes No No
MRI 170 Yes 103 to 10−5 Yes μg–mg No Yes
PET 1000 Yes 10−11 to 10−12 No ng No Yes
SPECT 150–2000 Yes 10−10 to 10−11 No ng No Yes
(determined by
collimator and
head orientation)
Ultrasound 30 Limited High (not Yes μg–mg Yes Yes
(small well
volume) characterized)
Photo-acoustic 44–75 (Small 10−7 Yes
volume)
Bioluminescence >1000 Limited 10−15 to 10−17 Yes μg–mg Yes (?) Yes (2D)
(semi-
quantitative)
2D fluorescence >1000 No 10−9 to 10−12 Yes μg–mg Yes Yes
Fluorescence <1000 Semi- 10−9 to 10−12 Yes μg–mg Yes No
tomography quantitative

utilization of various molecular imaging agents to probe specific pathways

(Table 5.1), it has become necessary to incorporate numerous modalities to investi-
gate the micro- and macro-­biologic system for drug development.

5.2 Development and Validation of Model Platforms

To understand a drugs interaction within a biologic system, pertinent animal models

have been established, from the simple subcutaneous (sc) injection of human cells
to orthotopic models to genetically modified mouse models (GEMMs) incorporat-
ing knock-in, knockout, and CRISPR [14–16] technologies. Development of mouse
model platforms for oncology drug development must also be validated utilizing
histotechnological processes with respect to pathological standards such as angio-
genesis (thymidine) and apoptosis (caspases, FITC-labeled Annexin V) [17].
Research groups such as the Biological Testing Branch, Division of Cancer
Treatment and Diagnosis, and the Center for Advanced Preclinical Research
(CAPR), Center for Cancer Research, both within NCI, have developed and fully
characterized various xenograft models incorporating both human cells and tumor
fragments and GEMM models (pancreas, lung, and ovarian) for testing new drugs
against standard clinical therapies [18, 19].
In addition to the primary tumor, model platforms for the study of metastasis
should be included in the repertoire [20]. Metastasis of cancer cells from a primary
tumor is a leading cause of death [21], and early detection for timely therapeutic
104 J. D. Kalen and J. L. Tatum

intervention with serial in vivo imaging would greatly improve clinical outcomes.
Bioluminescence imaging (BLI) has been shown to be sensitive with the ability to
image few cells, correlates to tumor volume as validated by gadolinium contrast
MRI, and provides rapid imaging for high throughput [22]. Unfortunately, it only
provides 2D images, and the depth penetration of light is limited to a few cm. On the
other hand, utilizing BLI to first determine the presence of a metastatic signal (pre-
screening technique), due to the BLI higher sensitivity and the higher throughput
with respect to 3D small-bore modalities, BLI can improve utilization of higher-­
cost 3D modalities. This demonstrates that multimodality imaging does not man-
date concurrent image acquisitions. Utilizing one modality to screen for presence of
metastasis can greatly enhance utilization of another modality. Furthermore, due to
the metastasis textural characteristics (i.e., echogenicity), a higher spatial resolution
scanner (i.e., 30 μm for ultrasound) might not detect the metastatic lesion, whereas
a lower spatial resolution scanner (i.e., 170 μm for MRI) can provide a higher-­
contrast signal (Fig. 5.1), such as in a T2* MRI sequence.
Further advancements in oncology animal models capturing the cell-autonomous
(genetic and epi-genetic) and non-cell autonomous (stromal) aspects of tumor het-
erogeneity improve the understanding of patient-specific responses to therapy (pre-
cision medicine); thus cancer researchers have instituted patient-derived xenograft
(PDX) animal model studies [23, 24]. The heterogeneity in the tumor fragment can
be attributed to the tumor matrix that is transplanted with the tumor. This matrix
tends to persist as the tumor grows, eventually becoming permeated and dissolving
into the tumor. While there is mild heterogeneity seen in cellular base xenografts,

a b

Fig. 5.1 These two images demonstrate the marked difference in image contrast (red arrow) for a
metastatic lesion (a) isoechoic with capsule in a B-mode ultrasound scanner (30 μm image resolu-
tion) and (b) high-contrast T2 signal in a 3 T MRI (150 μm image resolution) image with coils
specific for small animals. While the resolution of US is greater than MRI, it is more difficult to
compare lesions longitudinally on US. However, small animal US provides greater capability for
dynamic characterization of individual lesions
5 Small Animal Imaging in Oncology Drug Development 105

the heterogeneity in the fragment group is far more typical of what is seen through-
out the tumor growth in this group and is hypothesized to be more representative of
the native in vivo microenvironment.

5.3 In Vivo Imaging for Oncology Translational Research

Within the vast array of technologies for small animal imaging, there are many
opportunities to design imaging-based experiments to answer complex biological
questions. However, in drug/therapy development, the requirements become more
demanding and less flexible, shifting from an imaging experiment to an in vivo
assay. To institute an in vivo assay, three key elements must exist: (1) relevant bio-
marker matched to imaging capability, (2) highly reproducible results, and (3) effi-
cient throughput. Another aspect of importance is the ability to translate nonclinical
assays to clinical research as needed for future development.
As previously discussed, this requires rigorous SOPs, a validated animal model,
imaging equipment quality control, close attention to animal handling, quality con-
trol of probes, contrast agents, and radiopharmaceuticals, close adherence to acqui-
sition protocols and a standardized analysis. The routine incorporation of both
positive and negative pathological standards is also critical.
This conversion of the imaging experiment to an in vivo assay requires the devel-
opment of an imaging assay platform that incorporates the appropriate validated
model and a highly controlled imaging protocol designed to optimally measure the
biomarker of interest. For the assay to be practical, both logistical and physiologic
barriers need to be minimized by incorporation of such practices as keeping intrave-
nous administrations to a minimum, reducing anesthesia sessions, and using sys-
tems that allow concurrent imaging of animals but allowing constant monitoring.
The implementation of small animal imaging in oncology drug development will
now be described using three case study examples.

5.3.1 Case Studies

5.3.1.1 D  evelopment of an Imaging Platform Consisting

of an Animal Model and MR Technique
Colorectal cancer (CRC) is the third most common cancer in the United States, and
the American Cancer Society estimated that in 2018 there will be 97,220 new cases
and 50,630 deaths [25] and that chronic inflammation, such as ulcerative colitis and
Crohn’s disease, is associated with increased risk of CRC. To study colorectal can-
cer, animal models as an imaging assay platform can assist in the assessment of the
initial stages of cancer and therapy response.
Conventional micro-endoscopes can provide appropriate information on colorec-
tal cancer, i.e., imaging polyps, but risk perforating the colon or obstruction of the
image due to bleeding associated with colitis. 3D in vivo imaging (i.e., virtual colo-
noscopy) can image the early phases of cancer, and utilizing various imaging agents
106 J. D. Kalen and J. L. Tatum

can probe molecular pathways for characterization, including when altered by a

drug. Clinically, X-ray CT has been a standard in performing virtual colonoscopy
due to its rapid image acquisition. Preclinical X-ray CT scanners can provide infor-
mation on polyp growth [26, 27] but are unable to provide the high tissue contrast
necessary to discern normal and inflamed lumen tissue from surrounding tissue
without resulting in high-radiation doses, especially in preclinical studies when
serial imaging is required. Another method used in the clinic, magnetic resonance
colonography (MRC), can discern normal and inflammatory tissue utilizing the
dark lumen technique with either water or gas to expand the colon followed by IV
contrast (Gd-chelate)-enhanced T1w MRI sequence, where water, if used as an
enema, remains dark. T2w MRI is required to discern inflammatory tissue and also
provides a rapid acquisition for high-throughput; unfortunately the water enema
results in a bright signal. Preclinical MRI colorectal studies have utilized other tech-
niques such as fecal tagging [28], and water enema [29], which unfortunately cre-
ates a bright lumen in T2w images making it difficult to distinguish normal from
inflamed colonic mucosa, especially for researching drugs for chronic inflammation
(pre-CRC).
Thus, the aim was to develop an imaging platform consisting of a mouse model
and an adjusted MR protocol to study colorectal cancer. The in vivo MRI protocol
should provide several important components: noninvasive serial imaging to moni-
tor tumor progression and tissue inflammation, artifact-free imaging on T1w and
T2w MRI sequences, enhanced image contrast (high signal-to-noise ratio: SNR) to
discern inflamed and normal lumen tissue from surrounding tissue, quantitative
imaging, and high throughput.
FVB/N mice were dosed (10 mg/kg, IP route) with a chemical carcinogen
azoxymethane (AOM) and exposed 1 week later to the colonic irritant dextran
sodium (DSS) [1–5% DSS dissolved in drinking water] for five cycles (5 days
DSS and 16 days normal water) to develop the inflammation-induced colorectal
cancer mouse model [30]. The virtual MRI colonoscopy technique implemented
an enema procedure [31] using 1 mL of Fluorinert FC-770 (perfluorotri-n-butyl-
amine, molecular formula of C12F27N), which does not create a MRI signal due to
the lack of hydrogen atoms, and has been used in MRI imaging of human prostate
cancer. Prior to MRI imaging, the enema was administered with a 20-G gavage
syringe containing 0.6 mL of Fluorinert. The enema tubing was connected to a
syringe pump and maintained at a continuous rate of 25 μL/min to maintain the
colon distended during the imaging procedure. MRI (T1w and T2w) images were
acquired pre- and post-­ contrast (gadopentetate dimeglumine (Gd-DTPA),
0.2 mmol/kg, IV injection, 150 μL/min infusion). Figures 5.2 and 5.3 demonstrate
the utilization of a Fluorinert enema with MRI for virtual colonoscopy in the
development of an animal model assay platform for the study of inflamed
CRC. This standardized technique, incorporating both modality and animal
model, provides for an assay platform for the study of drug efficacy studies for
chronic inflammation (pre-CRC). In addition, the ability to fuse images from
other modalities to virtual MRC can further enhance the information on molecular
pathways in drug efficacy studies.
5 Small Animal Imaging in Oncology Drug Development 107

a b c d
T1w T1w
T2w T2w precontrast postcontrast

Healthy
mouse

Tumor-
bearing
mouse

Fig. 5.2 3.0 T MR images of the mouse colon of healthy (top) and tumor-bearing mice (bottom
row). (a) Coronal T2w image before Fluorinert enema infusion. (b–d) After Fluorinert enema:
T2w image (b); T1w pre-contrast (c); and T1w post-contrast (Gd-DTPA) image (d). Scale bar,
5 mm. Ileva L et al. Nature Protocols, (2014), 9(11), 178–2682–2692. DOI:https://doi.org/10.1038/
nprot.2014.178

5.3.1.2 Probe Validation: Nuclear Versus Optical Imaging

One major aspect for the development of an oncology drug as an in vivo assay is the
development of a relevant biomarker matched to an imaging capability. One such
drug, panitumumab (Vectibix), an anti-HER1 mAb, is a fully human mAb with
minimal immunogenicity when injected intravenously. It is FDA approved for the
treatment of HER1-expressing colorectal cancers, and is being evaluated in patients
with other types of HER1-expressing cancers, such as breast, lung, head and neck,
renal, and ovarian tumors [32]. The epidermal growth factor receptor (EGFR, erb1,
HER1) is a glycoprotein belonging to subclass I of the tyrosine kinase receptor
super family [33], disregulated in a variety of cancers [34], and is associated with
disease progression and treatment resistance. Panitumumab binds to the domain III
of HER1 and is rapidly internalized, leading to downregulation of cell surface
HER1. It also arrests the cell cycle and inhibits tumor growth by suppressing the
production of proangiogenic factors (VEGF, IL-8) by tumor cells [35].
To investigate a labeled panitumumab to risk-stratify clinical patients or as an
intraoperative diagnostic probe for image-guided surgery, various subcutaneous
108 J. D. Kalen and J. L. Tatum

DC SC R

T2W T1W precontrast T1W postcontrast

SC

DC

Fig. 5.3 The MRI sagittal colon plane (top) provides the relative positions in the transverse plane
for the rectum (R), sigmoid colon (SC) and descending colon (DC). The relative planes in the
transverse slices (bottom) are generated from the T1w and T2w coronal 3D images demonstrating
the Fluorinert enema and the enhancement of the lumen for studying inflammation in pre-CRC
drug studies. Ileva L et al. Nature Protocols, (2014), 9(11), 178–2682–2692. DOI:https://doi.
org/10.1038/nprot.2014.178

athymic nude female breast cancer xenograft tumor cell models have been devel-
oped with respect to HER1 expression (MDA-MB-469; high HER1, MDA-MB-231;
mid HER1, and BT-474; low HER1). Panitumumab can be dual-labeled with a fluo-
rescence dye (i.e., IRDye 800, optical component) and a radionuclide (i.e.,
111-indium, SPECT component and/or 89-zirconium, PET component) for testing
the various scenarios and models to visualize and quantify panitumumab uptake
into the tumor and organs. Dual labeling is more time-consuming and costly com-
pared to single labeling, with the caveat that quantified results of a single-labeled
drug can be compared (molecular probe uptake into the tumor and organs) between
modalities while utilizing standard animal handling techniques. Furthermore, dual
labeling of radionuclides (i.e., SPECT and PET) might not be feasible due to the
higher-energy PET photons (511 keV) penetration of the lower photon energy
SPECT collimators. Figure 5.4 demonstrates dual-modality PET/CT coronal slices
for the biodistribution of [89Zr] panitumumab in various tumor-bearing animal
5 Small Animal Imaging in Oncology Drug Development 109

a b c

5L

SUV (g/ml)
tumor
3

Cell line BT-474 MDA-MB-231 MDA-MB-468

HER1 level Very low Medium High

Fig. 5.4 Coronal slices demonstrating tumor uptake of 89Zr-panitumumab in various subcutane-
ous athymic nude female xenograft models; 10.18 ± 1.24 MBq of 89Zr-panitumumab were admin-
istered intravenously via tail vein, and a 5-min CT scan followed by a 30-min static PET scan was
performed at 96 h postinjection. The probe uptake into the tumor correlates with the HER1 protein
expression. Orange arrows point to the representative HER1 tumors. Sibaprasad Bhattacharyya
et al. Zirconium-89 labeled panitumumab: a potential immuno-PET probe for HER1-expressing
carcinomas. Nuclear Medicine and Biology, 2013, 40, 451–457, https://doi.org/10.1016/j.
nucmedbio.2013.01.007

models and correlation to the HER1 protein expression [36]. The same HER1
expression animal model(s) were later implemented in an epi-fluorescence imaging
study for the determination of [IRDye 800]-labeled panitumumab for an intraopera-
tive diagnostic probe for image-guided surgery [37]. Figure 5.5 demonstrates epi-­
fluorescence imaging for the HER1 animal models, resulting in similar probe uptake
into the tumor(s) with respect to the HER1 expression, and Fig. 5.6 exhibits the
correlation of the panitumumab imaging probe ([89Zr] and [IRDye 800]) uptake
between the different modalities. This case study demonstrates that providing a
modulated signal (i.e., HER1) will enable the quantitative investigation of the
underlining biomarker and that quantitation of the molecular biomarker with respect
to the modality (i.e., fluorescence or nuclear probe) provides for an accurate tech-
nique for comparison between modalities without the necessity of costly dual
labeling.

5.3.1.3 Multimodality Probe Development

Inhalation of asbestos fibers is the primary cause of malignant pleural mesothelium
(MPM), a highly lethal cancer affecting the lung pleural, and has been shown to
result in increased tissue HER1 expression [38]. X-ray CT and MRI have difficul-
ties distinguishing nonmalignant features of scarring and fibrosis from tumor tissue;
110 J. D. Kalen and J. L. Tatum

0.50
a b c
0.40

0.30

0.20

4.7 mm 4.7 mm 7.1 mm

0.10
MDA-MB-468 MDA-MB-231 BT-474

Fig. 5.5 2D epi-fluorescence images of panitumumab–IRDye800 of various HER1-expressing

tumor (MDA-MB-468, MDA-MB-231, BT474) bearing athymic nude female mouse models at
24 h postinjection (100 mL of 1 mg mL−1 conjugate). Red arrows point to the representative HER1
tumors. The probe uptake into the tumor correlates with the HER1 protein expression. Trace
amounts of tracer accumulated in ears probably due to the inflammation caused by ear punch.
Sibaprasad Bhattacharyya et al. Synthesis and biological evaluation of panitumumab–IRDye800
conjugate as a fluorescence imaging probe for EGFR-expressing cancers. Med. Chem. Commum.,
2014, DOI: https://doi.org/10.1039/c4md00116h

10.0
MDA-MB-468 (high HER1)
uptake

8.0 MDA-MB-231 (mid HER1)

BT-474 (Low HER1)
SUV (g/mL)
89Zr-panitumumab

6.0

4.0

2.0

0.0
0.0 5.0 10.0 15.0
Fluorescence Concentration (cts/energy/mm2)
Panitumumab-IRDye800 uptake

Fig. 5.6 Uptake of panitumumab labeled with IRDye800 or [89Zr] in different tumor xenografts
with high, medium, and low EGFR expression, as measured by radioactive counts or fluorescence,
is highly correlated. Sibaprasad Bhattacharyya et al. Synthesis and biological evaluation of panitu-
mumab–IRDye800 conjugate as a fluorescence imaging probe for EGFR-expressing cancers.
Med. Chem. Commum., 2014, DOI: https://doi.org/10.1039/c4md00116h

while the high specificity of [18F]FDG PET for tumor imaging is successful, unfor-
tunately high FDG uptake is also observed in benign inflammatory processes.
Unfortunately, the X-ray CT used in most preclinical scanners is designed for PET
photon attenuation correction and/or anatomical-functional image fusion that does
not provide for the ability to segment the various tissues (organs). Therefore, local-
ization of the molecular probe in the tumor is poorly differentiated from the sur-
rounding tissues, which can result in significant quantitative issues depending on
the preclinical animal model. Nyak et al. [39] studied MPM in an orthotopic (NCI-­
H226 and MSTO-211H mesothelium cells) MPM mouse model, fusing
5 Small Animal Imaging in Oncology Drug Development 111

CT+SPECT MRI SPECT+CT+MRI

Fig. 5.7 Representative coronal sections in female athymic (NCr) nu/nu mouse bearing ortho-
topic NCI-H226 cells injected intravenously via tail vein with 2.0 MBq of 111In-CHX-A”-DTPA–
panitumumab. Images were acquired 5 days after the injection of radiolabeled panitumumab.
Radiology 267, 2013: 173–182. DOI: https://doi.org/10.1148/radiol.12121021

panitumumab labeled with [111In] for SPECT/CT with anatomic images from a
3.0 T MRI, exhibited in Fig. 5.7, demonstrating the enhancement of a multimodality
study for both diagnostic and prognostic tools for the enhancement in the classifica-
tion and assessment of the patients’ disease state.

5.3.1.4 Case Studies Summary

The first examples (colorectal cancer and panitumumab) demonstrate the validation
strategy that coupling an imaging probe or technique with an imaging platform(s)
has the characteristics of an assay. Specifically, in the panitumumab case, providing
a modulated signal over the relevant biological scale will enable the quantitative
investigation of the underlining biomarker, which in this case is HER1. The last case
(multimodality probe development) demonstrates the enhancement of multimodali-
ties to improve tissue characterization by reducing the false-positive features of a
single modality. This case study further exemplifies in a multimodality study that
establishing SOPs for equipment QC, animal handling, anesthesia, and image quan-
titation that utilizes animal models that are validated with respect to pathological
standards is essential for the development of in vivo imaging assays.

5.3.1.5 Future
Multimodality preclinical imaging will be a requirement in oncology drug develop-
ment to understand the effect of treatment(s) on the various molecular pathways
transforming the tumor microenvironment. The incorporation of patient-derived
xenografts (PDX) into preclinical and co-clinical studies will correspondingly
require multimodalities due to the multi-scale in tumor heterogeneity.
112 J. D. Kalen and J. L. Tatum

To understand and analyze these spatially distinct regions within a tumor as

probed by multimodalities, investigators are incorporating texture analysis (fractals
and lacunarity) [40–42]. For example, Dominietto et al. evaluated pattern analysis
in a murine efficacy study investigating tumor angiogenesis [43]. MRI studies eval-
uated tumor blood volume and permeability at baseline and post-therapy. The
authors evaluated the MRI images utilizing both standard histogram analyses (aver-
age values within a region of interest (ROI)) and pattern analysis (shape and texture)
and concluded that the standard histogram was insensitive to determine therapeutic
response, while pattern analysis appeared to be sensitive to tumor textural changes
due to treatment. In addition to evaluating texture analysis in different modalities
for clinical and preclinical studies, other authors are evaluating the effect of recon-
struction algorithms on textural analysis, such as in PET imaging due to the limited
number of projections and higher noise component [44, 45].
To improve the integration of multimodalities, especially for co-clinical endeav-
ors, the National Cancer Informatics Program, which includes several NIH/National
Cancer Institute Divisions (Center for Biomedical Informatics and Information
Technology (CBIIT)), Cancer Imaging Program/Division of Cancer Treatment and
Diagnosis), and several academic and industry partners, implemented a working
group and developed a standard radiological image header (DICOM) for small ani-
mal imaging (Work Group 30: http://dicom.nema.org/dicom/geninfo/Strategy.pdf).
These standards provide a framework for quantitative comparison of multimodality
preclinical and clinical image sets utilizing identical image analysis algorithms.
These headers will also allow for the incorporation and fusion of ex vivo pathologi-
cal slides and molecular analysis, to include the full spectrum and multi-­scale
aspects for understanding the tumor microenvironment.
The future of multimodality imaging in small animals will provide an important
basis for in vivo assay development in oncology drug discovery, improving quanti-
tative measurement of therapeutic response, understanding the tumor microenviron-
ment, and the integration of emerging fields such as radiomics and radiogenomics
for improved patient outcomes.

Acknowledgments This project has been funded in whole or in part with federal funds
from the National Cancer Institute, National Institutes of Health, under Contract No.
HHSN261200800001E. The content of this publication does not necessarily reflect the views or
policies of the Department of Health and Human Services, nor does mention of trade names,
commercial products, or organizations imply endorsement by the U.S. Government Modified per
federal agency (NIH).
Frederick National Laboratory for Cancer Research is accredited by AAALAC International
and follows the Public Health Service Policy for the Care and Use of Laboratory Animals.
Animal care was provided in accordance with the procedures outlined in the “Guide for Care and
Use of Laboratory Animals” (National Research Council, 2011; National Academies Press,
Washington, D.C.).

References
1. DiMasi JA, Grabowski HG, Hansen RW. Innovation in the pharmaceutical industry:
new estimates of R&D costs. J Health Econ. 2016;47:20–33. https://doi.org/10.1016/j.
jhealeco.2016.01.012.
5 Small Animal Imaging in Oncology Drug Development 113

2. Thomas DW, Burns J, Audette J, Carroll A, et al. Clinical development success rates
2006–2015, BioIndustry analysis. http://www.amplion.com/clinical-development-success-
rates?hsCtaTracking=7e38cfe3-248d-440b-a7e4-c038acfa6eb2%7Ca6180579-5624-4deb-
ac76-35b512407bd1
3. Vanhove C, Bankstahl JP, Krämer SD, Visser E, Belcari N, Vandenberghe S. Accurate molecu-
lar imaging of small animals taking into account animal models, handling, anesthesia, quality
control and imaging system performance. EJNMMI Phys. 2015;2:31. https://doi.org/10.1186/
s40658-015-0135-y.
4. Kinahan P, Fletcher JW. PET/CT standardized uptake values (SUVs) in clinical practice and
assessing response to therapy. Semin Ultrasound CT MR. 2010;31(6):496–505. https://doi.
org/10.1053/j.sult.2010.10.001.
5. Sha W, Ye H, Iwamoto KS, Wong K-P, Wilks MQ, Stout D, McBride W, Huang S-C. Factors
affecting tumor 18F-FDG uptake in longitudinal mouse PET studies. EJNMMI Res. 2013;3:51.
https://doi.org/10.1186/2191-219X-3-51.
6. Adiseshaiah PP, Patel NL, Ileva LV, Kalen JD, Haines DC, McNeil SE. Longitudinal imaging
of cancer cell metastasis in two preclinical models: a correlation of noninvasive imaging to his-
topathology. Int J Molecul Imaging. 2014;2014:102702. https://doi.org/10.1155/2014/102702.
7. Fuchs K, Kukuk D, Mahling M, Quintanilla-Martinez L, Reischl G, Reutershan J, Lang F,
Rocken M, Pichler BJ, Kneilling M. Impact of anesthetics on 3′-[18F]fluoro-3′-deoxythymidine
([18F]FLT) uptake in animal models of cancer and inflammation. Mol Imaging. 2013:1–11.
https://doi.org/10.2310/7290.2012.00042.
8. Maier FC, Kneilling M, Reischl G, Cay F, Bukala D, Schmid A, Judenhofer MS, Röcken M,
Machulla H-J, Pichler BJ. Significant impact of different oxygen breathing conditions on non-
invasive in vivo tumor-hypoxia imaging using [18F]-fluoro-azomycinarabino- furanoside ([18F]
FAZA). Radiat Oncol. 2011;6:165. https://doi.org/10.1186/1748-717X-6-165.
9. Fueger BJ, Czernin J, Hildebrandt I, Tran C, Halpern BS, Stout D, Phelps ME, Weber
WA. Impact of animal handling on the results of 18F-FDG PET studies in mice. J Nucl Med.
2006;47(6):999–1006.
10. Fuchs K, Kukuk D, Reischl G, Foller M, Eichner M, Reutershan J, Lang F, Rocken M,
Pichler BJ, Kneilling M. Oxygen breathing affects 3′-deoxy-3’-18F-fluorothymidine uptake in
mouse models of arthritis and cancer. J Nucl Med. 2012;53:823–30. https://doi.org/10.2967/
jnumed.111.101808.
11. Hildebrandt IJ, Helen S, Weber WA. Anesthesia and other considerations for in vivo imaging
of small animals. ILAR. 2008;49(1):17–26. https://doi.org/10.1093/ilar.49.1.17.
12. Ileva LV, Bernardo M, Patel NL, Riffle LA, Graff-Cherry C, Robinson C, Difilippantonio S,
Kalen JD. Challenges in performing preclinical imaging in a large cohort therapeutic efficacy
study of murine cancer models. 64th AALAS National Meeting, Baltimore, MD, October 29,
2013.
13. Honndorf VS, Schmidt H, Wehrl HF, Wiehr S, Ehrlichmann W, Quintanilla-Martinez L,
Barjat H, Ricketts S-A, Pichler BJ. Quantitative correlation at the molecular level of tumor
response to docetaxel by multimodal diffusion-weighted magnetic resonance imaging
and [18F]FDG/[18F]FLT positron emission tomography. Mol Imaging. 2014;(1) https://doi.
org/10.2310/7290.2014.00045.
14. Yang H, Wang H, Shivalila CS, Cheng AW, Shi L, Jaenisch R. One-step generation of mice
carrying reporter and conditional alleles by CRISPR/cas-mediated genome engineering. Cell.
2013;154(6):1370–9. https://doi.org/10.1016/2013.08.022.
15. The Jackson Laboratory, Bar Harbor, ME USA, https://www.jax.org/.
16. Tentler JJ, Tan AC, Weekes CD, Jimeno A, Leong S, Pitts TM, Arcaroli JJ, Messersmith WA,
Gail Eckhardt S. Patient-derived tumor xenografts as models for oncology drug development.
Nat Rev Clin Oncol. 2012;9:338–50. https://doi.org/10.1038/nrclinonc.2012.61.
17. Elmore S. Apoptosis: a review of programmed cell death. Toxicol Pathol. 2007;35(4):495–516.
18. Biological Testing Branch, Division of Cancer Diagnostics and Treatment, NCI, NIH: https://
dtp.cancer.gov/organization/btb/default.htm
19. Center for Advanced Preclinical Research, Center for Cancer Research, NCI, NIH: https://ccr.
cancer.gov/capr
114 J. D. Kalen and J. L. Tatum

20. van Marion DMS, et al. Studying cancer metastasis: Existing models, challenges and
future perspectives. Crit Rev Oncol Hematol. 2015;97:107–17. https://doi.org/10.1016/j.
critrevonc.2015.08.00.
21. Chaffer CL, Weinberg RA. A perspective on cancer cell metastasis. Science.
2011;331(6024):1559–64. https://doi.org/10.1126/science.1203543.
22. Troy T, Jekic-McMullen D, Sambucetti L, Rice B. Quantitative comparison of the sensitivity
of detection of fluorescent and bioluminescence reporters in animal models. Mol Imaging.
2004;3(1):9–23.
23. Siolas D, Honnon GJ. Patient-derived tumor xenografts: transforming clinical samples
into mouse models. Cancer Res. 2013;73(17):5315–9. https://doi.org/10.1158/0008-5472.
CAN-13-1069.
24. Cassidy JW, Caldas C, Bruna A. Maintaining tumor heterogeneity in patient-derived
tumor xenografts. Cancer Res. 2015;75(15):2963–8. https://doi.org/10.1158/0008-5472.
CAN-15-0727.
25. American Cancer Society. Colorectal cancer. 2018.
26. Durkee BY, Weichert JP, Halberg RB. Small animal micro-CT colonography. Methods.
2010;50:36–41. https://doi.org/10.1016/j.ymeth.2009.07.008.
27. Boll H, Bag S, Nölte IS, Wilhelm T, Kramer M, Groden C, Böcker U, Brockmann MA. Double-­
contrast micro-CT colonoscopy in live mice. Int J Color Dis. 2011;26:721–7. https://doi.
org/10.1007/s00384-011-1181-0.
28. Larsson AE, et al. Magnetic resonance imaging of experimental mouse colitis and association
with inflammatory activity. Inflamm Bowel Dis. 2006;12:478–85.
29. Herborn CU, et al. Dark lumen magnetic resonance colonography in a rodent polyp model:
initial experience and demonstration of feasibility. Investig Radiol. 2004;39:723–7.
30. Ileva LV, Bernardo M, Young MR, Riffle LA, Tatum JL, Kalen JD, Choyke PL. In vivo MRI
virtual colonography in a mouse model of colon cancer. Nat Protoc. 2014;9(11):2682–92.
https://doi.org/10.1038/nprot.2014.178.
31. Young MR, Ileva LV, Bernardo M, Riffle LA, Jones YL, Kim YS, Colburn NH, Choyke
PL. Monitoring of tumor promotion and progression in a mouse model of inflammation-­
induced colon cancer with magnetic resonance colonography. Neoplasia. 2009;11(3):237–46.
https://doi.org/10.1593/neo.81326.
32. Wu M, Rivkin A, Pham T. Panitumumab: human monoclonal antibody against the epider-
mal growth factor receptors for the treatment of metastatic colorectal cancer. Clin Ther.
2008;30:14–30. https://doi.org/10.1016/j.clinthera.2008.01.014.
33. Burgess AW. EGFR family: structure physiology signaling and therapeutic targets. Growth
Factors. 2008;26:263–74. https://doi.org/10.1080/0897719080231284.
34. Ciardiello F, Tortora G. Anti-epidermal growth factor receptor drugs in cancer therapy. Expert
Opin Investig Drugs. 2002;11:755–68. https://doi.org/10.1517/13543784.11.6.755.
35. Yang XD, Xia XC, Corvalan JR, Wang P, Davis CG. Development of ABX-EGF, a fully
human anti-EGF receptor monoclonal antibody, for cancer therapy. Crit Rev Oncol Hematol.
2001;38:17–23. https://doi.org/10.1016/S1040-8428(00)00134-7.
36. Bhattacharyya S, Kurdziel K, Wei L, Riffle L, Kaur G, Hill GC, Jacobs PM, Tatum JL,
Dorosho JH, Kalen JD. Zirconium-89 labeled panitumumab: a potential immuno-PET probe
for HER1-expressing carcinomas. Nucl Med Biol. 2013;40:451–7. https://doi.org/10.1016/j.
nucmedbio.2013.01.007.
37. Bhattacharyya S, Patel NL, Wei L, Riffle LA, Kalen JD, Hill GC, Jacobs PM, Zinn KR,
Rosenthal E. Synthesis and biological evaluation of panitumumab–IRDye800 conjugate as
a fluorescence imaging probe for EGFR-expressing cancers. Med Chem Commum. 2014;
https://doi.org/10.1039/c4md00116h.
38. Faux SP, Houghton CE, Hubbard A, Pat- rick G. Increased expression of epidermal growth
factor receptor in rat pleural mesothelial cells correlates with carcinogenicity of mineral fibres.
Carcinogenesis. 2000;21(12):2275–80. https://doi.org/10.1093/carcin/21.12.2275.
39. Nayak TK, Bernardo M, Milenic DE, Choyke PL, Brechbiel MW. Orthotopic Pleural
Mesothelioma in Mice: SPECT/CT and MRI Imaging with HER1-and HER2-targeted
5 Small Animal Imaging in Oncology Drug Development 115

Radiolabeled Antibodies. Radiology. 2013;267:173–82. https://doi.org/10.1148/

radiol.12121021.
40. Asselin M-C, O’Connor JPB, Boellaard R, Thacker NA, Jackson A. Quantifying hetero-
geneity in human tumours using MRI and PET. Eur J Cancer. 2012;48:447–55. https://doi.
org/10.1016/j.ejca.2011.12.025.
41. Soares F, Janela F, Pereira M, Seabra J, Freire MM. 3D lacunarity in multifractal analysis of
breast tumor lesions in dynamic contrast-enhanced magnetic resonance imaging. IEEE Trans
Image Process. 2013;22(11):4422–35. https://doi.org/10.1109/TIP.2013.2273669.
42. Goh V, Sanghera B, Wellsted DM, Sundin J, Halligan S. Assessment of the spatial pattern of
colorectal tumor perfusion estimated at perfusion CT using two-dimensional fractal analysis.
Eur Radiol. 2009;19:1358–65. https://doi.org/10.1007/s00330-009-1304-y.
43. Dominietto M, Lehmann S, Keist R, Rudin M. Pattern analysis accounts for heterogeneity
observed in MRI studies of tumor angiogenesis. Magn Reson Med. 2013;70:1481–90. https://
doi.org/10.1002/mrm.24590.
44. Leijenaar RTH, Nalbantov G, Carvalho S, van Elmpt WJC, Troost EGC, Boellaard R,
Aerts HJWL, Gillies RJ, Lambin P. The efect of SUV discretization in quantitative FDG-­
PET Radiomics: the need for standardized methodology in tumor texture analysis. Sci Rep.
2015;5:11075. https://doi.org/10.1038/srep11075.
45. Buvat I, Orlhac F, Soussan M. J Nucl Med. 2015;56(11):1642–4. https://doi.org/10.2967/
jnumed.115.163469.
Investigation of Transporter-Mediated
Drug-Drug Interactions Using PET/MRI 6
Thomas Wanek, Alexander Traxl, Claudia Kuntner-Hannes,
and Oliver Langer

6.1 Introduction

Drug disposition consists of absorption, distribution, metabolism and excretion

(ADME). For these processes, drugs and drug metabolites have to cross cellular
membranes in different organs and tissues (e.g. intestine, liver, kidney, etc.). In
many cases, movement of drugs and drug metabolites across cellular membranes is
not just a passive diffusion-mediated process but occurs via saturable transmem-
brane transporters belonging either to the solute carrier (SLC) or adenosine triphos-
phate-binding cassette (ABC) families. When two drugs, which are recognized by
the same SLC or ABC transporters, are co-administered, one drug may induce
transporter inhibition/saturation and thereby change the disposition of the other
drug as compared to when the drugs are administered alone. This phenomenon has
been termed transporter-mediated drug-drug interaction (DDI), which is of great
concern in drug development as this may impact drug safety and efficacy [1]. In
many cases, changes in the activity of transmembrane transporters may lead to

T. Wanek (*)
AIT Austrian Institute of Technology GmbH, Center for Health & Bioresources,
Biomedical Systems, Seibersdorf, Austria
BioImaging Austria (CMI), Seibersdorf, Austria
e-mail: [email protected]
A. Traxl · C. Kuntner-Hannes
AIT Austrian Institute of Technology GmbH, Center for Health & Bioresources,
Biomedical Systems, Seibersdorf, Austria
O. Langer
AIT Austrian Institute of Technology GmbH, Center for Health & Bioresources,
Biomedical Systems, Seibersdorf, Austria
Departments of Clinical Pharmacology and Biomedical Imaging and Image-Guided Therapy,
Medical University of Vienna, Vienna, Austria

© Springer Nature Switzerland AG 2019 117

C. Kuntner-Hannes, Y. Haemisch (eds.), Image Fusion in Preclinical
Applications, https://doi.org/10.1007/978-3-030-02973-9_6
118 T. Wanek et al.

pronounced changes in drug tissue distribution (e.g. liver, kidneys, brain) without
changes in drug plasma pharmacokinetics [2]. To assess such tissue DDIs in vivo, a
methodology is needed to measure drug tissue concentration levels. The non-inva-
sive nuclear imaging method positron emission tomography (PET) is a very power-
ful tool for measuring tissue distribution of drugs radiolabelled with positron-emitting
radionuclides, such as carbon-11 (11C, half-life: 20.4 min) or fluorine-18 (18F, half-
life: 109.8 min), in animals or humans [3, 4]. Dedicated small-animal PET systems
allow to measure rodents (mice, rats) with high sensitivity and good spatial resolu-
tion. As PET does not provide anatomical information, it needs to be combined with
anatomical imaging to allow for better definition of organs or tissues of interest. In
this chapter, we provide a case report of how PET in combination with magnetic
resonance imaging (MRI) can be used to assess transporter-mediated DDIs in dif-
ferent organs of the mouse.

6.2 Sequential PET/MR Imaging

PET is a nuclear imaging technique, which allows the non-invasive determination of

the concentration of a radiolabelled drug in tissues. PET utilizes the radioactive
decay of positron-emitting radionuclides such as 18F or 11C to track the localization
of a radiolabelled drug in the body. Besides its non-invasive nature, the advantages
of PET are its high sensitivity (only a few μg of the radiolabelled compound is
required) and its high temporal resolution enabling the detection of concentration
changes in tissues within several seconds. However, PET has a limited spatial reso-
lution (~2 mm) and provides only limited anatomical information, which may make
it difficult to allocate the PET signal to designated organs or tissues. Contrary to
PET, MRI delivers anatomical information with high spatial resolution and excel-
lent soft-tissue contrast. Especially the organs involved in elimination processes of
drugs, such as the liver, gall bladder and kidneys, can be clearly visualized by MRI
without the use of additional contrast agents. Therefore, MRI adds truly comple-
mentary information to PET data by facilitating the definition of these organs when
used in a combined way.
Combining PET and MRI can be achieved by either software fusion or hardware
combination. Several combined or simultaneous preclinical PET/MRI systems are
now commercially available [5, 6]. PET/MRI images can also be achieved by scan-
ning the subjects on two stand-alone scanners in a serial manner and subsequent
image fusion. This is usually performed using software tools that load the image
data of both modalities and either use an automated or manually adjusted rigid or
non-rigid transformation matrix. Image fusion is simplified when fiducial markers
or anatomical landmarks are identified. In contrast to clinical imaging, where the
patient has to be moved between the two scanners, often resulting in different patient
positions on the bed, this is not an issue in preclinical imaging. Here, not the animal
alone but the whole imaging chamber (including the retained and anaesthetized
animal) can be moved from one scanner to the other. Thus, subject position usually
remains the same within the chamber, which simplifies image fusion as only
6 Investigation of Transporter-Mediated Drug-Drug Interactions Using PET/MRI 119

rotation, and translation can occur. Especially for benchtop MRI systems with low-
field strength and permanent magnets, this combination is ideal, as they can be
installed almost anywhere and at relatively low costs and due to the low-field
strength do not interfere with photomultiplier tube (PMT)-based PET scanners.
Placing them near dedicated small-animal PET scanners, these benchtop MRI sys-
tems enable sequential PET/MRI acquisitions, combining the advantages of both
modalities and thus providing results comparable to a full-scale sequential PET and
a high-field-strength MR measurement, with few restrictions in terms of MR image
quality and animal handling. Another advantage of using two separate scanners is
that both scanners can be utilized alone, especially for studies when no multimodal
image information is required, i.e. only PET or only MRI scans. Moreover, the
maintenance costs and technical challenges are lower for stand-alone systems com-
pared to combined scanners.
A prerequisite for sequential PET/MR imaging is a multimodality imaging
chamber, which provides heating, inhalation anaesthesia and animal monitoring
(i.e. temperature, respiratory and heart rate). Moving the animal from one bed to the
next results in changes in animal position, which complicates image fusion. To
overcome this issue, one can either use a multimodality imaging chamber with
proper connectors on each modality (also providing anaesthesia gas support, heat-
ing and animal monitoring) or use the imaging chamber from one modality in all
other modalities, by means of adapters. Benchtop MRI systems often have an ani-
mal bed where the RF coil is directly mounted over the animal, and the combination
is slid into the scanner. Thus, the construction of an adapter and mounting plate for
the PET scanner is necessary.
Another premise for sequential PET/MR imaging is the correct and precise co-
registration of the two data sets. The used methods and algorithms have to be practi-
cal but also robust so that they can be used on a daily basis. One of the main
challenges is that for certain radiotracers PET images show no or little anatomical
information. As a result, it may be difficult to co-register them with MR images just
based on morphological structures, which can be seen in both images. Furthermore,
new radiotracers for PET examinations become more and more specific in terms of
their interaction with the molecular imaging target, so that they show little retention
in nontarget tissues (e.g. radiolabelled antibodies). As mentioned before, apart from
anatomical landmarks, fiducial markers can be used to simplify image fusion.
Fiducial markers are small objects that are visible in the used modality, in this case
PET and MRI. Small glass capillaries (inner diameters of 0.1–0.5 mm) are typically
used that can be filled with radioactive solution and sealed with adhesive. At least
three markers are then attached to or inside the animal chamber next to the animal,
in such a way that they are visible in the final images of all modalities (means for
MRI that they have to be inside the RF coil) but do not interfere with the imaging
study. The drawback of using these markers is that they increase the time for animal
positioning and, even if they are small, may complicate positioning of the animals
inside the imaging chamber. Especially for the animal holder used in the benchtop
1T MRI system, space is very limited as the whole body RF coil is positioned very
close over the animal bed and animal, already restricting the size of the scanned
120 T. Wanek et al.

animal. Another method to generate a transformation matrix for image fusion is

phantom measurements using a multimodality phantom that incorporates landmarks
(e.g. intersection points) that are visible in all modalities. When imaging the phan-
tom with the identical protocol as used for the animal studies (same acquisition and
reconstruction protocol), it is possible to generate a rigid transformation matrix that
can afterwards be applied to the animal data.

6.3 Role of Transporters in Drug Disposition

Transport of drugs can be mediated by members of two large families of transmem-

brane transporter proteins, the SLC and the ABC transporter family. Both trans-
porter families use cellular energy, either in form of electrochemical gradients or
ATP hydrolysis to transport compounds against concentration gradients. These
pumps can move substrates in (influx) or out (efflux) of cells. In humans and rodents,
ABC and SLC transporters are for instance expressed in the liver, intestine, blood-
brain barrier (BBB), blood-testis barrier, placenta and kidney. For instance, in the
liver organic anion-transporting polypeptides (e.g. OATP1B1/SLCO1B1, OATP1B3/
SLCO1B3) in the basolateral membrane of hepatocytes mediate the uptake of drugs
from blood in the liver, whereas canalicular ABC transporters (e.g. multidrug resis-
tance-associated protein 2, MRP2/ABCC2 or breast cancer resistance protein,
BCRP/ABCG2) can promote the excretion of drugs and drug metabolites from
hepatocytes into bile [7]. Impaired function of basolateral uptake transporters due to
DDIs, disease or genetic polymorphisms can lead to marked increases in blood
concentrations of drugs, which can cause severe side effects (e.g. rhabdomyolysis
for statin drugs). Impaired function of canalicular efflux transporters, on the other
hand, can lead to drug accumulation in the liver, which may cause hepatotoxicity.
Erlotinib is a reversible tyrosine kinase inhibitor (TKI) of the epidermal growth
factor receptor (EGFR) that has been approved for treatment of advanced, meta-
static non-small cell lung cancer (NSCLC) and advanced, unresectable or metastatic
pancreatic cancer. Approximately 10% of NSCLC patients in the Western popula-
tion harbour an activating mutation in their EGFR genes (e.g. the exon 19 deletion
delE746-A750 or the exon 21 point mutation L858R) resulting in higher response
rates to treatment with erlotinib or gefitinib, another EGFR-inhibiting TKI, as com-
pared with patients with wild-type EGFR [8].
Erlotinib undergoes extensive metabolism in humans and is mainly excreted via
the hepatobiliary pathway leading to high concentration levels in the liver and bile
[9]. Erlotinib is a substrate of BCRP, P-glycoprotein (Pgp/ABCB1), organic anion
transporter 3 (OAT3/SLC22A8) and organic cation transporter 2 (OCT2/SLC22A2).
Efflux transport by P-gp and BCRP at the BBB results in a low extent of brain dis-
tribution of erlotinib [10–12]. Less information is currently available about how far
drug transporters influence distribution of erlotinib to other organs than the brain
and to what extent transporters are involved in erlotinib excretion.
In this study, we assessed the influence of transporters on 11C-erlotinib organ
distribution and excretion. To accomplish this, we performed sequential MRI/PET
6 Investigation of Transporter-Mediated Drug-Drug Interactions Using PET/MRI 121

measurements with a microdose of 11C-erlotinib (<5 μg) without and with the co-
injection of a pharmacologic dose of unlabelled erlotinib or pretreatment with elac-
ridar, a dual P-gp/BCRP inhibitor, in wild-type mice. In addition, we scanned P-gp/
Bcrp knockout (Abcb1a/b(−/−)Abcg2(−/−)) mice. The recorded images were co-regis-
tered, and organ concentration levels were further quantitated in order to obtain
pharmacokinetic data. The technical challenges of this project lie in the use of two
different scanner systems and the development of devices and procedures which
allow a rapid and secure exchange of the animals from one scanner setup to the
other including anaesthesia, animal heating and monitoring systems. In order to
obtain images suitable for image co-registration, animals have to be moved from
one scanner to the other in a way that no body movement of the animal occurs. In
addition, for optimal image co-registration, a transformation matrix to correct for
the position mismatch of the reconstructed MRI and PET images has to be
calculated.

6.4 Material and Methods

6.4.1 Phantom Study

For calculation of a transformation matrix for image co-registration, a multimodal-

ity phantom was constructed. This phantom consists of a polypropylene cylinder
filled with a mixture of H2O, CuSO4 (1 g/L) and NaCl (3.6 g/L) in which a flexible
tube containing 18F-fluoride solution was inserted. The tube was positioned inside
the phantom in such a way that intersection points at different planes were created.
These intersection points were used as markers to define the position of the phantom
in the MRI or PET scanner. In total, the measurement phantom contained 15 inter-
section points. The last one was located 42 mm away from the first intersection
point. Every kind of intersection point appeared three times. The phantom was posi-
tioned on the MRI animal bed with the whole body mouse coil, and images were
acquired with the PET and the MR using the same acquisition and reconstruction
protocols as used for the in vivo study. To co-register the phantom PET and MR
images, the data set alignment wizard in AMIDE [13] was used, which allows the
alignment of one image data set with another. For alignment, the rigid body registra-
tion method using fiducial markers was selected. Before the data alignment wizard
could be used, the fiducial markers were defined on the PET and MR images at the
positions of the intersection points. Transverse, coronal and sagittal views of the
data set in AMIDE were used to identify the correct position of the fiducial markers
as good as possible. For every fiducial marker, the centre location was documented
before and after the PET, and MRI data sets were aligned. Afterwards the alignment
was calculated by AMIDE using the Procrustes rigid body alignment algorithm
without scaling. Position mismatch was calculated by alignment of the intersection
points of the flexible tubes within the phantom in the MR and PET images resulting
in a rigid transformation matrix that can be applied to all data sets. Here, the PET
data set was defined as the fixed and the MR data set as the moving one, because the
122 T. Wanek et al.

moving data set is interpolated after the transformations were applied. This should
be avoided for the PET data set, as it is used for further analysis of the drug distribu-
tion data and interpolation could lead to falsified results. Evaluation of image co-
registration accuracy was performed by calculation of the fiducial registration error
(FRE) and the target registration error (TRE).

6.4.2 Animals

Female FVB and Abcb1a/b(−/−)Abcg2(−/−) mice were obtained from Charles River
and Taconic. At the time of experiment, animals were 10–15 weeks old and weighed
25.7 ± 2.7 g. Animals were housed in groups in individual ventilated cages (IVCs)
under controlled environmental conditions. Animals were allowed to acclimatize
>1 week prior to the experiments. The study was approved by the local Animal
Welfare Committee in accordance with the Austrian Animal Experiments Act. All
efforts were made to minimize the number of animals as well as pain or
discomfort.

6.4.3 Anaesthesia and Monitoring

Anaesthesia was initiated using 2.5–3% isoflurane in oxygen and maintained at

1–1.5% for the imaging procedure. Body temperature and respiratory rate were con-
stantly monitored during the whole experiment using an MR compatible small-ani-
mal monitoring and gating system. Animals were positioned prone on the mouse
bed, front teeth were positioned inside the tooth bar and body temperature was
maintained by warm water heating.

6.4.4 Experimental Design

Wild-type animals were divided into three groups: In the first group (n = 4), a
dynamic 11C-erlotinib PET scan was performed without any pretreatment. The sec-
ond group of mice (n = 6) underwent dynamic 11C-erlotinib PET scans at 20 min
after intravenous pretreatment with the dual P-gp/BCRP inhibitor elacridar (10 mg/
kg). In the third group, mice (n = 4) underwent dynamic 11C-erlotinib PET scans, in
which a pharmacologic dose of unlabelled erlotinib (10 mg/kg) was co-injected
with 11C-erlotinib. In addition, Abcb1a/b(−/−)Abcg2(−/−) mice (n = 4) underwent a
dynamic 11C-erlotinib PET scan without any pretreatment.

6.4.5 MRI

Anatomic MR imaging was performed using a benchtop MRI scanner (ICON 1 T,

Bruker BioSpin GmbH, Ettlingen, Germany). The 1T MRI is equipped with a 1T
6 Investigation of Transporter-Mediated Drug-Drug Interactions Using PET/MRI 123

permanent magnet with negligible fringe field. Its gradient coil provides 450 mT/m.
The animal bed is equipped with a nose-cone for inhalation anaesthesia, fixation
system (tooth bar), heating system (built-in water tubing in the animal bed), tem-
perature and respiration monitoring. After positioning the animal on the animal bed
and connection of all cables, the mouse whole body RF coil (length 80 mm) was slid
over the animal and animal bed. Images were acquired using a modified three-
dimensional T1-weighted gradient echo sequence (T1-fast low-angle shot) with the
following parameters: echo time = 5 ms; repetition time = 25 ms; flip angle = 25°;
field of view = 76 × 28 × 24 mm; matrix = 253 × 93; 32 slices; slice thick-
ness = 0.75 mm; and scan time = 6.25 min.

6.4.6 PET

After MRI, the animal holder was transferred into the gantry of a Siemens microPET
Focus220 system and secured using a custom-made mounting plate. A 10-min
transmission scan using a 57Co point source was followed by a 90-min emission
scan which was started at time of intravenous injection of 11C-erlotinib (26 ± 9 MBq,
1 ± 1 nmol, 0.10 mL, n = 18). The scanner acquires the data in list mode format with
a 6 ns coincidence timing window and an energy window of 250–750 keV. The
reported spatial image resolution was 1.8 mm FWHM within 1 cm radial offset and
the detection sensitivity was 3% [14]. The PET emission data were sorted into 25
frames, which incrementally increased in time length from 5 s to 20 min. Images
were reconstructed using Fourier rebinning followed by two-dimensional filtered
backprojection. The reconstructed PET images consist of a 128 × 128 × 95 matrix
with a final voxel size of 0.4 × 0.4 × 0.8 mm3. The standard data correction protocol
(normalization, decay correction, injection decay correction and attenuation correc-
tion) was applied to all data sets.

6.4.7 Image Analysis

Images were co-registered using the medical data examiner software AMIDE [13].
After applying the transformation matrix calculated from the phantom study to the
animal data, volumes of interest (VOIs) were manually drawn over the whole brain,
right lung, left ventricle of the heart, left kidney, liver, gallbladder, intestine and
urinary bladder in the fused MR/PET images, and time-activity concentration
curves expressed in standardized uptake values (SUV) from 0 to 90 min after injec-
tion of the radiolabelled compound were derived. It was assumed that the sum of
radioactivity in the gallbladder and the intestine represented radioactivity in the bile
excreted from the liver. From the time-activity curves, the area under the curve from
time 0 to 90 min (AUC) was calculated using Prism 5.0 software (GraphPad
Software).
Furthermore, a graphical analysis approach (integration plot) was used to esti-
mate the rate constants for cerebral, hepatic, renal and pulmonary uptake (kuptake, brain,
124 T. Wanek et al.

kuptake, liver, kuptake, kidney and kuptake, lung) and the rate constants for biliary and urinary
excretion (kbile, kurine) of 11C-erlotinib. Rate constants for uptake were measured from
0.3 to 3.5 min after tracer injection for the brain, liver and kidney and from 0.6 to
4.5 min after tracer injection for the lung using the integration plot method [15] and
the following equation:

X t , brain AUC0 -t Ct , organ AUC0 -t , blood

= CL uptake , brain ´ + VE , brain = kuptake ´ + VE
Ct , blood Ct , blood Ct , blood Ct , blood

where Ct, organ is the radioactivity concentration in the brain, liver, kidney or lung at
time t and Ct, blood is the radioactivity concentration in the left ventricle of the heart
at time t. AUC0-t, blood represents the area under the concentration-time curve in the
left ventricle of the heart from time 0 to time t. Kuptake can be obtained by performing
linear regression analysis of a plot of Ct, organ/Ct, blood versus AUC0-t, blood/Ct, blood and
calculating the slope of the regression line. The unit of kuptake is mL blood per min
per gram tissue (mL/min/g tissue), and it thus corresponds to K1 from kinetic model-
ling of PET data [16]. VE is the y-intercept of the integration plot.
Rate constants for biliary and urinary excretion (kbile and kurine) of 11C-erlotinib
were measured from 8.8 to 65 min and 12.5 to 65 min after tracer injection, respec-
tively, using the integration plot method and the following equations:

Ct ,intestine = kbile ´ AUC0 -t , liver + VE

and
Ct , urine = kurine ´ AUC0 -t , kidney + VE

where Ct, intestine and Ct, urine are the radioactivity concentrations in the intestine
(including the gall bladder and duodenum) and urine, respectively, at time t. AUC0-t,
liver and AUC0-t, kidney represent the area under the concentration-time curve in the liver
and the kidney, respectively, from time 0 to time t. kbile and kurine can be obtained by
performing linear regression analysis of a plot of Ct, intestine versus AUC0-t, liver and Ct,
urine versus AUC0-t, kidney, respectively, and calculating the slope of the regression line.
VE is the y-intercept of the integration plot.

6.5 Results and Discussion

Both employed imaging systems are located in the same room within close proxim-
ity (~3 m) enabling a quick transfer of animals from one scanner system to the other.
The external stray field of the 1T MRI system is negligible and therefore no RF cage
is needed. The integration of anaesthesia, animal warming and respiratory trigger in
the MRI animal holder allows switching between both scanner systems without
interruption of the entire animal handling systems. Hence, anaesthesia and warm
water for temperature stabilization were constantly supplied, and animal vital
parameters were monitored. All PET scans were performed after MRI with the
6 Investigation of Transporter-Mediated Drug-Drug Interactions Using PET/MRI 125

mouse bed and mounted mouse whole body coil. This setup ensures that movement
of the scanned subject is limited; however, the mounted coil increased attenuation
and scatter for the PET measurements which has to be corrected by performing a
transmission scan prior to the PET measurement. Including the attenuation informa-
tion into PET image reconstruction leads to the same quantitative data as with PET-
only scans [17].
With the aim to safely and reproducibly mount the MRI animal holder onto the
PET system, a mounting plate was constructed that allows exact positioning of the
completely assembled animal holder onto the PET system. This mounting plate
consists of a bottom plate which exactly fits to the bed moving mechanism of the
PET scanner, with a fixable clamp to secure the animal holder and a mounting guide
to prevent any rotation of the animal holder and four adjustable screws (Fig. 6.1a).
The mounting plate can be easily removed from the PET scanner if not needed. By

a b

c d

Fig. 6.1 (a) Custom-made mounting plate for reproducible positioning of the Bruker animal
holder on the PET system. (b) Bruker animal bed holder with mouse bed and mouse whole body
coil mounted on a Siemens Focus220 microPET system. The animal bed holder can be moved
from the ICON MRI to the PET system within seconds without interruption of anesthesia or ani-
mal warming. The employed system ensures that the animal position in the PET system corre-
sponds exactly to the position in the previous MRI scan. (c) Home-made position finder for
reproducible positioning of the Bruker animal holder in the MRI system. (d) Bruker animal bed
holder with mouse bed and mouse whole body coil inserted into the ICON MRI
126 T. Wanek et al.

using this setup, the animal bed holder could be transferred from the MRI to the
PET scanner within seconds without interruption of anaesthesia, animal warming or
monitoring systems. Using the same horizontal and vertical positioning parameters
of the bed moving mechanism of the PET scanner, a reproducible position inside the
centre field of view could be achieved for all conducted experiments (Fig. 6.1b). To
ensure a reproducible position of the animal bed inside the MRI scanner, a position
finder was designed that can be easily attached to the Bruker ICON (Fig. 6.1c, d).
All MRI and attenuation measurements could be performed within the 20-min pre-
treatment phase with elacridar prior to injection of 11C-erlotinib.
Figure 6.2a shows the mouse whole body PET/MRI phantom that was con-
structed to calculate a transformation matrix for image co-registration of corre-
sponding MRI and PET images. All 15 aligned intersections were clearly visible in

Fig. 6.2 (a) Mouse whole body phantom used for the calculation of a transformation matrix to
correct for images mismatch in fused MR and PET images. The phantom consists of a 15 mL
polypropylene falcon tube filled with a solution of CuSO4 and NaCl in water. In this falcon tube
holes were drilled and a thin, flexible tube (inner diameter 0.8 mm) was threaded in such a way that
it forms a series of intersections in all anatomical orientations. Prior to measurements the flexible
tube was filled with a solution containing 18F-fluoride and the end of the flexible tube was secured
with screw caps. (b) Transversal PET (left) and T1-weighted gradient echo MRI (middle) views of
the mouse whole body phantom acquired using the Bruker animal holder equipped with the mouse
whole body coil. The flexible tube filled with 18F-fluoride solution can be clearly distinguished
from the background of the MRI phantom in both imaging modalities. A transformation matrix
was calculated by placing fiducial markers at the 5 intersections visible in the PET and MR images
and further rotation and transformation of one dataset to exactly match the position of the fiducial
markers (right)
6 Investigation of Transporter-Mediated Drug-Drug Interactions Using PET/MRI 127

all MRI and PET images, which allowed placement of sufficient fiducial markers
from which the position mismatch between reconstructed MRI and PET images
could be calculated (Fig. 6.2b). The final calculated mean registration error from the
centre location between both reconstructed MRI and PET images was <0.2 mm
(FREmean = 0.18 ± 0.02 mm and TREmean = 0.18 ± 0.02 mm). Deviations caused by
rotation errors were almost negligible.
Quantitative data of organ concentration of 11C-erlotinib were derived after
image co-registration of corresponding MRI and PET images. The transformation
matrix calculated from the phantom studies could be used for all scanned animals
without further modifications.
We assessed the effect of transporters on tissue distribution and excretion of
11
C-erlotinib. In Fig. 6.3 representative co-registered PET/MR images and in Fig. 6.4

a L
I

UB 15

SUV

1
d

7.5 - 10 min 20 - 30 min 50 - 60 min 70 - 90 min

Fig. 6.3 Serial coronal whole-body PET/MR images of 11C-erlotinib in a wild-type mouse (a), a
wild-type mouse pretreated at 20 min before PET with 10 mg/kg elacridar (b), a wild-type mouse
co-injected with a pharmacologic dose (10 mg/kg) of erlotinib (c), and an Abcb1a/b(−/−)Abcg2(−/−)
mouse. Anatomical structures are indicated by arrows: L = liver, I = intestine; UB = urinary
bladder
128 T. Wanek et al.

a 12 b 8.0
WT
10 WTelacr
6.0

Kidney (SUV)
Liver (SUV)

8 WTerlot

Abcb1a/b(-/-)Abcg2(-/-)
6 4.0
4
2.0
2
0 0.0
0 20 40 60 80 0 20 40 60 80
Time (min) Time (min)
c d
12 60

Urinary bladder (SUV)

10 50
Intestine (SUV)

8 40
6 30
4 20
2 10
0 0
0 20 40 60 80 0 20 40 60 80
Time (min) Time (min)

Fig. 6.4 Time-activity curves (mean SUV ± SD) of 11C-erlotinib in (a) liver, (b) kidney, (c) intes-
tine and (d) urinary bladder of wild-type mice (WT), wild-type mice pretreated at 20 min before
PET with 10 mg/kg elacridar (WTelacr), wild-type mice co-injected with a pharmacologic dose
(10 mg/kg) of erlotinib (WTerlot), and Abcb1a/b(−/−)Abcg2(−/−) mice

time-activity curves in different organs are shown. There was a clear visual differ-
ence in the biodistribution of 11C-erlotinib between wild-type and Abcb1a/
b(−/−)Abcg2(−/−) mice, in that there was prolonged retention of radioactivity in the
liver of Abcb1a/b(−/−)Abcg2(−/−) mice with less radioactivity in the intestine and
more radioactivity in the urinary bladder (Figs. 6.3 and 6.4). Also, in erlotinib co-
injected wild-type mice, a prolonged liver retention and less radioactivity content in
the intestine were observed. Blood radioactivity concentrations were significantly
higher in erlotinib co-injected mice (AUC in left ventricle of the heart, wild-type
microdose: 91.0 ± 2.4 SUV*min, pharmacologic dose: 135.8 ± 13.3 SUV*min).
Integration plot analysis revealed significantly lower kbile values in Abcb1a/
b(−/−)Abcg2(−/−) mice and in erlotinib co-injected wild-type mice as compared with
wild-type mice which received a microdose of 11C-erlotinib (Fig. 6.5d). kurine values
were significantly increased in Abcb1a/b(−/−)Abcg2(−/−) mice (Fig. 6.5e). Taken
together this suggests that Bcrp and P-gp at the canalicular membrane of hepato-
cytes mediated biliary excretion of 11C-erlotinib and/or its radiolabelled metabo-
lites. These transporters appear to become saturated when a pharmacologic dose of
erlotinib was co-injected leading to a reduction in kbile. Interestingly, the P-gp/BCRP
inhibitor elacridar appeared to be not able to inhibit hepatic P-gp/Bcrp (Fig. 6.5d).
 a 1.0 b 0.5 c 0.10
6

0.8 0.4 0.08

0.6 0.3 0.06

0.4 0.2 0.04

Kuptake,liver
Kuptake,lung

Kuptake,kidney

(mL/min/g tissue)
(mL/min/g tissue)
0.2 0.1 0.02

(mL/min/g tissue)
0.0 0.0 0.00
c r ot /-) c r o t /-) r ot /-)
WT ela T erl 2(- WT ela erl 2(- WT e lac T erl 2(-
WT W c g WT WT g WT W c g
Ab )A bc b
-/-) (-/- )/A
/b( /b / b(-/-
1a 1a 1a
cb cb c b
Ab Ab Ab
d 0.05 e 0.5
0.4
0.04
0.3
0.03
0.2

Kbile
Kurine
0.02
0.1

(mL/min/g tissue)

(mL/min/g tissue)
0.01 0.0
0.00 -0.1
cr ot ) r ot /-)
WT ela erl (-/- WT e lac erl 2(-
WT WT c g2 WT WT g
Ab /Abc
-/-) (-/-)
a /b( a/b
c b1 1
cb
Investigation of Transporter-Mediated Drug-Drug Interactions Using PET/MRI

Ab Ab

Fig. 6.5 Rate constants for (a) hepatic, (b) renal, and (c) pulmonary uptake (mean kuptake,organ ± SD) as well as for (d) biliary and (e) urinary excretion (mean kfluid
± SD) of 11C-erlotinib in wild-type mice (WT), wild-type mice pretreated at 20 min before PET with 10 mg/kg elacridar (WTelacr), wild-type mice co-injected
129

with a pharmacologic dose (10 mg/kg) of erlotinib (WTerlot), and Abcb1a/b(−/−)Abcg2(−/−) mice. ***P < 0.001, 1-way ANOVA with Bonferroni’s multiple com-
parison test
130 T. Wanek et al.

The increase in kurine values in Abcb1a/b(−/−)Abcg2(−/−) mice was consistent with a

shift from hepatobiliary to renal excretion in absence of canalicular Bcrp/P-gp
activity (Figs. 6.4d and 6.5e).
In the brain, AUCs and kuptake, brain values were significantly higher in elacridar-
treated wild-type mice as well as in Abcb1a/b(−/−)Abcg2(−/−) mice (Fig. 6.6),
which confirmed that P-gp and Bcrp restricted brain distribution of 11C-erlotinib
and that elacridar can completely inhibit P-gp/Bcrp at the mouse
BBB. Interestingly, also co-injection of unlabelled erlotinib increased kuptake, brain,
which indicated partial transporter saturation at the BBB and which suggests a
potential utility of high-dose erlotinib to enhance brain uptake of other dual
P-gp/BCRP substrates.
Integration plot analyses were conducted to obtain the rate constants for uptake
of 11C-erlotinib from blood into the liver, kidney and lung (Fig. 6.5). In the liver and
kidney, kuptake was significantly lower for the pharmacologic dose than for the micro-
dose, consistent with saturation of basolateral uptake transporters in hepatocytes
and kidney cells, such as OATPs or organic anion transporters. This is in line with
previous findings that TKIs including erlotinib are competitive inhibitors of these
SLC transporters [18]. This most likely also accounted for the significantly higher
blood AUCs for the pharmacologic dose as compared with the microdose. In the
lungs, kuptake was significantly increased in animals receiving the pharmacologic
dose (Fig. 6.5a–c).
In other organs than the brain and the lungs, elacridar pretreatment exerted no
significant effect on 11C-erlotinib distribution. In the liver, there was a trend for
decreased kuptake values following elacridar, which indicates that elacridar may
inhibit erlotinib uptake transporters in hepatocytes.
An advantage of performing transporter studies in mice is the fact that the whole
animal is within the field of view of the PET scanner so that all mouse organs can

a 1.6 b 0.16
WT
WTelacr
(nL/min/g tissue)

1.2 0.12
Brain (SUV)

WTerlot
Kuptake,brain

Abcb1a/b(-/-)Abcg2(-/-)
0.8 0.08

0.4 0.04

0.0 0.00
)
0 20 40 60 80 (-/-
WT ela
cr
erl
ot
g2
Time (min) WT WT bc
(-/- A
)
/b
1a
cb
Ab

Fig. 6.6 (a) Time-activity curves (mean SUV ± SD) and (b) rate constants for whole brain uptake
(mean kuptake,brain ± SD) of 11C-erlotinib in wild-type mice (WT), wild-type mice pretreated at 20 min
before PET with 10 mg/kg elacridar (WTelacr), wild-type mice co-injected with a pharmacologic
dose (10 mg/kg) of erlotinib (WTerlot), and Abcb1a/b(−/−)Abcg2(−/−) mice. ***P < 0.001, 1-way
ANOVA with Bonferroni’s multiple comparison test
6 Investigation of Transporter-Mediated Drug-Drug Interactions Using PET/MRI 131

be dynamically measured at the same time to assess transporter effects. This also
allows obtaining an image-derived blood input function, e.g. from the left ventricle
of the heart, which is needed for quantitative analysis of the PET data to obtain
quantitative parameters of transporter activity (e.g. kbile, kuptake). In addition, the com-
mercial availability of transgenic transporter knockout mice enables to elucidate the
effect of complete absence of one or several transporters on drug disposition. The
additional benefit of including the MRI measurements into this study was the easier
delineation of the organs of interest to extract the time-activity curves from the PET
image data. Especially the definition of the urinary bladder in the erlotinib co-
injected group was challenging without MR images as there was neither radioactiv-
ity uptake in the bladder nor in adjacent regions. Using the MRI data for the
definition of the VOIs leads to a more precise extraction of PET information result-
ing in reproducible quantitative data. Moreover, it also enables the user to create an
organ atlas template based on the MR images that could be then used to facilitate
analysis of PET-only image data.

6.6 Conclusion

The performed imaging study could confirm that Bcrp, P-gp and SLC uptake
transporters (e.g. OATPs) influence in vivo disposition of 11C-erlotinib and thereby
affect its distribution to normal and potentially also tumour tissue. Saturable
transport of erlotinib leads to non-linear pharmacokinetics, which needs to be
considered when attempting to predict the organ distribution of erlotinib in tumour
patients using PET scans with a microdose of 11C-erlotinib. Moreover, erlotinib
may be a perpetrator of transporter-mediated DDIs when being combined with
other drugs that are transported by BCRP, P-gp and OATPs. This may for instance
lead to changes in hepatic disposition of victim drugs. Inhibition of P-gp and
BCRP at the BBB appears to be a promising approach to enhance brain distribu-
tion of erlotinib to increase its efficacy in the treatment of NSCLC brain
metastases.
The motivation for combining PET with MR in this study was the interest in the
in vivo whole body disposition of 11C-erlotinib. A special focus was on the extrac-
tion of uptake and excretion rate constants in multiple organs, which made it neces-
sary to define the whole brain, lung, left ventricle of the heart, kidney, liver,
gallbladder, intestine and urinary bladder on the PET images. Some of these organs
(e.g. gallbladder) were only clearly visible in the MR images, and thus image data
analysis without the corresponding individual anatomical image would have been
impossible. As the desired quantitative PET parameters that need to be generated
were clear from the beginning of the study, the inclusion of the MR was already
fixed in the study planning phase.
Moreover, this study demonstrated that a 1T benchtop MRI system, which offers
many features of high-field strength preclinical MRI at low running costs, enables
high-quality PET/MRI studies when the MRI is installed in close proximity to the
dedicated PET scanner. The time difference between the scans is negligible and
132 T. Wanek et al.

animal monitoring and heating is constantly provided. The setup introduced in this
study showed that combined PET and anatomical MR imaging is feasible and can
be a valuable research tool in elucidating the role of transporters in drug
disposition.
For future dual modality experiments, the following list of items needs to be
considered in the study planning phase:

1. Selection of the animal model based on the research question

(a) Definition of age and sex.
(b) Inclusion of a control group.
(c) Adequate sample size calculation.
2. Selection of the functional imaging modality based on the research question
(a) PET versus SPECT.
(b) Available radioisotopes and labelling procedures.
(c) Biological versus physical half-life.
3. Selection of the anatomical imaging modality based on the research question
(a) Definition of organs (regions) of interest—thereafter selection of modality
(CT or MR) and selection of MR coils (head or whole body).
(b) Needed resolution—definition of voxel sizes.
4. Definition if integrated or sequential dual modality imaging is needed.
5. Definition if individual (especially MR) imaging is needed.

References
1. Nies AT, Schwab M, Keppler D. Interplay of conjugating enzymes with OATP uptake trans-
porters and ABCC/MRP efflux pumps in the elimination of drugs. Expert Opin Drug Metab
Toxicol. 2008;4(5):545–68.
2. Kusuhara H, Sugiyama Y. In vitro-in vivo extrapolation of transporter-mediated clearance in
the liver and kidney. Drug Metab Pharmacokinet. 2009;24(1):37–52.
3. Langer O. Use of PET imaging to evaluate transporter-mediated drug-drug interactions. J Clin
Pharmacol. 2016;56(Suppl 7):S143–56.
4. Wagner CC, Langer O. Approaches using molecular imaging technology -- use of PET in
clinical microdose studies. Adv Drug Deliver Rev. 2011;63(7):539–46.
5. Nagy K, Toth M, Major P, Patay G, Egri G, Haggkvist J, et al. Performance evaluation of the
small-animal nanoScan PET/MRI system. J Nucl Med. 2013;54(10):1825–32.
6. Wehrl HF, Amend M, Thielcke A. Multimodal imaging and image fusion. In: Kiessling F,
Pichler BJ, Hauff P, editors. Small Animal imaging basics and practical guide. Berlin: Springer
International Publishing; 2017. p. 491–507.
7. Kalliokoski A, Niemi M. Impact of OATP transporters on pharmacokinetics. Br J Pharm.
2009;158(3):693–705.
8. Hammerman PS, Janne PA, Johnson BE. Resistance to epidermal growth factor receptor tyro-
sine kinase inhibitors in non-small cell lung cancer. Clin Cancer Res. 2009;15(24):7502–9.
9. Ling J, Johnson KA, Miao Z, Rakhit A, Pantze MP, Hamilton M, et al. Metabolism and excre-
tion of erlotinib, a small molecule inhibitor of epidermal growth factor receptor tyrosine
kinase, in healthy male volunteers. Drug Metab Dispos. 2006;34(3):420–6.
10. Agarwal S, Manchanda P, Vogelbaum MA, Ohlfest JR, Elmquist WF. Function of the blood-
brain barrier and restriction of drug delivery to invasive glioma cells: findings in an orthotopic
rat xenograft model of glioma. Drug Metab Dispos. 2013;41(1):33–9.
6 Investigation of Transporter-Mediated Drug-Drug Interactions Using… 133

11. de Vries NA, Buckle T, Zhao J, Beijnen JH, Schellens JH, van Tellingen O. Restricted brain
penetration of the tyrosine kinase inhibitor erlotinib due to the drug transporters P-gp and
BCRP. Investig New Drugs. 2012;30(2):443–9.
12. Kodaira H, Kusuhara H, Ushiki J, Fuse E, Sugiyama Y. Kinetic analysis of the cooperation of
P-glycoprotein (P-gp/Abcb1) and breast cancer resistance protein (Bcrp/Abcg2) in limiting the
brain and testis penetration of erlotinib, flavopiridol, and mitoxantrone. J Pharmacol Exp Ther.
2010;333(3):788–96.
13. Loening AM, Gambhir SS. AMIDE: a free software tool for multimodality medical image
analysis. Mol Imaging. 2003;2(3):131–7.
14. Tai YC, Ruangma A, Rowland D, Siegel S, Newport DF, Chow PL, et al. Performance evalu-
ation of the microPET focus: a third-generation microPET scanner dedicated to animal imag-
ing. J Nucl Med. 2005;46(3):455–63.
15. Takashima T, Wu C, Takashima-Hirano M, Katayama Y, Wada Y, Suzuki M, et al. Evaluation
of breast cancer resistance protein function in hepatobiliary and renal excretion using PET
with 11C-SC-62807. J Nucl Med. 2013;54(2):267–76.
16. Takano A, Kusuhara H, Suhara T, Ieiri I, Morimoto T, Lee YJ, et al. Evaluation of in vivo
P-glycoprotein function at the blood-brain barrier among MDR1 gene polymorphisms by
using 11C-verapamil. J Nucl Med. 2006;47(9):1427–33.
17. Schmid A, Schmitz J, Mannheim JG, Maier FC, Fuchs K, Wehrl HF, et al. Feasibility of
sequential PET/MRI using a state-of-the-art small animal PET and a 1 T benchtop MRI. Mol
Imaging Biol. 2013;15(2):155–65.
18. Johnston RA, Rawling T, Chan T, Zhou F, Murray M. Selective inhibition of human solute
carrier transporters by multikinase inhibitors. Drug Metab Dispos. 2014;42(11):1851–7.
Multimodality Preclinical Imaging
in Inflammatory Diseases 7
Paul D. Acton

7.1 Inflammation and Its Role in Disease

Inflammation is a complex biological response to invasion of the host by pathologi-

cal organisms and other harmful stimuli, such as bacterial infection, damaged cells,
wounds, or irritants. The inflammatory response typically is the first in a cascade of
defense mechanisms, and plays a vital role in the elimination of harmful pathogens,
and in the initial repair of damaged tissue.
Acute inflammation is a normal response of the body to maintain homeostasis
after injury or host invasion. The early inflammatory response is considered part of
the innate immune system, since it is not specific to any particular pathogen [1]. The
process typically occurs rapidly after the initial harmful stimulus and involves the
local activation of a number of immune cells, particularly macrophages, dendritic
cells, Kupffer cells, and mast cells. Depending on the stimulus, different cell types
recognize different features of the stimulus or pathogen and release pro-inflamma-
tory cytokines, such as interleukin-1 (IL-1) and tumor necrosis factor alpha (TNFα),
and chemokines, such as interleukin-8 (IL-8). The resulting vasodilation and
increase in vascular permeability triggers accumulation of plasma proteins, such as
fibrin and antibodies, at the site of injury. This leads to many of the classical features
of inflammation, such as redness from the increase in blood flow to the tissue, and
edema and swelling from the accumulation of fluid and proteins.
Leukocytes, including macrophages and neutrophils, extravasate into tissue,
migrating to the site of injury along a concentration gradient of the locally released
chemoattractants [2]. Depending on the nature of the tissue damage, phagocytosis
of pathogens or cellular debris occurs. If the inflammatory response was triggered
by a wound, aggregation of platelets and fibrin occurs, leading to coagulation and
clot formation and eventually tissue repair.

P. D. Acton (*)
Janssen Research and Development, Johnson & Johnson, Spring House, PA, USA
e-mail: [email protected]

© Springer Nature Switzerland AG 2019 135

C. Kuntner-Hannes, Y. Haemisch (eds.), Image Fusion in Preclinical
Applications, https://doi.org/10.1007/978-3-030-02973-9_7
136 P. D. Acton

The inflammatory response from an acute stimulus will continue for as long as
the stimulus is present. Many of the mediators of inflammation degrade rapidly in
tissue. The termination of the acute inflammatory response is vital to prevent tissue
damage from the host’s own immune system and an accelerating cycle of stimulus
and immune response. Resolution of the inflammation is strictly regulated and
involves the cessation of the response by the release of anti-inflammatory agents,
such as interleukin-10 (IL-10), and the downregulation of pro-inflammatory gene
expression. If the stimulus continues for a longer period, a shift to chronic inflam-
mation can be triggered, driving a change in the type of cells present in the target
tissue [3]. During chronic inflammation, the healing process occurs, while the dam-
aging effects of the external stimulus are ongoing.
While innate immunity offers security against a wide range of pathogens and
other harmful stimuli, a more versatile system, known as adaptive immunity, has
evolved to provide increased protection [1]. The adaptive immune system offers an
immunological memory after an initial pathogenic invasion, which allows an
enhanced response to any subsequent challenges from the same pathogen. In the
presence of antigens, such as cell surface proteins on a bacterium, lymphocytes (B-
and T-cells) are activated. B-cells secrete antibodies, which are highly selective to
the presenting antigen, and prevent binding of the pathogen to host cells. Cytotoxic
T-cells induce the death of an invading or tumor cell, while other cell types regulate
the complex interaction between the immune system and the pathogen. A class of
memory cells provide antigen-specific immunity that persists long after the stimulus
has been removed, allowing any subsequent challenge from the same pathogen to be
dealt with rapidly. Similar to the innate immune system, careful control of the adap-
tive immune system is required to ensure the destruction of invading pathogens is
complete, while preventing damage to healthy tissue.
Failure of the immune system to manage the delicate cycle between an adequate
response to harmful stimuli and removal of that response once the threat has been
dealt with can cause a range of inflammatory disorders. Dysregulation of the
immune system can lead to an abnormal, aberrant, or out-of-control inflammatory
response to external stimuli or even to substances that occur naturally in the body.
Allergies are a result of hypersensitivity to an external stimulus, or allergen, leading
to an excessive inflammatory response that can be life-threatening [4]. In contrast,
autoimmune diseases, such as rheumatoid arthritis (RA) and inflammatory bowel
disease (IBD), are caused by the body’s immune system attacking its own organs
[5]. Indeed, conditions such as obesity, diabetes, and other metabolic diseases,
which previously were thought to be unrelated to the immune system, are believed
now to be caused by complex interactions between the immune system, adipose tis-
sue, and pancreatic beta cells and can be considered as chronic inflammatory disor-
ders [6–8].
Neuroinflammation represents a special category of immune response. The brain
is protected by the blood-brain barrier (BBB) and was considered an immune-priv-
ileged organ. However, it is known now that the brain and central nervous system
(CNS) are in direct connection with the immune system [9]. Specialized cells, such
as microglia, are the brain’s own immune cells and act as the main defense against
7 Multimodality Preclinical Imaging in Inflammatory Diseases 137

pathological invasion [10]. They are active scavengers of cellular debris, damaged
tissue, and neuritic plaques, and, like macrophages, microglia are phagocytic and
release cytokines to trigger an inflammatory response. However, similar to the
peripheral immune system, chronic neuroinflammation can lead to an uncontrolled
destructive microglial response, worsening the underlying tissue damage. A number
of neurological and psychiatric disorders are thought now to have a neuroinflamma-
tory pathogenesis, including Parkinson’s disease [11], multiple sclerosis [12],
Alzheimer’s disease [13], and chronic depression [14, 15], and there may be links
between peripheral metabolic disorders and central neuroinflammation [16].

7.2 Imaging Inflammation

In order to image inflammation, it is necessary to understand the features of the

inflammatory response that are amenable to imaging. A summary of these features
is presented in Table 7.1.
One of the most fundamental changes that occurs at the site of inflammation dur-
ing an immune response is a dramatic change in tissue metabolism [17]. The rapid
influx of inflammatory cells, coupled with the increase in vascular permeability and
blood flow, leads to a significant increase in the demand for nutrients and oxygen.
Myeloid cells, such as neutrophils and macrophages, expend large amounts of
energy during recruitment to the site of inflammation and during the process of
phagocytosis. This energy demand is met through local depletion of nutrients, such
as glucose, and increased oxygen consumption. Neutrophils in particular derive
much of their energy from glycolysis [18], which makes them suitable targets for
metabolic imaging tracers, such as 2-deoxy-2-(18F)fluoro-D-glucose (18F-FDG)
[19]. However, 18F-FDG is non-specific, and any tissue or organ that is highly meta-
bolic, such as the brain or a tumor, will show increased uptake.
The standard of care for imaging infection and inflammation uses radiolabeled
autologous leukocytes. This process involves withdrawal and separation of blood
cells from the subject, radiolabeling the white blood cells, followed by reinjection
of the tagged cells [20]. The radiolabeled cells, typically tagged with 99mTc or 111In,
accumulate at the site of inflammation and can be imaged using single-photon emis-
sion computed tomography (SPECT) [21, 22]. While it has been suggested that
18
F-FDG positron emission tomography (PET) could supersede SPECT imaging of
radiolabeled leukocytes, making use of the superior performance of PET, in many
situations questions remain over the accuracy and suitability of 18F-FDG in this
application [23]. However, in vitro radiolabeling of autologous leukocytes with 18F-
FDG, making use of the leukocyte’s avidity for glucose, and then injection into the
subject and imaging with PET could provide the best of both worlds—the superior
performance of PET coupled with the specificity of labeled leukocyte imaging [24].
Increased blood flow and vascular permeability at the site of inflammation, like
changes in metabolism, are amenable to imaging using many conventional tech-
niques [25]. Contrast-enhanced (CE) magnetic resonance imaging (MRI) and X-ray
computed tomography (CT), with blood pool contrast agents, can demonstrate
138 P. D. Acton

Table 7.1 List of potential targets for imaging inflammation and some probes and imaging
modalities that have been used
Probe(s) and imaging
Target modality Comments
Cell metabolism 18
F-FDG PET Non-specific—tumors and other
highly metabolic tissues also show
increased 18F-FDG uptake
Leukocytes In vitro cell labeling with Low spatial resolution—non-specific
99m
Tc or 111In and SPECT uptake in infection as well as
inflammation
Ultrasound microbubbles Low sensitivity
Blood flow CE-CT or CE-MRI Leakage of contrast into tissue highly
DCE-CT or DCE-MRI visible. DCE-MRI or CT provides
Ultrasound quantitative measures of blood flow
Edema Conventional MRI or CT Hyperintense T2 signal, reduced CT
density
DWI Changes in ADC good indicator of
edema
Hypoxia 18
F-FMISO, 18F-FAZA, Slow clearance and non-specific
18
F-HX4, 64Cu-ATSM PET binding of 18F-FMISO. High
Optical probes sensitive to radiation dose for 64Cu
nitroreductase
Phagocytosis SPIO particles and MRI Low sensitivity
19
F perfluorocarbons and Virtually no background signal, but
MRI very poor sensitivity at clinical field
strengths
CD8+ T-cells 89
Zr-Df-IAB22M2C PET Can detect a few million CD8+
and other labeled T-cells in preclinical studies.
antibodies, minibodies, Residualization of 89Zr can lead to
etc. high background signals in clearance
organs
Various cell surface Various labeled antibodies, Can be used with both optical and
markers on immune cells minibodies, fragments, etc. PET/SPECT imaging depending on
(CD40, MHC, CD11b, label
etc.)
TSPO 11
C-(R)-PK11195 PET Low brain uptake, high non-specific
binding
18
F-PBR06, 11C-PBR28, Higher specific binding, but
etc., PET susceptible to TSPO polymorphism
in humans
Proteases ProSense optical Activatable fluorophore, so light only
emitted in presence of target enzyme.
Slow kinetics
Matrix metalloproteinase MMPSense optical Activatable fluorophore, so light only
(MMP) emitted in presence of target enzyme.
Slow kinetics
PET tracers for MMP
Bone density, volume, CT Structural changes visible on CT
surface area, roughness often lag behind the inflammatory
response
7 Multimodality Preclinical Imaging in Inflammatory Diseases 139

changes in blood volume around a site of inflammation and any leakage into the
interstitial space. Dynamic contrast-enhanced (DCE) imaging allows the generation
of quantitative measures of vascular blood flow and leakage. Areas of edema lead to
a significant increase in the T2-weighted signal on non-contrast MRI, while CT
shows a reduced region of X-ray density. Diffusion-weighted (DW) MRI also
exhibits increased apparent diffusion coefficient (ADC), due to the free diffusion of
water molecules extravasated from blood vessels in areas of edema.
Inflammation in the brain often leads to disruption of BBB integrity, leading to
vascular leakage into the brain. This can be visualized using very similar methods
to those utilized for imaging peripheral vascular permeability, including CE-MRI
and CE-CT, and DWI. Radioactive tracers also are used to image BBB leakage.
Typically, these tracers are not lipid soluble, so will not cross the intact BBB, but are
readily extravasated into the brain where the barrier has been compromised by
inflammation [25, 26]. In a similar manner, near-infrared (NIR) fluorescent probes
have been developed which normally are retained in the blood vessels but will accu-
mulate in the brain in the presence of vascular leakage [27].
Inflamed tissue often becomes hypoxic [28, 29], driven by elevated levels of
hypoxia-inducible factors (HIF) [30, 31]. The rapid influx of invading pathogens
and immune cells and the release of oxygen-consuming enzymes into an area of
inflammation lead to an imbalance in the supply and demand of oxygen, nutrients,
and metabolites [32]. While the high metabolic demand of neutrophils makes them
a good target for imaging probes such as 18F-FDG, the presence of hypoxia also
provides a suitable target for imaging, and a number of optical and PET probes have
been developed which target features of the hypoxic environment. Most often, these
probes rely on being taken up and retained by hypoxic cells, while being cleared
rapidly by normoxic tissue, giving a high target-to-background ratio.
One of the most widely studied PET imaging agents for hypoxia is 18F-FMISO
[33–35]. Like most of the probes based on nitroimidazoles, this enters cells by pas-
sive diffusion, where it is reduced by nitroreductase enzymes. In the presence of
normoxic levels of oxygen, the radical is reoxidized rapidly, and free 18F-FMISO
diffuses out of the cell and may be cleared. However, in hypoxic cells, where the
partial pressure of oxygen (pO2) is less than approximately 10 mmHg, the reduction
products bind to intracellular macromolecules, and the radiolabeled material
becomes trapped [36]. While 18F-FMISO was one of the first PET hypoxia imaging
agents, it has drawbacks related to slow clearance and non-specific uptake [37].
Second-generation nitroimidazoles have been developed to improve on the perfor-
mance of 18F-FMISO, including 18F-FAZA and 18F-HX4 [38]. A number of fluores-
cent probes make use of the same nitroreductase enzymes to cleave a fluorescent
substrate or to trigger Förster resonance energy transfer (FRET), allowing NIR
imaging of hypoxia [39–41]. The metal chelate 64Cu-ATSM [42] also relies on intra-
cellular reduction of the radiolabeled complex, which is reoxidized back to the par-
ent molecule under normoxic conditions. In the presence of hypoxic cells, the
reduction products dissociate into 64Cu(I) which becomes trapped irreversibly inside
the cell. Despite a relatively high radiation dose, due to the long half-life of 64Cu, the
imaging performance of 64Cu-ATSM appears to be superior to 18F-FMISO [43].
140 P. D. Acton

While the majority of imaging studies using these probes have concentrated on
tumor hypoxia, the presence of hypoxic inflammatory lesions also should be ame-
nable to this type of imaging [44, 45].
The presence of phagocytes at the site of inflammation is amenable to imaging,
by making use of the ability of the cell to take up small particles that are identified
as alien to the host. Phagocytosis of small and ultra-small superparamagnetic iron
oxide (SPIO) particles by monocytes at the site of inflammation leads to an accumu-
lation of the iron particles, which is visible with MRI. These iron-rich macrophages
can be visualized by an increase in image contrast on T1-weighted MRI or a reduc-
tion in signal on T2 images [46]. In a similar manner, targeted lipid microbubbles,
used for perfusion imaging with ultrasound, can interact with leukocytes, leading to
an accumulation in inflamed tissue, which is visible as increased contrast on ultra-
sound [47].
An alternative to SPIO particles are the perfluorocarbons, which allow an 19F
signal to be detected on a conventional 1H MRI by retuning the radiofrequency coils
[48]. The main advantage of 19F imaging with MRI is the lack of background signal,
due to the absence of naturally occurring 19F in the soft tissues of the body. In addi-
tion, the 19F signal can be acquired in the same scanning session as the 1H image,
which provides a functional 19F map overlaid on a spatially aligned anatomical
image [49]. Following administration of 19F-perfluorocarbon, the small emulsion
droplets are taken up by phagocytosis into monocytes and macrophages at the site
of inflammation [50, 51]. This technique has been used to assay macrophage activ-
ity in mouse models of IBD [52], to study treatment response in RA [53], and for
measuring peripheral nerve inflammation [54]. The main disadvantage of the 19F
technique is the low sensitivity, particularly at clinical magnetic field strengths,
requiring millimolar quantities of contrast agent in each image voxel [49]. This is to
be compared with the picomolar concentrations that are detectable with PET.
Phagocytic cells also express surface proteins, which are upregulated in the pres-
ence of inflammation. The 18 kDa translocator protein (TSPO), formerly known as
the peripheral benzodiazepine receptor, has been identified as a suitable target for
imaging inflammation in the periphery and CNS [25]. The radioligand
11
C-(R)-PK11195 has been used for PET imaging of TSPO expression for many
years [55, 56], although it has relatively poor brain penetration and high non-spe-
cific uptake compared to more recent probes [57–59]. These newer tracers include
the phenoxyarylacetamides, such as 11C-DAA1106, and compounds such as 11C-
PBR28 and 18F-PBR06 (for a review and comparison of these, see [60]).
Unfortunately, a problem with these new-generation TSPO PET tracers was identi-
fied, in which a polymorphism in the TSPO gene causes changes in binding of the
probes [61–63]. However, if each subject is genetically tested for this polymor-
phism, interpretation of the imaging data may be more accurate.
Proteases are found in all organisms and are responsible for a vast array of pro-
cesses involving the breakdown of proteins. Specifically for the immune response,
proteases are involved in the blood clotting process in response to tissue damage
and are active in the complement system [64]. Various proteases are upregulated
during inflammation and are a suitable target for imaging probes. Some optical
7 Multimodality Preclinical Imaging in Inflammatory Diseases 141

imaging agents make use of the protease enzyme to cleave the probe, separating the
fluorophore from the quencher, leading to light emission. These activatable probes
are particularly powerful, as they emit light only in the presence of the activating
enzyme.
Matrix metalloproteinases (MMPs) are a family of protease enzymes that are
involved in a number of cellular processes, including degrading the extracellular
matrix, cell proliferation and migration, apoptosis, tissue repair, and immune
response. Generally, similar to other proteases, MMPs are upregulated during
inflammation and have been used as a target for both optical and radionuclide imag-
ing [65–68]. While early MMP imaging probes were pan-MMP selective, more
recently they have been developed to target specific members of the MMP family
[69, 70].
The response of the adaptive immune system to pathological invasion results in
the influx of large numbers of B- and T-cells to the damaged tissue. These cells typi-
cally rely on oxidative phosphorylation as their source of energy, utilizing amino
acid, glucose, and lipid metabolism [17]. Many immune cells express cell surface
receptors that are amenable to imaging, particularly when the inflammatory process
induces a significant upregulation of receptor expression. Cytotoxic T-cells express
the CD8 glycoprotein on the cell surface, which is a promising target for imaging
T-cell influx and activation at the site of inflammation [71, 72]. Similarly, imaging
the expression of CD40, which is upregulated on the surface of many immune cells
during inflammation, can be achieved using optical imaging methods [73], while
antibody fragments against class II major histocompatibility complex (MHC) and
CD11b have been visualized using PET [74].

7.3 Imaging Neuroinflammation

Neuroinflammation has been implicated in the pathophysiology of a number of psy-

chiatric and neurological disorders, such as Alzheimer’s disease, depression, stroke,
and multiple sclerosis. Imaging of the inflammatory process in the brain could pro-
vide an early indication of disease or reveal the extent of damage, allowing interven-
tion or treatment [75, 76]. Clinical studies of TSPO expression using PET could
provide some of the most valuable insights into neuroinflammation-related diseases.
However, many have been inconclusive, where some studies have observed signifi-
cant increases in TSPO in schizophrenia [77] and depression [78], while others have
failed to demonstrate any signs of neuroinflammation [79–82]. The absence of neu-
roinflammation could be due to the imaging studies being performed too late in the
course of the disease, where the inflammatory component already has caused sig-
nificant loss of brain function, but the underlying inflammation has resolved.
Alternatively, neuroinflammation may be present only during extreme periods of
symptomatic behavior, such as acute episodes of major depression or psychosis, and
imaging TSPO expression during those periods could reveal increased uptake.
Similar conflicting results have been observed in neurological disorders, such as
Alzheimer’s disease [83–86].
142 P. D. Acton

Animal models of neuroinflammation have become vital in both the develop-

ment of imaging probes to study features of the inflammatory process and also in
studying the underlying disease and interactions between the peripheral and central
immune system [87–89]. Multimodality imaging of neuroinflammation can provide
information in individual animals on disruptions to the BBB, molecular indications
of the underlying inflammation, such as TSPO, and the expression of inflammatory
genes. As an example, Fig. 7.1 shows a preclinical PET-MR study, in which the
excitotoxin, kainic acid, was administered peripherally to rats to induce neuroin-
flammation, and imaging was performed using the TSPO PET tracer 18F-PBR06
[90–95], and MRI performed to assess cerebral edema. High uptake of 18F-PBR06
is observed in the thalamus, hippocampus, and amygdala of kainic acid animals,
which is absent in healthy controls, and corresponds to regions confirmed in other
studies [96]. The brain regions exhibiting increased uptake with PET correspond
closely with areas of T2 hyperintensity on MRI, indicating edema (Fig. 7.2).
Interestingly, the edema also appears to manifest itself as highly enlarged ventricles,
shown clearly as the dark regions of hypointensity on T1 MRI. Many of these fea-
tures of PET-MR imaging correlate significantly with other indicators of disease

Fig. 7.1 PET-MR imaging of the TSPO tracer 18F-PBR06 in a rat model of neuroinflammation,
showing increased uptake of the tracer in the brain after kainic acid administration, overlaid on the
corresponding MRI slice (CPUT caudate-putamen, AMYG amygdala, HIPP hippocampus, CER
cerebellum)
7 Multimodality Preclinical Imaging in Inflammatory Diseases 143

Fig. 7.2 MRI in a kainic acid rat model of neuroinflammation demonstrates hyperintensity on
T2-weighted scans, corresponding to areas of cerebral edema and enlarged ventricles. Yellow
arrows indicate fluid accumulation in the ventricles, leading to enlargement. Blue and green arrows
show T2 hyperintensity in the hippocampus and amygdala, respectively

severity, such as weight loss. High-field MRI volumetric studies have demonstrated
similar results, where kainic acid-treated animals had a significantly smaller hip-
pocampus and increased ventricle size [97].
One significant advantage small animal studies have over clinical applications
in humans is the ability to use transgenic animals, in which one or more genes are
altered to provide detailed information on inflammatory pathways. A particularly
useful mouse model developed recently uses a firefly luciferase (luc) reporter gene,
coupled to a 12 kb fragment of the glial fibrillary acidic protein (GFAP) promoter
and a 850 bp human β-globin intron 2, to allow bioluminescence imaging of GFAP
expression [98]. The GFAP protein is expressed by a number of cell types in the
CNS, including astrocytes, and is known to be significantly increased during neu-
roinflammation. Indeed, GFAP immunohistochemistry is one of the standard meth-
ods for studying astrocyte activation in the presence of neurotoxins, neuronal
injury, and inflammation, although the precise mechanism of GFAP upregulation is
unclear.
The GFAP-luc mouse (FVB/N-Tg(Gfap-luc)53Xen) has been studied exten-
sively in a number of inflammatory models [90, 98–101]. Peripheral administration
of pro-inflammatory agents, such as lipopolysaccharide (LPS), kainic acid, and
144 P. D. Acton

Fig. 7.3 Imaging bioluminescence light emission from Gfap-luc transgenic mice after adminis-
tration of increasing doses of the pro-inflammatory compound kainic acid

TNFα, leads to increased expression of the GFAP gene and the emission of biolu-
minescence light in the presence of the luciferin substrate (Fig. 7.3). The amount of
light emitted increases by up to two orders of magnitude from baseline and exhibits
a steep dose-response (Fig. 7.4), which is highly correlated with seizure score and
other measures of sickness behavior, such as weight loss. The time course of light
emission after a single dose of kainic acid exhibits a rapid increase up to 24 h post-
administration, which remains stable for several days and then begins to resolve
slowly after 3–4 days (Fig. 7.5), which is mirrored in measures such as weight loss.
One consequence of such a steep dose-response could be the impact on therapeu-
tic studies. At the higher doses of kainic acid, the induced inflammation and neuro-
nal damage could be so severe that it would be impossible for any therapeutic to
recover the lost brain function. Therefore, using lower doses, around the region of
rapid increase of the dose-response curve, could be more useful in studies of treat-
ment effect. With a dose of 20 mg/kg kainic acid, a study was able to demonstrate
inhibition of astrogliosis in the GFAP-luc mouse model with treatment by colony-
stimulating factor 1 (CSF1), but only when treated prior to, or up to 6 h after, kainic
acid [102]. Any further delay in treatment, beyond 6 h, could no longer rescue dam-
aged neurons.
It should be noted also that the sensitivity to pro-inflammatory agents, such as
kainic acid, is highly age- and strain-dependent for both rats [103] and mice [104,
105]—for example, DBA/2J mice are very sensitive to chemically induced seizures,
while C57BL/6J are quite resistant.
7 Multimodality Preclinical Imaging in Inflammatory Diseases 145

Fig. 7.4 Dose-response of 24 h time point

bioluminescence light 5
emission from the brain of
Gfap-luc transgenic mice
after administration of
kainic acid. Dose range of 4
kainic acid is 5–30 mg/kg,

Log(BLI)
and ED50 (the dose at
which 50% of the
maximum response is 3
observed) is 10.6 ± 0.2 mg/
kg
2

0.5 1.0 1.5

Log(dose)

100000

10000

vehicle
% baseline

5 mg/kg
1000 10 mg/kg
12 mg/kg
15 mg/kg
100 20 mg/kg
25 mg/kg
30 mg/kg

10
0 20 40 60 80 100
Time post-kainic acid (h)

Fig. 7.5 Time course of bioluminescence light emission after acute administration of varying
doses of kainic acid

7.4 Imaging Rheumatoid Arthritis

Rheumatoid arthritis (RA) is a debilitating chronic autoimmune disease that affects

the joints, resulting in pain, stiffness, and swelling [106]. Inflammation of the
synovial membrane eventually leads to tendon tethering and destruction of the
cartilage and bone surface in the joint, causing loss of function and deformity.
146 P. D. Acton

While primarily a disease of the joint, RA can lead to reactions in the skin, lungs,
kidneys, heart, eyes, and other organs. Diagnosis of RA typically is done using
conventional imaging methods, such as X-rays of the affected joints looking for
swelling of the soft tissue and early bone erosion. However, structural changes due
to RA, which are visible on anatomical imaging modalities, often occur long after
the underlying chronic inflammation has taken hold. The use of molecular imaging
methodologies, looking for the specific signatures of inflammation, may enable
much earlier treatment, before joint damage has occurred [107–109]. Similarly,
detailed analysis of DCE-MRI and diffusion tensor imaging (DTI), coupled with
the use of novel contrast agents, could provide an early signal from changes in the
synovial space [110, 111]. Ultrasound offers similar opportunities for imaging the
structure of the affected joint and changes in blood flow resulting from the inflam-
matory response [112].
A number of genetic and environmental factors may play a role in the patho-
physiology of RA, although the precise mechanism of the initial cellular activation
is unknown [113]. The initiating inflammatory trigger leads to a T-cell-mediated
amplification of the inflammation and a chronic phase leading to joint injury and
bone erosion. Cytokines such as IL-1, TNFα, and IL-6 are released at high concen-
trations into the synovial fluid, which accelerate the tissue damage. TNFα appears
to be one of the main pro-inflammatory cytokines in RA, which has led to the
development of a number of therapeutics targeting this chemical, primarily anti-
TNFα antibodies and fusion proteins. However, since TNFα is known to inhibit
tumorigenesis and viral replication, the inhibition of this cytokine can have serious
side effects, such as certain cancers and a vulnerability to opportunistic infection
and tuberculosis (TB). While biologic therapeutics have been successful in treating
RA, these potential side effects from systemic exposure can lead to therapy discon-
tinuation and, in the most severe cases, death. Therefore, strategies to develop
more targeted therapies to the joint, while reducing systemic toxicity, have been
developed, and imaging will play a vital role in measuring joint penetration of
these treatments [114].
Despite their inherent limitations, animal models of RA have been vital to prog-
ress our understanding of the disease and in the development of new therapeutics
[115, 116]. Collagen-induced arthritis (CIA) is one of the most widely used models,
utilizing immunization of animals with type II collagen to induce symptoms very
similar to human RA. Histological examination reveals an influx of cells into syno-
vial tissue that resembles RA and destruction of bone and cartilage. While disease
penetration is strain-dependent, one advantage of the CIA model is that it can be
used not only in rodents but also in nonhuman primates (NHP) [117].
Imaging has been used extensively in the CIA model, using methods to probe the
underlying inflammation and also structural imaging of bone loss. Although a non-
specific marker of the inflammatory process, 18F-FDG PET is a useful method to
quantify the increased metabolic activity induced by joint inflammation [118].
Uptake of 18F-FDG in the joint is correlated with clinical and histopathological
scores of disease severity throughout the onset and progression of disease [119].
7 Multimodality Preclinical Imaging in Inflammatory Diseases 147

Dual-probe fluorescence imaging in mice has allowed the simultaneous acquisition

of signals from MMP activity, using the commercially available probe MMPSense,
and macrophage influx, using a fluorescently tagged antibody to CD11b [120]. The
response to treatment can be monitored using molecular imaging of a number of
inflammatory targets, including fibroblast activation protein (FAP), macrophages,
and αvβ3 integrins [121, 122]. Similarly, 19F-MRI of perfluorocarbons in the joint of
CIA rats is correlated with disease severity and is an indicator of response to therapy
[53]. Combining imaging modalities allows the greatest flexibility, providing imag-
ing of joint metabolism with 18F-FDG PET and optical imaging of NIR fluorescent
probes to visualize targets such as MMP, proteases, cathepsins, and others [123].
Figure 7.6 demonstrates the application of PET and optical imaging of the hind paw
in a CIA mouse model of RA. Interestingly, 18F-FDG uptake can be measured using
conventional PET imaging and also by detecting the Cerenkov light emission pro-
duced by the positron [124, 125].
PET imaging of 18F-fluoride uptake by bone provides valuable information on
bone mineral deposition and osteoblastic activity. Although used primarily for
detection and quantification of metastatic bone cancer, fluoride PET can detect
murine arthritis as early as 14 days after administration of pro-inflammatory agents
[126]. Comparison with CT of the same joints revealed a reasonable correlation
between clinical scores or bone surface measurements and 18F-fluoride uptake.

Fig. 7.6 Imaging inflammation in the joints of CIA mice (arrowed). 18F-FDG uptake can be
observed using both conventional PET imaging and optical imaging of Cerenkov light emission.
In the same animals, MMP expression is measured using MMPSense and NIR fluorescence imag-
ing. Note the field of view of the PET scan is smaller than the other optical images, and does not
cover the whole mouse. Also, the right paw exhibits increased FDG and MMPSense uptake in all
images, indicating a greater degree of inflammation in that paw compared to the other
148 P. D. Acton

Fig. 7.7 High-resolution CT imaging of a hind paw demonstrates dramatic destruction of bone in
the CIA mouse, leading to decreased bone density and surface erosion (left). A map of the surface
roughness indicates the joint regions most affected by the disease (center), while detailed examina-
tion of the joint reveals bone thinning at the growth plate boundary and loss of cartilage and col-
lapse of joint space (right). F femur, T tibia, Fi fibula, FC condyle of femur, TC condyle of tibia

Bone erosion in animal models of RA can be visualized using high-resolution

CT imaging of the joint (Fig. 7.7). While a general decrease in bone density in
the affected regions is expected, other features, such as joint space reduction due
to cartilage damage and changes in bone volume and surface area, also are
resolved quite clearly [127, 128]. In addition, quantification of bone erosion has
been performed by measuring the roughness of the bone surface in the vicinity of
the joint [129, 130]. This roughness is defined in a number of ways, including the
depth of features from a plane parallel to the local surface or from the angle of
intercepting polygons drawn on a surface rendering. The distribution of rough-
ness values demonstrates features consistent with the progression of disease,
such as a shift to higher absolute roughness values in more severe disease
(Fig. 7.8).
The potentially severe side effects from conventional treatments for RA, such as
anti-TNFα antibodies, have driven the search for more targeted therapies, which
will maximize the concentration of the drug at the affected joint, while minimizing
systemic exposure and the resulting toxicity [114]. Most of the targeted therapies
make use of features of the arthritic joint, such as increased vascular permeability,
changes in lymphatic drainage, alterations in temperature or pH, and increased
expression of certain enzymes or other proteins. A variety of novel drug technolo-
gies have been employed, including bispecific biologics, targeted nanoparticles, and
polymer-drug conjugates. Radiolabeling these therapeutics, and imaging with PET,
allows the quantitative measurement of uptake in a rodent model of RA, demon-
strating improved joint targeting (Fig. 7.9).
7 Multimodality Preclinical Imaging in Inflammatory Diseases 149

Fig. 7.8 Analysis of the 25000 Group mean roughness

surface roughness
distribution in CIA mice
shows the shift of the 20000
distribution to the right,
indicating more severe
disease and bone

Frequency
destruction 15000

Control
10000 Disease

5000

0
0 10 20 30 40 50 60
Roughness (degrees)

Fig. 7.9 PET imaging of a radiolabeled therapeutic demonstrates enhanced uptake into the
affected joint in RA (right picture, arrowed), which is confirmed on a fluorescently tagged version
of the same compound (center picture, arrowed). A healthy control animal exhibits significantly
lower uptake into the joint (left picture)
150 P. D. Acton

7.5 Imaging Inflammatory Bowel Disease

Inflammatory bowel disease (IBD) comprises both Crohn’s disease and ulcerative
colitis. IBD is characterized by inflammation of the ileum, colon, and rectum and
can involve both mucosal and transmural inflammation. It is believed that a
genetic predisposition to IBD leads to an inappropriate or unregulated response to
the intestinal microbiome, although there also is a strong environmental compo-
nent of the disease relating to diet, smoking behavior, and use of certain medi-
cines, particularly antibiotics [131, 132]. IBD is a known risk factor for the
development of colorectal cancer [133, 134], and, as such, methods for imaging
inflammation in the colon could provide an early indication of potential tumor
development [135].
Many of the conventional techniques for imaging inflammation, such as MRI,
CT, CE-MRI, CE-CT, DWI, 18F-FDG PET-CT, and autologous leukocytes, are
applicable in IBD [136–146]. MRI and CT contrast agents administered enterically
increase the contrast between the lumen and bowel wall, allowing better visualiza-
tion of bowel wall thickening, abscesses, edema, and mucosal enhancement. Like
in other inflammatory diseases, 18F-FDG PET can be used to study IBD, making
use of the influx of inflammatory cells and increased cellular metabolism at the site
of local disease. Multimodality PET-CT imaging is vital in this application to iden-
tify accurately the location, by CT, of any inflammatory lesions seen on
PET. However, the non-specific nature of 18F-FDG could lead to the misidentifica-
tion of a potential tumor as an IBD lesion, making this technique less useful for
primary diagnosis.
IBD and colon cancer offer an almost unique opportunity for imaging, since the
location of the inflammation is amenable to endoscopic examination, and could be
included as part of a routine colonoscopy [147]. This opens up the possibility of
using a variety of optical imaging techniques, which would be impossible in most
other organs due to the depth of tissue being imaged, and for topical administration
of any optical contrast agents [148–151]. Some methodologies used in endoscopic
imaging of the GI tract include conventional white light endoscopy, NIR fluores-
cence imaging, confocal laser endomicroscopy, Raman endoscopy, optical coher-
ence tomography, narrow-band imaging, and hyperspectral imaging [152–155].
Advances in nanotechnology have led to the development of targeted nanoparticles
that carry a diagnostic or therapeutic payload. Tagging these nanoparticles with a
fluorescent dye or radioisotope allows in vivo imaging of cellular and molecular
processes occurring in IBD [156].
A number of animal models of IBD have been developed that provide valuable
information on the pathophysiology of intestinal inflammation and the mechanisms
that drive loss of mucosal and epithelial barrier homeostasis [157–160]. One of the
most common chemically induced models involves the administration of the chelat-
ing agent dextran sulfate sodium (DSS) in the drinking water of mice. This causes
acute inflammation of the colon, leading to weight loss, diarrhea, and rectal bleed-
ing. A number of genetically engineered models of IBD also have been developed,
7 Multimodality Preclinical Imaging in Inflammatory Diseases 151

including those with innate immunity defects or aberrant adaptive immune cell
response [161, 162]. However, there are vast differences between these mouse mod-
els and human IBD, particularly as the human disease is driven by a wide array of
genetic and environmental factors, while experimental models tend to result from a
single genetic deletion or transgene overexpression. Therefore, these mouse models
should be used with caution in the development of novel therapeutics, as they can-
not represent the complex genetic, immune system, and environmental interactions
seen in humans [163].
Imaging studies of preclinical models of IBD have utilized a variety of imaging
modalities and probes to understand the inflammatory response. A study to compare
a number of different NIR fluorescent probes in a DSS mouse model of colitis found
that most of the activatable probes that are sensitive to protease and MMP enzymes
gave a good signal in areas of inflammation [164]. While the fluorescence signal
was correlated with histology and colitis scores, the high background signals may
still provide a problem during in vivo optical imaging. Similarly, a study using a
targeted γ-glutamyltranspeptidase (GGT) probe, which is activated in the presence
of the GGT enzyme, demonstrated high uptake in a DSS mouse model [165]. The
advantage of this optical imaging agent was that the probe was delivered topically,
via a spray-on method, significantly reducing systemic exposure and any potential
toxicity.
An extensive comparison of different imaging modalities and probes was per-
formed in the DSS colitis model, which revealed that many of the conventional
clinical imaging methods are equally applicable to mouse models [137]. PET
imaging of DSS mice has been performed to compare the non-specific metabolic
tracer, 18F-FDG (Fig. 7.10), with the TSPO imaging agent, 18F-DPA-714 [166].
While 18F-FDG has been shown to be useful for quantifying colon cancer [167], the
inflammation from acute DSS administration was found to be lower, and not accu-
rately resolved with 18F-FDG. However, TSPO imaging demonstrated significantly
increased uptake of the tracer, which could provide a more sensitive measure to
monitor the progression of disease and response to anti-inflammatory treatment.
Other PET imaging agents specific to targets known to be increased in IBD also
could have value, such as the inducible form of the cyclooxygenase enzyme,
COX-2 [168].
Similar to RA, anti-TNFα therapies have been effective in treating some patients
with IBD. However, not all patients respond to anti-TNFα treatment, which could
be due to variations in the expression of the target in the bowel. Molecular imaging
of TNFα expression prior to treatment, using a fluorescently tagged antibody to
membrane-bound TNFα, has demonstrated that patients with high levels of TNFα
show significantly improved response to treatment and has the potential to predict
response and select suitable patients for this type of treatment [169]. Methods to
deliver targeted therapies to the GI tract for IBD would have significant value in
reducing the systemic exposure and associated toxicities [170–173]. Visualization
of the delivery of the drug with imaging would be vital to ensure it reaches all
affected parts of the ileum, colon, and rectum [174]. Indeed, contrast agents
152 P. D. Acton

Fig. 7.10 18F-FDG uptake in a DSS mouse model of IBD, showing increased uptake in the GI
tract (left), which is confirmed by ex vivo Cerenkov imaging (right). The four GI tract sections
shown are (left to right) the duodenum, jejunum, ileum, and colon (cecum has been removed).
Image courtesy of Dr. Peter King, Janssen Research and Development

encapsulated into the same targeted delivery technology as the therapeutic com-
bines both treatment and imaging in a single package [175].

7.6 Conclusions

Inflammation is implicated in the pathophysiology of a vast range of diseases, from

Alzheimer’s disease and depression to chronic autoimmune disorders such as RA
and IBD, and even diabetes and other metabolic disorders. Imaging the inflamma-
tory response and immune system is becoming increasingly important in the under-
standing of these disorders and for the development of novel therapeutics. Many
targets associated with inflammation are amenable to imaging, including non-spe-
cific changes in metabolic activity or blood flow, alterations in enzymatic activity,
and expression of cell surface markers. However, multimodality imaging, incorpo-
rating features of both anatomical and molecular imaging, targeting a variety of
processes involved in the inflammatory response will maximize our understanding
of these disorders.
7 Multimodality Preclinical Imaging in Inflammatory Diseases 153

References
1. Murphy K, Weaver C. Janeway’s Immunobiology. 9th ed. New York: Garland Science; 2016.
2. Kolaczkowska E, Kubes P. Neutrophil recruitment and function in health and inflammation.
Natl Rev. 2013;13:159–75.
3. Murakami M, Hirano T. The molecular mechanisms of chronic inflammation development.
Front Immunol. 2012;3:323.
4. O’Hehir RE, Holgate ST, Sheikh A. Middleton’s allergy essentials. San Diego: Elsevier;
2016.
5. Mackay IR, Rose NR, editors. The autoimmune diseases. San Diego: Elsevier; 2014.
6. Donath MY, Shoelson SE. Type 2 diabetes as an inflammatory disease. Nat Rev Immunol.
2011;11:98–107.
7. Gonzalez-Chavez A, et al. Pathophysiological implications between chronic inflammation
and the development of diabetes and obesity. Cir Cir. 2011;79:190–7.
8. Lumeng CN, Saltiel AR. Inflammatory links between obesity and metabolic disease. J Clin
Invest. 2011;121:2111–7.
9. Ader R, Felten D, Cohen N. Interactions between the brain and the immune system. Annu
Rev Pharmacol Toxicol. 1990;30:561–602.
10. Nayak D, Roth TL, McGavern DB. Microglia development and function. Annu Rev Immunol.
2014;32:367–402.
11. Vivekanantham S, Shah S, Dewj R, et al. Neuroinflammation in Parkinson’s disease: role in
neurodegeneration and tissue repair. Int J Neurosci. 2015;125:717–25.
12. Ellward E, Zipp F. Molecular mechanisms linking neuroinflammation and neurodegeneration
in MS. Exp Neurol. 2014;262:8–17.
13. Heneka MT, Carson MJ, El Khoury J, et al. Neuroinflammation in Alzheimer’s disease.
Lancet Neurol. 2015;14:388–405.
14. Bhattacharya A, Derecki NC, Lovenberg T, et al. Role of neuro-immunological factors in the
pathophysiology of mood disorders. Psychopharmacologia. 2016;233:1623–36.
15. Dantzer R, O’Connor JC, Freund GG, et al. From inflammation to sickness and depression:
when the immune system subjugates the brain. Nat Rev Neurosci. 2008;9:46–56.
16. Van Dijk G, van Heijnengen S, Reijne AC, et al. Integrative neurobiology of metabolic dis-
eases, neuroinflammation, and neurodegeneration. Front Neurosci. 2015;9:173. https://doi.
org/10.3389/fnins.2015.00173.
17. Kominsky DJ, Campbell EL, Colgan SP, et al. Metabolic shifts in immunity and inflamma-
tion. J Immunol. 2010;184:4062–8.
18. Borregaard N, Herlin T. Energy metabolism of human neutrophils during phagocytosis. J
Clin Invest. 1982;70:453–65.
19. Wu C, Li F, Niu G, et al. PET imaging of inflammation biomarkers. Theranostics.
2013;3:448–66.
20. Sampson CB. Labeled cells for imaging infection. In: Cox PH, Buscombe JR, editors. The
imaging of infection and inflammation. Heidelberg: Springer; 1998. p. 31–60.
21. Rodie ME. Imaging inflammation with Tc-99m hexamethyl propylene amine oxime
(HMPAO) labeled leucocytes. Radiology. 1988;166:767–72.
22. Thakur ML. Indium-111 labeled leukocytes for the localization of abscesses: preparation,
analysis, tissue distribution and comparison with gallium-67 citrate in dogs. Lab Clin Med.
1977;89:217–28.
23. Glaudemans AWJM, Signore A. FDG-PET/CT in infections: the imaging method of choice?
Eur J Nucl Med Mol Imaging. 2010;37:1986–91.
24. Forstrom LA, Mullan BP, Hung JC, et al. 18F-FDG labeling of human leukocytes. Nucl Med
Commun. 2000;21:691–4.
25. Jacobs AH, Tavitian B, INMiND Consortium. Noninvasive molecular imaging of neuroin-
flammation. J Cereb Blood Flow Metab. 2012;32:1393–415.
26. Wunder A, Klohs J, Dirnagl U. Non-invasive visualization of CNS inflammation with nuclear
and optical imaging. Neuroscience. 2009;158:1161–73.
154 P. D. Acton

27. Klohs J, Steinbrink J, Bourayou R, et al. Near-infrared fluorescence imaging with fluores-
cently labeled albumin: a novel method for non-invasive optical imaging of blood-brain bar-
rier impairment after focal cerebral ischemia in mice. J Neurosci Methods. 2009;180:126–32.
28. Biddlestone J, Bandarra D, Rocha S. The role of hypoxia in inflammatory disease (Review).
Int J Mol Med. 2015;35:859–69.
29. Eltzschig HK, Carmeliet P. Hypoxia and inflammation. N Engl J Med. 2011;364:656–65.
30. Colgan SP, Campbell EL, Kominsky DJ. Hypoxia and mucosal inflammation. Annu Rev
Pathol. 2016;11:77–100.
31. Cummins EP, Keogh CE, Crean D, et al. The role of HIF in immunity and inflammation. Mol
Asp Med. 2016;47-48:24–34.
32. Koeppen M, Eckle T, Eltzschig HK. The hypoxia-inflammation link and potential drug tar-
gets. Curr Opin Anaesthesiol. 2011;24:363–9.
33. Dubois L, Landuyt W, Haustermans K, et al. Evaluation of hypoxia in an experimental rat
tumor model by [(18)F]fluoromisonidazole PET and immunohistochemistry. Br J Cancer.
2004;91:1947–54.
34. Prekeges JL, Rasey JS, Grunbaum Z, et al. Reduction of fluoromisonidazole, a new imaging
agent for hypoxia. Biochem Pharmacol. 1991;42:2387–95.
35. Tang G, Wang M, Tang X, et al. Fully automated one-pot synthesis of [(18)F]fluoromisonida-
zole. Nucl Med Biol. 2005;32:553–8.
36. Yip C, Blower PJ, Goh V, et al. Molecular imaging of hypoxia in non-small cell lung cancer.
Eur J Nucl Med Mol Imaging. 2015;42:956–76.
37. Krohn KA, Link JM, Mason RP. Molecular imaging of hypoxia. J Nucl Med. 2008;49(suppl
2):129S–48S.
38. Peeters SGJA, Zegers CML, Lieuwes NG, et al. A comparative study of the hypoxia PET
tracers [18F]HX4, [18F]FAZA, and [18F]FMISO in a preclinical tumor model. Int J Radiat
Oncol Biol Phys. 2015;91:351–9.
39. Guo T, Cui L, Shen J, et al. A highly sensitive long-wavelength fluorescence probe for
nitroreductase and hypoxia: selective detection and quantification. Chem Commun.
2013;49:10820–2.
40. Kiyose K, Hanaoka K, Oushiki D, et al. Hypoxia-sensitive fluorescent probes for in vivo real-
time fluorescence imaging of acute ischemia. J Am Chem Soc. 2010;132:15846–8.
41. Xu K, Wang F, Pan X, et al. High selectivity imaging of nitroreductase using near-infrared
fluorescence probe in hypoxic tumor. Chem Commun. 2013;49:2554–6.
42. Vavere AL, Lewis JS. Cu-ATSM: a radiopharmaceutical for PET imaging of hypoxia. Dalton
Trans. 2007;43:4893–902.
43. Bourgeois M, Rajerison H, Guerard F, et al. Contribution of [64Cu]-ATSM PET in molecular
imaging of tumor hypoxia compared to classical [18F]-MISO—a selected review. Nucl Med
Rev Cent East Eur. 2011;14:90–5.
44. Buscombe JR. Exploring the nature of atheroma and cardiovascular inflammation in vivo
using positron emission tomography (PET). Brit J Radiol. 2015;88:20140648.
45. Liu R-S, Chou T-K, Chang C-H, et al. Biodistribution, pharmacokinetics and PET imaging of
[18F]FMISO, [18F]FDG and [18F]FAc in a sarcoma- and inflammation-bearing mouse model.
Nucl Med Biol. 2009;36:305–12.
46. Stoll G, Bendszus M. Imaging of inflammation in the peripheral and central nervous system
by magnetic resonance imaging. Neuroscience. 2008;158:1151–60.
47. Lindner JR, Song J, Xu F, et al. Noninvasive ultrasound imaging of inflammation using
microbubbles targeted to activated leukocytes. Circulation. 2000;102:2745–50.
48. Ruiz-Cabello J, Barnett BP, Bottomley PA, et al. Fluorine (19F) MRS and MRI in biomedi-
cine. NMR Biomed. 2011;24:114–29.
49. Hu L, Hockett FD, Chen J, et al. A generalized strategy for designing 19F/1H dual-frequency
MRI coil for small animal imaging at 4.7 Tesla. J Magn Reson Imaging. 2011;34:245–52.
50. Ahrens ET, Young W-B, Xu H, et al. Rapid quantification of inflammation in tissue samples
using perfluorocarbon emulsion and fluorine-19 nuclear magnetic resonance. BioTechniques.
2011;50:229–34.
7 Multimodality Preclinical Imaging in Inflammatory Diseases 155

51. Jacoby C, Temme S, Mayenfels F, et al. Probing different perfluorocarbons for in vivo inflam-
mation imaging by 19F MRI: image reconstruction, biological half-lives and sensitivity.
NMR Biomed. 2014;27:261–71.
52. Kadayakkara DK, Ranganathan S, Young WB, et al. Assaying macrophage activity in a murine
model of inflammatory bowel disease using fluorine-19 MRI. Lab Investig. 2012;92:636–45.
53. Balducci A, Helfer BM, Ahrens ET, et al. Visualizing arthritic inflammation and therapeu-
tic response by fluorine-19 magnetic resonance imaging (19F MRI). J Inflamm. 2012;9:24.
https://doi.org/10.1186/1476-9255-9-24.
54. Weise G, Basse-Luesebrink TC, Wessig C, et al. In vivo imaging of inflammation in the
peripheral nervous system by 19F MRI. Exp Neurol. 2011;229:494–501.
55. Pike VW, Halldin C, Crouzel C, et al. Radioligands for PET studies of central benzodiazepine
receptors and PK (peripheral benzodiazepine) binding sites—current status. Nucl Med Biol.
1993;20:503–25.
56. Venneti S, Lopresti BJ, Wiley CA. The peripheral benzodiazepine receptor (translocator pro-
tein 18kDa) in microglia: from pathology to imaging. Prog Neurobiol. 2006;80:308–22.
57. Ching ASK, Kuhnast B, Damont A, et al. Current paradigm of the 18-kDa translocator pro-
tein (TSPO) as a molecular target for PET imaging in neuroinflammation and neurodegenera-
tive diseases. Insights Imaging. 2012;3:111–9.
58. Turkheimer FE, Rizzo G, Bloomfield PS, et al. The methodology of TSPO imaging with
positron emission tomography. Biochem Soc Trans. 2015;43:586–92.
59. Vivash L, O’Brien TJ. Imaging microglial activation with TSPO PET: lighting up neurologi-
cal disease? J Nucl Med. 2016;57:165–8.
60. Chauveau F, Boutin H, Van Camp N, et al. Nuclear imaging of neuroinflammation: a compre-
hensive review of [11C]PK11195 challengers. Eur J Nucl Med Mol Imaging. 2008;35:2304–19.
61. Kreisl WC, Fujita M, Fujimara Y, et al. Comparison of [11C]-(R)-PK 11195 and [11C]PBR28,
two radioligands for translocator protein (18 kDa) in human and monkey: implications
for positron emission tomographic imaging of this inflammation biomarker. NeuroImage.
2009;49:2924–32.
62. Owen DR, Yeo AJ, Gunn RN, et al. An 18-kDa translocator protein (TSPO) polymorphism
explains differences in binding affinity of the PET radioligand PBR28. J Cereb Blood Flow
Metab. 2012;32:1–5.
63. Rojas C, Stathis M, Coughlin JM, et al. The low-affinity binding of second generation radio-
tracers targeting TSPO is associated with a unique allosteric binding site. J Neuroimmune
Pharmacol. 2018;13:1–5.
64. Bunnett NW. Protease-activated receptors: how proteases signal to cells to cause inflamma-
tion and pain. Semin Thromb Hemost. 2006;32(Suppl 1):39–48.
65. Klohs J, Baeva N, Steinbrink J, et al. In vivo near-infrared fluorescence imaging of
matrix metalloproteinase activity after cerebral ischemia. J Cereb Blood Flow Metab.
2009;29:1284–92.
66. Leahy AA, Esfahani SA, Foote AT, et al. Following the trajectory of osteoarthritis develop-
ment through serial near infrared fluorescence imaging of MMP activities. Arthritis Rheum.
2015;67:442–53.
67. Matusiak N, Waarde A, Bischoff R, et al. Probes for non-invasive matrix metalloproteinase-
targeted imaging with PET and SPECT. Curr Pharmaceutical Design. 2013;19:4647–72.
68. Wagner S, Breyholz H-J, Faust A, et al. Molecular imaging of matrix metalloproteinases in vivo
using small molecule inhibitors for SPECT and PET. Curr Med Chem. 2006;13:2819–38.
69. Bordenave T, Helle M, Beau F, et al. Synthesis and in vitro and in vivo evaluation of MMP-12
selective optical probes. Biconjugate Chem. 2016;27(10):2407–17. https://doi.org/10.1021/
acs.bioconjchem.6b00377.
70. Qin H, Zhao Y, Zhang J, et al. Inflammation-targeted gold nanorods for intravascular photo-
acoustic imaging detection of matrix metalloproteinase-2 (MMP2) in atherosclerotic plaques.
Nanomedicine. 2016;12:1765–74.
71. Tavare R, McCracken MN, Zettlitz KA, et al. Engineered antibody fragments for immune-
PET imaging of endogenous CD8+ T cells in vivo. Proc Natl Acad Sci. 2014;111:1108–13.
156 P. D. Acton

72. Tavare R, Escuin-Ordinas H, Mok S, et al. An effective immune-PET imaging method to

monitor CD8-dependent responses to immunotherapy. Cancer Res. 2015;76:73–82.
73. Klohs J, Grafe M, Graf K, et al. In vivo imaging of the inflammatory receptor CD40 after
cerebral ischemia using a fluorescent antibody. Stroke. 2008;39:2845–52.
74. Rashidian M, Keliher EJ, Bilate AM, et al. Noninvasive imaging of immune response. Proc
Natl Acad Sci. 2015;112:6146–51.
75. Albrecht DS, Granziera C, Hooker JM, et al. In vivo imaging of human neuroinflammation.
ACS Chem Neurosci. 2016;7:470–83.
76. Pulli B, Chen JW. Imaging neuroinflammation—from bench to bedside. J Clin Immunol.
2014;5 https://doi.org/10.4172/2155-9899.1000226.
77. Bloomfield PS, Selvaraj S, Veronese M, et al. Microglial activity in people at ultra high risk
of psychosis and in schizophrenia: an [11C]PBR28 PET brain imaging study. Am J Psychiatry.
2016;173:44–52.
78. Setiawan E, Wilson AA, Mizrahi R, et al. Role of translocator protein density, a marker
of neuroinflammation, in the brain during major depressive episodes. JAMA Psych.
2015;72:268–75.
79. Hafizi S, Tseng H-H, Rao N, et al. Imaging microglial activation in untreated first-episode
psychosis: a PET study with [18F]FEPPA. Am J Psychiatry. 2017;174(2):118–24. https://doi.
org/10.1176/appi.ajp.2016.16020171.
80. Hannestad J, DellaGioia N, Gallezot J-D, et al. The neuroinflammation marker translocator
protein is not elevated in individuals with mild-to-moderate depression: a [11C]PBR28 PET
study. Brain Behav Immun. 2013;33:131–8.
81. Kenk M, Selvanathan T, Rao N, et al. Imaging neuroinflammation in gray and white
matter in schizophrenia: an in vivo PET study with [18F]-FEPPA. Schizophr Bull.
2015;41:85–93.
82. van der Doef TF, de Witte LD, Sutterland AL, et al. In vivo (R)-[11C]PK11195 PET imag-
ing of 18kDa translocator protein in recent onset psychosis. NPJ Schizophr. 2016;2:16031.
https://doi.org/10.1038/npjschz.2016.31.
83. Golla SSV, Boellaard R, Oikonen V, et al. Quantification of [18F]DPA-714 binding in the
human brain: initial studies in healthy controls and Alzheimer’s disease patients. J Cereb
Blood Flow Metab. 2015;35:766–72.
84. Suridjan I, Pollock BG, Verhoeff NPLG, et al. In-vivo imaging of grey and white matter neu-
roinflammation in Alzheimer’s disease: a positron emission tomography study with a novel
radioligand, [18F]-FEPPA. Mol Psychiatry. 2015;20:1579–87.
85. Varrone A, Oikonen V, Forsberg A, et al. Positron emission tomography imaging of the
18-kDa translocator protein (TSPO) with [18F]FEMPA in Alzheimer’s disease patients and
control subjects. Eur J Nucl Med Mol Imaging. 2015;42:438–46.
86. Yokokura M, Terada T, Bunai T, et al. Depiction of microglial activation in aging and demen-
tia: positron emission tomography with [11C]DPA713 versus [11C](R)PK11195. J Cereb
Blood Flow Metab. 2017;37(3):877–89. https://doi.org/10.1177/0271678X16646788.
87. Castanon N, Luheshi G, Laye S. Role of neuroinflammation in the emotional and cognitive
alterations displayed by animal models of obesity. Front Neurosci. 2015;9:229. https://doi.
org/10.3389/fnins.2015.00229.
88. Chiu CC, Liao YE, Yang LY, et al. Neuroinflammation in animal models of traumatic brain
injury. J Neurosci Methods. 2016;272:38–49. https://doi.org/10.1016/j.jneumeth.2016.06.018.
89. Nazem A, Sankowski R, Bacher M, et al. Rodent models of neuroinflammation for
Alzheimer’s disease. J Neuroinflammation. 2015;12:74. https://doi.org/10.1186/
s12974-015-0291-y.
90. Biesmans S, Acton PD, Cotto C, et al. Effect of stress and peripheral immune activation on
astrocyte activation in transgenic bioluminescent Gfap-luc mice. Glia. 2015;63:1126–37.
91. Briard E, Zoghbi SS, Simeon FG, et al. Single-step high yield radiosynthesis and evaluation
of a sensitive 18F-labeled ligand for imaging brain peripheral benzodiazepine receptors with
PET. J Med Chem. 2009;52:688–99.
92. Buck JR, McKinley ET, Hight MR, et al. Quantitative, preclinical PET imaging of TSPO
expression in glioma using [18F]PBR06. J Nucl Med. 2011;52:107–14.
7 Multimodality Preclinical Imaging in Inflammatory Diseases 157

93. Fujimaru Y, Zoghbi SS, Simeon FG, et al. Quantification of translocator protein (18
kDa) in the human brain with PET and a novel radioligand (18)F-PBR06. J Nucl Med.
2009;50:1047–53.
94. Iamaizumi M, Briard E, Zoghbi SS, et al. Kinetic evaluation in nonhuman primates of two new
PET ligands for peripheral benzodiazepine receptors in brain. Synapse. 2007;61:595–605.
95. Lartey FM, Ahn G-O, Shen B, et al. PET imaging of stroke-induced neuroinflammation in
mice using [18F]PBR06. Mol Imaging Biol. 2014;16:109–17.
96. Dedeurwaerdere S, Callaghan PD, Pham T, et al. PET imaging of brain inflammation during
early epileptogenesis in a rat model of temporal lobe epilepsy. EJNMMI Res. 2012;2:60.
https://doi.org/10.1186/2191-219X-2-60.
97. Wolf OT, Dyakin V, Patel A, et al. Volumetric structural magnetic resonance imaging (MRI)
of the rat hippocampus following kainic acid (KA) treatment. Brain Res. 2002;934:87–96.
98. Zhu L, Ramboz S, Hewitt D, et al. Non-invasive imaging of GFAP expression after neuronal
damage in mice. Neurosci Lett. 2004;367:210–2.
99. Biesmans S, Bouwknecht JA, Ver Donck L, et al. Peripheral administration of tumor necro-
sis factor-alpha induces neuroinflammation and sickness but not depressive-like behavior in
mice. Biomed Res Int. 2015;2015:716920. https://doi.org/10.1155/2105/716920.
100. Biesmans S, Meert TF, Bouwknecht JA, et al. Systemic immune activation leads to neuroin-
flammation and sickness behavior in mice. Mediat Inflamm. 2013;2013:271359. https://doi.
org/10.1155/2013/271359.
101. Cordeau P Jr, Lalancette-Hebert M, Weng YC, et al. Live imaging of neuroinflamma-
tion reveals sex and estrogen effects on astrocyte response to ischemic injury. Stroke.
2008;39:935–42.
102. Luo J, Elwood F, Britschgi M, et al. Colony-stimulating factor 1 receptor (CSF1R) signaling
in injured neurons facilitates protection and survival. J Exp Med. 2013;210:157–72.
103. Golden GT, Smith GG, Ferraro TN, et al. Rat strain and age differences in kainic acid induced
seizures. Epilepsy Res. 1995;20:151–9.
104. Ferraro TN, Golden GT, Smith GG, et al. Mapping murine loci for seizure response to kainic
acid. Mamm Genome. 1997;8:200–8.
105. McKhann GM, Wenzel HJ, Robbins CA, et al. Mouse strain differences in kainic acid sensitiv-
ity, seizure behavior, mortality, and hippocampal pathology. Neuroscience. 2003;122:551–61.
106. Smolen JS, Aletaha D, McInnes IB. Rheumatoid arthritis. Lancet. 2016;388(10055):P2023–
38. https://doi.org/10.1016/S0140-6736(16)30173-8.
107. Put S, Westhovens R, Lahoutte T, et al. Molecular imaging of rheumatoid arthritis: emerging
markers, tools, and techniques. Arthritis Res Ther. 2014;16:208.
108. Wang S-C, Xie Q, Lv W-F. Positron emission tomography/computed tomography imaging
and rheumatoid arthritis. Int J Rheum Dis. 2014;17:248–55.
109. Wunder A, Straub RH, Gay S, et al. Molecular imaging: novel tools in visualizing rheumatoid
arthritis. Rheumatology. 2005;44:1341–9.
110. Borrero CG, Mountz JM, Mountz JD. Emerging MRI methods in rheumatoid arthritis. Nat
Rev Rheumatol. 2011;7:85–95.
111. Rogers JL, Tarrant T, Kim J. Nanoparticle-based diagnostic imaging of inflammation in rheu-
matoid disease. Curr Rheumatol Rev. 2014;10:3–10.
112. Clavel G, Marchiol-Fournigault C, Renault G, et al. Ultrasound and Doppler micro-imaging
in a model of rheumatoid arthritis. Ann Rheum Dis. 2008;67:1765–72.
113. McInnes IB, Schett G. The pathogenesis of rheumatoid arthritis. N Engl J Med.
2011;365:2205–19.
114. Ferrari M, Onuoha SC, Pitzalis C. Trojan horses and guided missiles: targeted therapies in the
war on arthritis. Nat Rev Rheumatol. 2015;11:328–37.
115. Asquith DL, Miller AM, McInnes IB, et al. Animal models of rheumatoid arthritis. Eur J
Immunol. 2009;39:2040–4.
116. McNamee K, Williams R, Seed M. Animals models of rheumatoid arthritis: how informative
are they? Eur J Pharmacol. 2015;759:278–86.
117. Vierboom MPM, Jonker M, Tak PP, et al. Preclinical models of arthritic disease in non-
human primates. Drug Discov Today. 2007;12:327–35.
158 P. D. Acton

118. Kundu-Raychaudhuri S, Mitra A, Datta-Mitra A, et al. In vivo quantification of mouse auto-

immune arthritis by PET/CT. Int J Rheum Dis. 2016;19:452–8.
119. Matsui T, Nakata N, Nagai S, et al. Inflammatory cytokines and hypoxia contribute to 18F-
FDG uptake by cells involved in pannus formation in rheumatoid arthritis. J Nucl Med.
2009;50:920–6.
120. Cho H, Bhatti F-U-R, Yoon TW, et al. Non-invasive dual fluorescence in vivo imaging for
detection of macrophage infiltration and matrix metalloproteinase (MMP) activity in inflam-
matory arthritic joints. Biomed Optics Express. 2016;7:1842–52.
121. Terry SYA, Koenders MI, Franssen GM, et al. Monitoring therapy response of experimental
arthritis with radiolabeled tracers targeting fibroblasts, macrophages, or integrin αvβ3. J Nucl
Med. 2016;57:467–72.
122. van der Geest T, Laverman P, Gerrits D, et al. Liposomal treatment of experimental arthritis
can be monitored non-invasively with radiolabeled anti-fibroblast activation protein antibod-
ies. J Nucl Med. 2017;58(1):151–5. https://doi.org/10.2967/jnumed.116.177931.
123. Mountz JM, Alavi A, Mountz JD. Emerging optical and nuclear medicine imaging methods
in rheumatoid arthritis. Nat Rev Rheumatol. 2012;8:719–28.
124. Robertson R, Germanos MS, Li C, et al. Optical imaging of Cerenkov light generation from
positron-emitting radiotracers. Phys Med Biol. 2009;54:N355–66.
125. Thorek DLJ, Robertson R, Bacchus WA, et al. Cerenkov imaging—a new modality for
molecular imaging. Am J Nucl Med Mol Imaging. 2012;2:163–73.
126. Irmler IM, Gebhardt P, Hoffmann B, et al. 18F-fluoride positron emission tomography/com-
puted tomography for noninvasive in vivo quantification of pathophysiological bone metabo-
lism in experimental murine arthritis. Arthritis Res Ther. 2014;16:R155.
127. Barck KH, Lee WP, Diehl LJ, et al. Quantification of cortical bone loss and repair for thera-
peutic evaluation in collagen-induced arthritis, by micro-computed tomography and auto-
mated image analysis. Arthritis Rheum. 2004;50:3377–86.
128. Sevilla RS, Cruz F, Chiu C-S, et al. Development and optimization of a high-throughput
micro-computed tomography imaging method incorporating a novel analysis technique to
evaluate bone mineral density of arthritic joints in a rodent model of collagen induced arthri-
tis. Bone. 2015;73:32–41.
129. Silva MD, Ruan J, Siebert E, et al. Application of surface roughness analysis on micro-
computed tomography images of bone erosion: examples using a rodent model of rheumatoid
arthritis. Mol Imaging. 2006;5:475–84.
130. Silva MD, Savinainen A, Kapadia R, et al. Quantitative assessment of micro-CT imaging and
histopathological signatures of experimental arthritis in rats. Mol Imaging. 2004;3:312–8.
131. Abraham C, Cho JH. Inflammatory bowel disease. N Engl J Med. 2009;361:2066–78.
132. de Souza HSP, Fiocchi C. Immunopathogenesis of IBD: current state of the art. Nat Rev
Gastroenterol Hepatol. 2016;13:13–27.
133. Johnson CM, Wei C, Ensor JE, et al. Meta-analysis of colorectal cancer risk factors. Cancer
Causes Control. 2013;24:1207–22.
134. Triantafillidis JK, Nasioulas G, Kosmidis PA. Colorectal cancer and inflammatory bowel dis-
ease: epidemiological, risk factors, mechanisms of carcinogenesis and prevention strategies.
Anticancer Res. 2009;29:2727–37.
135. Ananthakrishnan AN, Cagan A, Cai T, et al. Colonoscopy is associated with a reduced risk for
colon cancer and mortality in patients with inflammatory bowel diseases. Clin Gastroenterol
Hepatol. 2015;13:322–9.
136. Al-Hawary MA, Kaza RK, Platt JF. CT enterography: concepts and advances in Crohn’s
disease imaging. Radiol Clin N Am. 2013;51:1–16.
137. Beltzer A, Kaulisch T, Bluhmki T, et al. Evaluation of quantitative imaging biomarkers in the
DSS colitis model. Mol Imaging Biol. 2016;18:697–704.
138. Gee MS, Harisinghani MG. MRI in patients with inflammatory bowel disease. J Magn Reson
Imaging. 2011;33:527–34.
139. Kilcoyne A, Kaplan JL, Gee MS. Inflammatory bowel disease imaging: current practice and
future directions. World J Gastroenterol. 2016;22:917–32.
7 Multimodality Preclinical Imaging in Inflammatory Diseases 159

140. Masselli G, Gualdi G. CT and RM enterography in evaluating small bowel diseases: when to
use which modality? Abdom Imaging. 2013;38:249–59.
141. Mentzel H-J, Reinsch S, Kurzai M, et al. Magnetic resonance imaging in children
and adolescents with chronic inflammatory bowel disease. World J Gastroenterol.
2014;20:1180–91.
142. Panes J, Bouhnik Y, Reinisch W, et al. Imaging techniques for assessment of inflammatory
bowel disease: joint ECCO and ESGAR evidence-based consensus guidelines. J Crohn's
Colitis. 2013;7:556–85.
143. Perlman SB, Hall BS, Reichelderfer M. PET/CT imaging of inflammatory bowel disease.
Semin Nucl Med. 2013;43:420–6.
144. Ream JM, Dillman JR, Adler J, et al. MRI diffusion-weighted imaging (DWI) in pediatric
small bowel Crohn disease: correlation with MRI findings of active bowel wall inflammation.
Pediatr Radiol. 2013;43:1077–85.
145. Saverymuttu SH, Peters AM, Hodgson HJ, et al. Indium-111 autologous leucocyte scanning:
comparison with radiology for imaging the colon in inflammatory bowel disease. Brit Med J.
1982;285:255–7.
146. Treglia G, Quartuccio N, Sadeghi R, et al. Diagnostic performance of fluorine-18-fluoro-
deoxyglucose positron emission tomography in patients with chronic inflammatory bowel
disease: a systematic review and meta-analysis. J Crohn's Colitis. 2013;7:345–54.
147. Kozarek R, Chiorean M, Wallace M, editors. Endoscopy in inflammatory bowel disease.
Heidelberg: Springer; 2015.
148. Burggraaf J, Kamerling IMC, Gordon PB, et al. Detection of colorectal polyps in humans
using an intravenously administered fluorescent peptide targeted against c-Met. Nat Med.
2015;21:955–61.
149. Knieling F, Waldner MJ. Light and sound—emerging techniques for inflammatory bowel
disease. World J Gastroenterol. 2016;22:5642–54.
150. Liu J, Dlugosz A, Neumann H. Beyond white light endoscopy: the role of optical biopsy in
inflammatory bowel disease. World J Gastroenterol. 2013;19:7544–51.
151. Subramanian V, Ragunath K. Advanced endoscopic imaging: a review of commercially avail-
able technologies. Clin Gastroenterol Hepatol. 2014;12:368–76.
152. Coda S, Thillainayagam AV. State of the art in advanced endoscopic imaging for the detec-
tion and evaluation of dysplasia and early cancer of the gastrointestinal tract. Clin Exp
Gastroenterol. 2014;7:133–50.
153. Kumashiro R, Konishi K, Chiba T, et al. Integrated endoscopic system based on optical
imaging and hyperspectral data analysis for colorectal cancer detection. Anticancer Res.
2016;36:3925–32.
154. Pence I, Mahadevan-Jansen A. Clinical instrumentation and applications of Raman spectros-
copy. Chem Soc Rev. 2016;45:1958–79.
155. Tontini GE, Rath T, Neumann H. Advanced gastrointestinal endoscopic imaging for inflam-
matory bowel disease. World J Gastroenterol. 2016;22:1246–59.
156. Wu Y, Briley K, Tao X. Nanoparticle-based imaging of inflammatory bowel disease.
NanoBiotechnology. 2016;8:300–15.
157. Goyal N, Rana A, Ahlawat A, et al. Animal models of inflammatory bowel disease: a review.
Inflammopharmacology. 2014;22:219–33.
158. Kolios G. Animal models of inflammatory bowel disease: how useful are they really? Curr
Opin Gastroenterol. 2016;32:251–7.
159. Neurath MF. Animal models of inflammatory bowel diseases: illuminating the pathogenesis
of colitis, ileitis and cancer. Dig Dis. 2012;30(suppl 1):91–4.
160. Valatas V, Bamias G, Kolios G. Experimental colitis models: insights into the pathogenesis of
inflammatory bowel disease and translational issues. Eur J Pharmacol. 2015;759:253–64.
161. Liu T-C, Stappenbeck TS. Genetics and pathogenesis of inflammatory bowel disease. Annu
Rev Pathol. 2016;11:127–48.
162. Mizoguchi A, Takeuchi T, Himuro H, et al. Genetically engineered mouse models for study-
ing inflammatory bowel disease. J Pathol. 2015;238:205–19.
160 P. D. Acton

163. Valatas V, Vakas M, Kolios G. The value of experimental models of colitis in predicting the
efficacy of biological therapies for inflammatory bowel diseases. Am J Physiol Gastrointest
Liver Physiol. 2013;305:G763–85.
164. Ding S, Blue RE, Morgan DR, et al. Comparison of multiple enzyme activatable near-infra-
red fluorescent molecular probes for detection and quantification of inflammation in murine
colitis models. Inflamm Bowel Dis. 2014;20:363–77.
165. Mitsunaga M, Kosaka N, Choyke PL, et al. Fluorescence endoscopic detection of murine
colitis-associated colon cancer by topically applied enzymatically rapid-activatable probe.
Gut. 2013;62:1179–86.
166. Bernards N, Pottier G, Theze B, et al. In vivo evaluation of inflammatory bowel disease with
the aid of μPET and the translocator protein 18 kDa radioligand [18F]DPA-714. Mol Imaging
Biol. 2015;17:67–75.
167. Heijink DM, Kleibeuker JH, Nagengast WB, et al. Total abdominal 18F-FDG uptake reflects
intestinal adenoma burden in Apc mutant mice. J Nucl Med. 2011;52:431–6.
168. Tietz O, Wuest M, Marshall A, et al. PET imaging of cyclooxygenase-2 (COX-2) in a
pre-clinical colorectal cancer model. EJNMMI Res. 2016;6:37. https://doi.org/10.1186/
s13550-016-0192-9.
169. Atreya R, Neumann H, Neufert C, et al. In vivo imaging using fluorescent antibodies to tumor
necrosis factor predicts therapeutic response in Crohn’s disease. Nat Med. 2014;20:313–8.
170. Hua S, Marks E, Schneider JJ, et al. Advances in oral nano-delivery systems for colon tar-
geted drug delivery in inflammatory bowel disease: selective targeting to diseased versus
healthy tissue. Nanomedicine. 2015;11:1117–32.
171. Karrout Y, Dubuquoy L, Piveteau C, et al. In vivo efficacy of microbiota-sensitive coatings
for colon targeting: a promising tool for IBD therapy. J Control Release. 2015;197:121–30.
172. Takedatsu H, Mitsuyama K, Torimura T. Nanomedicine and drug delivery strategies for treat-
ment of inflammatory bowel disease. World J Gastroenterol. 2015;21:11343–52.
173. Zhang S, Emann J, Zhou A, et al. An inflammation-targeting hydrogel for local drug delivery
in inflammatory bowel disease. Sci Transl Med. 2015;7:300ra128. https://doi.org/10.1126/
scitranslmed.aaa5657.
174. Ali H, Weigmann B, Collnot E-M, et al. Budesonide loaded PLGA nanoparticles for targeting
the inflamed intestinal mucosa—pharmaceutical characterization and fluorescence imaging.
Pharm Res. 2016;33:1085–92.
175. Jin M, Yu D-G, Geraldes CFGC, et al. Theranostic fibers for simultaneous imaging and drug
delivery. Mol Pharm. 2016;13:2457–65.
Preclinical Multimodality Imaging
and Image Fusion in Cardiovascular 8
Disease

James T. Thackeray

8.1 Introduction

The rapid expansion and deployment of multimodality imaging instrumentation

including PET-CT, SPECT-CT, and more recently PET/MR has resulted in a wide
range of image fusion possibilities for cardiovascular imaging [1]. In general, these
fusion approaches have been limited to combining anatomic imaging with physio-
logic imaging modalities, particularly the fusion of CT or MR images with PET or
SPECT images. Indeed, the combination of nuclear myocardial perfusion imaging
with CT coronary angiography has become common in clinical practice [2]. In addi-
tion, combination of multiple radiotracers targeted to distinct physiologic pathways
is more frequently applied to obtain complementary information about pathophysi-
ology of cardiovascular disease, similar to clinical application combining perfusion
and viability [3]. The principle challenge to impage fusion in preclinical applica-
tions is the frequency of image acquisition using separate imaging systems with
different spatial resolution, animal positioning, anesthesia connections, gating capa-
bilities, and dataset file type. Software packages provided with individual cameras
allow for image fusion between modalities but often require conversion of the image
file for semiautomated or manual co-registration. The same operations may be nec-
essary for third-party software packages [4]. Nonetheless, image fusion bears sig-
nificant potential to clarify regional image analysis and maximize the capacity of
molecular imaging approaches in cardiovascular disease models. Moreover, multi-
modality imaging probes including nanoparticles and other synthetic nanoparticle-
like structures have gained prominence over the last decade, which may have
significant impact in refining image analysis and fusion tools. In this chapter, strate-
gies and techniques for image fusion between modalities and systems will be
broadly discussed in the context of cardiovascular disease, with particular attention

J. T. Thackeray (*)
Department of Nuclear Medicine, Hannover Medical School, Hannover, Germany
e-mail: [email protected]

© Springer Nature Switzerland AG 2019 161

C. Kuntner-Hannes, Y. Haemisch (eds.), Image Fusion in Preclinical
Applications, https://doi.org/10.1007/978-3-030-02973-9_8
162 J. T. Thackeray

paid to the relevant application of these techniques to optimize the complementary

nature of sequential or simultaneous physiologic and anatomic imaging.

8.2 PET-CT

8.2.1 Localization of Signal

The primary challenge in cardiovascular imaging is the localization of a cardiac

signal without requisite anatomic information. This challenge is particularly signifi-
cant in cases where the imaging target does not encompass the complete myocar-
dium. Accordingly, approaches to visualize myocardial space in concert with the
PET image are desirable. In some cases, this may be accomplished by fusing the
PET image to a fast CT scan, providing sufficient anatomic contrast to properly
localize the PET signal.
The simplest approach for cardiac localization is the alignment of the heart within
the boundaries of the lung and diaphragm, which can be identified using low contrast
fast CT. Vascular activity can be localized by contrast-enhanced CT acquisitions, illu-
minating the vessel lumen to compare with tracer distribution in the aorta, carotid, or
femoral arteries. Vascular CT contrast agents used in preclinical applications tend to
be iodine- or metal-rich synthetic compounds and provide exquisite definition of the
arterial and venous blood pool [5, 6]. Coronary vasculature represents a formidable
challenge due to the dual factors of cardiac and respiratory motion. While some
motion and partial volume correction algorithms have been introduced for clinical
application [7], use in small animals with faster heart rates has not been evaluated.
Blood pool contrast agents also have the benefit of improved soft tissue contrast,
allowing for clearer definition of cardiac anatomy and left ventricle contours.

8.2.2 Fusion

There are several commercially available software packages dedicated to the fusion
of multimodality images. Such software packages include the vendor-supplied soft-
ware installed with the multimodal cameras (e.g., Siemens Inveon Research
Workplace (IRW) and 3D Image Viewer), clinical or preclinical licensed image
analysis software packages (e.g., PMOD, Hermes Hybrid Viewer, Syntegra,
InVivoScope), or open-source image analysis tools (e.g., AMIDE, OsiriX) [4, 8].
The choice of software depends on the image file formats generated, available com-
puting power and operating systems, the desired applications, and operator prefer-
ence. For most preclinical cardiovascular applications, the vendor-provided analysis
software is sufficient for basic fusion techniques. In cases with multiple fused data-
sets (i.e., >3 modalities or tracers) or for vascular fusion, more sophisticated soft-
ware may be required. If segmental data is desired from the datasets, optimal
software can provide a polar map based on contour detection, ray detection, and/or
sampling point definitions (e.g., Munich Heart, FlowQuant, Segment) [9–12].
8 Preclinical Multimodality Imaging and Image Fusion in Cardiovascular Disease 163

8.2.3 Transformation

PET images can be rigidly matched to the CT template, relying on anatomic land-
marks to align the images. This process is simplified with the full body in the field
of view for both PET and CT, as can be achieved in mice with a single bed position
for acquisition [13, 14]. Lowering the activity threshold display allows the boundar-
ies of the animal thoracic cavity to be visualized, which can then be easily matched
to the boundaries of the animal defined by CT. The geometric coordinates of the
PET image are realigned to the CT space, allowing for a fused image to be gener-
ated. Limitations in the soft tissue contrast of standard fast-sequence small animal
CT can limit the effectiveness of automated or manual fusion of radionuclide images
to the CT image, particularly when the field of view is limited.
In cases with low PET signal, sequential administration of additional tracers can
be of benefit. For instance, subsequent injection of 18F-sodium fluoride (18F-NaF)
may be used to obtain a bone PET map that is more easily fused to the skeleton
defined on CT (Fig. 8.1). If the animal positioning in the PET acquisition is not
modified between administrations of the two tracers, the transformation can be
cross-applied to the original image. Similar approaches can be applied using bone
imaging with SPECT-CT. Khmelinskii et al. demonstrated segmentation and visual

CT NaF FDG

Rigid Matching Cross-Apply

Transformation
CT

NaF

FDG

NaF 0 1 FDG 0 0.75

MBq /cc MBq /cc

Fig. 8.1 PET-CT image fusion. Three-step procedure involves first the fusion of the 18F-fluoride
PET image to sequential fast CT scan and subsequent application of transformation coordinates to
the inflammation 18F-FDG PET image. Rigid matching automatically matches the bone signal of
18
F-fluoride to the skeleton defined on CT as shown in coronal and sagittal slices. The generated
transformation is then cross-applied to the 18F-FDG image acquired earlier in the sequence. Image
fusion carried out using Siemens Inveon Research Workplace (IRW) 3.0 software
164 J. T. Thackeray

analysis of mouse bone structures using SPECT-CT with 99mTc-methylene diphos-

phonate. SPECT distribution data provided the basic skeleton, which was initially
registered to the full mouse atlas and then more accurately registered by individual
bone using the iterative closest point approach [15]. The articulated planar reforma-
tion algorithm then reformatted the segmented data into the mouse atlas to map
space [15]. The transformation can then be cross-applied to a cardiac-specific
image, enabling appropriate image fusion.

8.2.4 Co-registration for Multi-camera

Difficulties for moving between cameras may be compounded when different bed
systems and anesthesia connections are present. Optimally, the preclinical imaging
facility employs a common bed system, allowing for consistent bed alignment
between cameras, though this is not always the case. Customized beds that can be
moved between cameras are desirable and are available from some third-party ven-
dors. Alternatively, animal-holding inserts that can be moved back and forth between
the vendor-supplied bed systems may overcome this complication [16].

8.2.5 Fiducial Markers

A number of additional approaches may be employed to facilitate the fusion of

images acquired using separate cameras. Fiducial markers containing CT-visualized
material may be further filled with a small amount of radioactivity to visualize the
marker using multiple modalities [17]. A point source placed on the scanning bed or
insert that can be transferred between cameras facilitates three-dimensional co-
localization of the signal between images. Such a point source may include polyeth-
ylene tubing or glass capillary tubes filled with radioactivity. These fiducial markers
should be fixed to the bed and are optimally placed in multiple directions to facili-
tate rigid matching, without interfering with the targeted PET signal. For example,
external fiducial markers optimized the positioning of the animal and bed insert
between two separate dedicated scanners for sequential PET and CT acquisition.
Using the point-source fiducial marker as a guide, contrast (Exia)-enhanced CT
images were fused to the 18F-FDG acquisition to define the localization of the vena
cava to calculate the venous input function [18].

8.3 SPECT-CT

Many of the same principles used for PET-CT fusion can be applied for SPECT-CT
image fusion. Image transformation and the use of fiducial markers are key factors
in obtaining optimal image fusion between SPECT and CT images acquired sequen-
tially or using separate cameras (Fig. 8.2).
8 Preclinical Multimodality Imaging and Image Fusion in Cardiovascular Disease 165

Coronal Sagittal Transaxial

Contrast
Enhanced
CT

Fuse

4000

99mTc-MIBI

Pinhole Arb.
SPECT Units

100

Fig. 8.2 SPECT-CT image fusion. Contrast-enhanced CT images acquired immediately after
administration of alkaline earth metal-based nanoparticle contrast agent (ExiTron nano 12,000)
show high soft tissue contrast to delineate the left ventricle and ischemic region (light band) 3d
after transient coronary artery occlusion. Subsequent acquisition of a perfusion SPECT study
using pinhole collimation is automatically co-registered to the CT from the fused system. 99mTc-
sestamibi accumulates in the perfused myocardium, providing clear delineation of the perfusion
defect in the infarct territory. Image fusion conducted using open-source AMIDE software

8.3.1 Localization of Signal

Fusion of SPECT images to CT may be complicated depending on the type of col-

limation employed for the SPECT acquisition. When using slit collimation, the
positioning of individual organs is rendered more challenging due to poor image
contrast and limited resolution of the image. Alternatively, improved image resolu-
tion using pinhole collimation severely limits the field of view, which may compli-
cate precise co-registration with CT images. When using fused camera systems, the
SPECT and CT images are automatically co-registered and require only minor post-
acquisition manipulation to match the images to account for animal movement. The
principles of alignment and co-registration described for PET above are translatable
for SPECT image fusion [19].
166 J. T. Thackeray

8.3.2 Fusion

The image fusion within the limited space can be challenging and requires a suffi-
cient field of view to co-localize to the CT image. The CT acquisition should exceed
the boundaries of the pinhole SPECT field of view. O’Neill et al. used an antibody
approach to image CD169-positive macrophages labeled with 99mTc-pertechnetate.
SPECT-CT images were analyzed using InVivoScope software and fused manually.
Identification of the liver, spleen, and bone marrow was possible from fusion with
CT images [20].

8.3.3 Transformation

Image transformation is achieved in a similar manner to PET-CT imaging, where

the positioning of the SPECT image is adjusted based on rigid matching and fine-
tuned with manual adjustment to fit the CT boundaries. Whether limited to the posi-
tioning of the heart within the boundaries of the lungs and diaphragm or employing
higher resolution CT with or without contrast agent for soft tissue definition, rigid
matching and manual adjustment perform adequately. Bone imaging agents can
provide additional assistance in rigid matching [15] and may employ isotopes of
different energy spectra so as not to compromise the target image. Image transfor-
mations can be saved in vendor-provided or third-party software and loaded with
the image files to achieve fused images.

8.3.4 Fiducial Markers

Fiducial markers are applied similar to PET, with the benefit of a broader spectrum
of detected activity such that different isotopes can be used for fiducial markers that
are not visible in the target image reconstruction. As such, a different isotope can be
used for the fiducial marker without concern for contaminating the SPECT images,
regardless of the proximity to the animal in the imaging bed.

8.4  ulti-isotope Fusion: PET-PET, PET-SPECT,

M
and SPECT-SPECT

Short-lived PET isotopes, different energy spectra, and complementary radiotracers

open an additional category of image fusion in cardiology, namely, multi-isotope or
multi-tracer fusion. Imaging protocols are designed to take advantage of these char-
acteristics. Moreover, the emergence of next generation SPECT cameras featuring
enhanced energy resolution has enabled the simultaneous acquisition and isolation
of energy spectra from multiple isotopes. Clear energy resolution between thal-
lium-201 (68 keV), technetium-99m (140 keV), iodine-123 (159 keV), and
8 Preclinical Multimodality Imaging and Image Fusion in Cardiovascular Disease 167

indium-111 (171 keV) allows for multi-isotope acquisition [21]. Injected activity
may be administered sequentially or in a mixed bolus to minimize venous damage
and to reduce variability in tracer delivery.

8.4.1 Regional Mismatch

In cardiology, the primary reasoning for multi-isotope studies is to isolate multiple

physiological processes in a single imaging session. Typical imaging protocols
would combine a molecular marker with a perfusion agent to provide definition of
myocardial contours and coronary vascular territories. Combination of perfusion
with glucose metabolism using 18F-FDG allows the delineation of matched defects
(scar) from regions characterized by viable myocardium and hypoperfusion,
termed hibernating myocardium [22]. Polar map analysis in properly fused PET-
PET or PET-SPECT images facilitates regional analysis of these mismatch territo-
ries. The same principle has been applied for identifying altered myocardial
sympathetic nervous integrity [23, 24] or elevated tissue inflammation following
myocardial infarction or in chronic heart failure [9, 10, 25]. PET-PET fusion
images rely on the short half-life of the initial imaging agent (e.g., 13N-ammonia or
15
O-water for perfusion) and subsequent or delayed imaging using an agent with
the longer half-life (e.g., 11C-acetate or 18F-FDG). Optimally, six half-lives should
be allowed to pass between image acquisitions, but with a higher injected activity
of the second tracer, the duration between scans can be shortened depending on the
expected signal strength and distribution of the imaging agent. Movement of the
subject between acquisitions, due to ambient settling during sustained anesthesia,
injection of the second agent, or transfer between cameras in the case of PET-
SPECT fusion, should be adjusted for when co-registering the acquired images.
Dual isotope SPECT has the added benefit of acquiring images simultaneously,
such that the images are perfectly fused without body motion. Reconstructions
applying appropriate scatter windows can ideally isolate each isotope signal with-
out loss of counts, especially when using next generation cameras equipped with
cadmium-zinc-telluride (CZT) detectors.
Co-localization of an equivocal PET signal may be improved with a secondary
PET image acquired with higher activity administration. For example, Thackeray
et al. used 68Ga-pentixafor (~10 MBq) to image CXCR4 in activated leukocytes in
the ischemic territory of myocardial infarction mouse hearts. To localize the signal,
18
F-FDG (~25 MBq) was injected under isoflurane anesthesia to obtain the myocar-
dial contours in identical position. Polar map analysis using matched sampling
points facilitated the interpretation of regional distribution of the inflammation
tracer with a wide positron range and low background [25]. When using 18F-FDG
under ketamine-xylazine anesthesia to suppress cardiomyocyte 18F-FDG uptake, a
perfusion PET scan using 13N-ammonia ~30 min prior to the administration of 18F-
FDG allowed for definition of myocardial contours and subsequent fusion of perfu-
sion and inflammation signal [9]. The relative signal intensities of the perfusion or
168 J. T. Thackeray

cardiomyocyte PET signal being at least an order of magnitude higher than the
inflammatory cell PET signal, there is limited cross contamination of the signal. The
benefit of precise image fusion facilitates placement of regions of interest for the
molecular radioligand image.

8.4.2 Reporter-Perfusion

These approaches have particular benefit for tracking of labeled autologous or

exogenous cells for inflammation imaging or assessing experimental cell therapy.
Direct cell labeling may be achieved by 18F-FDG, 111In-oxine, or 99mTc-HMPAO,
allowing cell tracking with PET or SPECT. Alternatively, cells may be genetically
modified to express a reporter gene that can be imaged using targeted agents. The
most commonly used reporter gene constructs are the sodium iodide symporter
(NIS) and the herpes simplex virus thymidine kinase (HSV-tk), imaged using
99m
Tc-pertechnetate/123I-NaI or 18F-FHBG/18F-FHPG, respectively [26]. The for-
mer allows for multi-isotope SPECT imaging using either 201Tl or 99mTc-based per-
fusion agents in combination with 123I-NaI and, when combined with CZT detectors,
may enable clear regional distribution of reporter gene-expressing cells to be real-
ized. Appropriate co-localization of the HSV-tk-targeted PET images may be
achieved in combination with 13N-ammonia to define the myocardial contours and
bears potential for clear spatial orientation of signal.
Dual-energy imaging was performed in inflammation-associated amyloid in
transgenic mice using a multi-isotope SPECT-CT platform. 125I-labeled serum amy-
loid P component (SAP) was injected with the synthetic polybasic peptide 99mTc-
p5+14 and acquired simultaneously in dual-energy SPECT, with a subsequent
acquisition of CT for anatomic co-registration [27]. Imaging defined higher splenic
localization of SAP versus the polybasic peptide, whereas cardiac accumulation of
the synthetic peptide was higher, suggesting the potential for imaging amyloidosis
with limited background [27].
Similarly, a herpes simplex virus thymidine kinase (HSV-tk) reporter gene
tracked transplanted mesenchymal stromal cells using 18F-FEAU over a 5-month
period. Serial PET assessed focal intramyocardial cell injection sites over time,
without clear delineation of the myocardial contours. Axial CT images of the thorax
effectively defined the myocardial location, though precise regional information
could not be noninvasively determined [28].
Comparable approaches have been applied clinically to monitor localization of
indium-111-labeled white blood cells in the setting of myocarditis [29]. CZT
SPECT imaging using 111In-leukocytes with 99mTc-sestamibi allowed for improved
reader confidence and clearer definition of the myocardial contours and valve plane
for diagnosis of localized inflammation [29]. The fundamental infrastructure for
preclinical SPECT using a CZT-equipped camera allows the same approach but
may be limited to sequential imaging in the absence of CZT detectors and their
inherent energy resolution [30].
8 Preclinical Multimodality Imaging and Image Fusion in Cardiovascular Disease 169

8.5 PET-MR-CT

PET-MR fusion imaging presents unique challenges with regard to complex acqui-
sition protocols and the potential interference of PET instrumentation with MR
acquisition. The superior soft tissue contrast afforded by MR is attractive for cardio-
vascular applications, as functional measurements are simplified without the
requirement of contrast injection as for CT. However, unlike the fused PET-CT and
SPECT-CT cameras, MR is frequently operated as a stand-alone system and gener-
ates a range of image files in different orientations which present challenges to
fusion operations. Simultaneous acquisition of PET and MR data allows for optimal
image fusion with limited subject movement and enables accurate and directly com-
parable gating of the acquired images. Nevertheless, concerns about interference
with image quality have persisted.
The feasibility of PET-MR image acquisition and fusion was assessed using a
standard 7T small animal MR scanner with a custom-built PET insert using lutetium
oxyorthosilicate (LSO) crystals and avalanche photodiodes. In mice following myo-
cardial infarction, 18F-FDG PET was performed in the center field of view with
simultaneous acquisition of serial MRI sequences. Differences in the geometric
coordinates between the cameras necessitated user interaction to manually align
images. Initial evaluation using rod phantoms provided the basic transformation
operations, but exact registration was required for subsequent animal image fusion.
AMIDE software was employed for precise co-registration of images. While simul-
taneous acquisition of PET and MR measurements was feasible, the PET insert
resulted in lower image resolution and sensitivity [31].
More recently, Weissler et al. reported a thorough investigation of PET and MR
performance in a small animal MR camera equipped with a PET insert. Initial tests
revealed a reduction of PET sensitivity inside a limited radius, with increased scat-
ter induced by RF coils [32]. Simultaneous operation of PET and MR was initially
tested using a hot-rod phantom with six hot-rod arrays with diameters and gaps of
0.8–2.0 mm filled with 18F-FDG. No difference in the PET or MR resolution was
detected for individual or simultaneous acquisition. Further testing with a multi-
nuclei phantom with a 1H/19F coil assessed the sensitivity of MR for soft tissue
contrast and allowed differentiation between air, water, isopropyl alcohol, perflu-
oro-15-crown-5-ether, and olive oil, without loss of PET sensitivity. In vivo testing
using custom-designed electrocardiogram electrodes demonstrated successful
simultaneous gating of PET and MR images in eight bins. Gated PET images
showed good image uniformity and excellent spatial and temporal alignment. By
contrast, sequential acquisition using different camera systems requires indepen-
dent gating, which results in less precise fusion in individual bins. The PET/RF
Hyperion IID insert using digital silicon photomultiplier technology exhibited capa-
bility for simultaneous PET and MR image acquisition with minimal disturbance of
the magnetic field and PET sensitivity [32]. The practicality of such a system, par-
ticularly considering differences in the required acquisition time for PET and MR,
has not yet been established, but a clear benefit for image fusion independent of
170 J. T. Thackeray

third-party software and fiducial markers offers opportunities for improvement of

image analysis, gating, and reconstruction.
Nieman et al. implemented a simple modification to 3D gradient echo sequence
to include a short delay between the start of the readout diphase gradient pulse and
the two phase-encode pulses for imaging cardiac development in utero. Serially
acquired embryo images were registered to adjust the position over the course of the
scan, using freeware from the Montreal Neurological Institute. A manually drawn
mask defined the embryo heart, which was then transformed for each phase of the
cardiac cycle [33].
The accuracy of multimodality image fusion was previously assessed using a
motion enabled targeting phantom. The phantom was molded from a human heart
with comparable mechanical properties and a detachable targeting slab insert at the
apex which can be customized for multiple targets. Image co-registration between
MR, ultrasound, and electromagnetic signal was assessed using the phantom to pro-
vide the framework for future (pre)clinical applications [34]. Taken together, these
approaches demonstrate the power of nuclear-MR image fusion approaches to com-
bine the high spatial resolution of MR with the physiologic molecular specificity of
PET or SPECT imaging.

8.6 Applications

8.6.1 Signal Localization, ROI Guidance

The primary functionality of image fusion in cardiovascular preclinical imaging is

the localization of an equivocal radiotracer signal in anatomic space. Fusion of
nuclear images with CT or MR can be used to guide and accurately define the posi-
tioning of appropriate regions of interest. For the myocardium, focused regional
activity can be assigned to the ventricular wall based on the identification of the
ventricle cavity or surrounding soft tissue. The definition of anatomy takes on higher
importance for imaging of vasculature, where contrast-enhanced CT or MR images
facilitate visualization of the vessel lumen. The aortic arch can be defined in this
manner and, in some cases, can then be segmented to obtain planes in ascending,
transverse, and descending aorta that can be compared to tissue specimens [35, 36].
For aortic aneurysm, the location of the vascular injury can be identified from the
anatomic image, and the region of interest positioned independent of the radiotracer
activity [37]. In these cases, fused images and guided ROI placement minimize
potential bias in the image analysis and increase reader confidence in the anatomic
structures.
Acquisition of 18F-FDG images followed by short CT x-ray acquisitions for
attenuation correction and image co-localization was performed in C57Bl/6 mice
with myocardial infarction and ischemic heart failure. CT successfully guided ROI
placement in 18F-FDG images, despite reduction of glucose uptake signal by ket-
amine-xylazine anesthesia or conscious tracer uptake [38]. Rigid matching, manual
fine-tuning, and system-matched CT images were also employed in order to
8 Preclinical Multimodality Imaging and Image Fusion in Cardiovascular Disease 171

accurately overlay PET images with SPECT images acquired using separate camera
systems. 11C-methionine images of macrophages in the ischemic territory of mice
after myocardial infarction were fused to subsequently acquired 99mTc-sestamibi
perfusion images (Fig. 8.3). PET images were first co-registered to a fast CT scan
acquired on the PET-CT system, and SPECT images were co-registered to an inde-
pendent fast CT scan acquired using the SPECT-CT system. The two CT images
were then fused using rigid matching, and the transformation coordinates were
cross-applied to the PET and SPECT images to obtain an accurate fusion of the
images [10]. Perfusion images acquired using the PET scanner with 13N-ammonia

MET MIBI Fuse

reorient

SA HLA VLA
MET

MET
MIBI

MIBI
Fuse

Fuse

0 100 0 1.5
% Max MBq/cc

Fig. 8.3 PET-SPECT image fusion. Fusion of PET image acquired at 30 min after injection of
11
C-methionine (MET) with sequentially acquired pinhole SPECT perfusion image at 30 min after
injection of 99mTc-sestamibi (MIBI) in a mouse 3d after permanent occlusion of the left coronary
artery. Pinhole SPECT field of view is restricted compared to whole body PET (left in coronal
(upper) and sagittal view (lower)). Liver activity (blue arrow) is used to fine-tune the co-registra-
tion of the images from separate camera systems. The fused images are then reoriented to the
cardiac axis and analyzed in matched and fused short axis (SA), horizontal long axis (HLA), and
vertical long axis (VLA) slices (right). 11C-methionine accumulates in activated macrophages
localized to the perfusion defect. Note the typical circular artifact around the SPECT field of view,
resulting from scatter generated by pinhole collimation and the limited field of view, which does
not interfere with accurate image co-registration. Images generated and co-registered using open-
source AMIDE software. Modified from Thackeray et al. Theranostics. 2016;6:1768–79
172 J. T. Thackeray

may also provide accurate definition of the myocardial contours for co-localization
of the inflammation signal [9] but requires an on-site cyclotron due to the short half-
life of nitrogen-13. In the absence of perfusion images, co-registered CT can also be
utilized to assign the suppressed myocardial 18F-FDG signal to the cardiac region
[39], though the assignment of regions of interest using this method is less precise.

8.6.2 Polar Map Analysis

Multi-isotope or multi-tracer fused images lend themselves to the analysis of

regional left ventricular tracer distribution, particularly when the images can be
acquired simultaneously or sequentially in the same imaging session. Complementary
information on myocardial perfusion, metabolism, and other molecular imaging tar-
gets enables exquisite spatial analysis of tracer distribution in the left ventricle.
Dedicated polar map analysis software has been developed by several working
groups and companies, often providing for interactive image co-registration and
side-by-side analysis (Fig. 8.4). PET-PET, SPECT-SPECT, and PET-SPECT polar
map fusion have wide-ranging applications.
Perfusion-Metabolism. Concurrent assessment of myocardial perfusion with
regional changes in glucose or fatty acid metabolism allows for assessment of heart
failure progression and monitoring of metabolism-targeted therapies. In ischemic
heart failure, myocardial scar or area at risk may be defined by either PET or cardiac
MR, which can allow further differentiation of the substrate. Elevated metabolism
within the perfusion defect indicates repetitive stunning and/or hibernating myocar-
dium, which bears clinical significance for revascularization [40]. In small animals,
definition of altered metabolism is complicated by continuous anesthesia which
affects glucose uptake by cardiomyocytes [41]. Nevertheless, combined analysis of
perfusion and metabolism is feasible and provides complementary data in assessing
cardiac health.
Perfusion-Innervation. Clinical and preclinical studies have emphasized the
importance of myocardial sympathetic innervation in the progression of heart fail-
ure, sudden cardiac death, and particularly lethal ventricular arrhythmias [42, 43].
Combination of a perfusion agent (13N-ammonia) with a sympathetic nerve-targeted
tracer such as 11C-meta-hydroxyephedrine or 11C-epinephrine allows the definition
of not only the perfusion defect, but also denervated or dysinnervated myocardium
[44]. Preclinical imaging studies in pigs have confirmed the larger area of sympa-
thetic innervation defect compared to the perfusion defect, employing a sequential
imaging protocol of 13N-ammonia followed by 11C-meta-hydroxyephedrine or
11
C-epinephrine [45]. Polar map analysis facilitates the comparison between the
imaging agents and requires accurate fusion of the sequentially acquired PET
images, facilitated by co-registered CT.
Perfusion-Inflammation. Localized inflammation within the infarct territory
may be accurately defined by combined perfusion and targeted inflammatory cell
imaging. Lee et al. characterized leukocyte infiltration to the infarct territory fol-
lowing myocardial ischemia in mouse models, using 18F-FDG imaging under
8 Preclinical Multimodality Imaging and Image Fusion in Cardiovascular Disease 173

68Ga-
18F-FDG Fuse
Pentixafor
68Ga-
18F-FDG Fuse
Pentixafor

Reorientation to cardiac axis

Anterior Apex Anterior Anterior Apex Anterior

Inferior Base Inferior Inferior Base Inferior

Ant Ant Ant

Sep Lat Sep Lat Sep Lat
Inf Inf Inf
18F-FDG 68Ga-Pentixafor CXCR4 / FDG
0 3 0 1.5 0 6
MBq /cc x 10 5 Bq /cc Arb. Units

Fig. 8.4 Polar map analysis. Sequential acquisition of PET images using first 68Ga-pentixafor
targeted to CXCR4 and second 18F-FDG under isoflurane to show viable myocardium. Whole body
images (top) are reoriented to the cardiac axis (blue lines), and ray detection (center) through the
ventricular long axes is used to generate polar maps for analysis (bottom). The infarct region
defined by 18F-FDG co-localizes to increased accumulation of 68Ga-pentixafor showing CXCR4
upregulation in the inflammatory region. Correction of the 68Ga-pentixafor image by the infarct
confirms localized increase of CXCR4. Image co-registration and polar map generation performed
using Munich Heart software. Modified from Thackeray et al. JACC Cardiovasc Imaging
2015;8:1417–26

ketamine-xylazine to suppress cardiomyocyte uptake in combination with mag-

netic resonance imaging. A fiducial vest was combined with a specially designed
bed to allow the placement of the animal in identical positions between two scan-
ners. The bed design facilitated no-touch transfer between scanner systems. The
fiducial vest was filled with 15% iodine in water, which permitted visualization
with CT and MRI, with the PET-CT fusion carried out as part of standard protocol.
Fusion accuracy was tested using a18F-filled phantom with cross-correlation func-
tion, showing a peak correlation coefficient of 0.91 over five repetitions. The fidu-
cial vest allowed fine adjustment of the co-registration due to subtle changes in the
animal position during anesthesia and transfer. Such methods were necessitated by
174 J. T. Thackeray

the lack of clear myocardial contours under effective ketamine-xylazine suppres-

sion and allowed the localization of the signal to the infarct territory. Validation of
the signal source was performed using flow cytometry, identifying CD11b-positive
inflammatory cells as the primary contributors to the 18F-FDG signal in the infarct
region [16]. Similarly, sequential imaging of macrophage-expressed mitochondrial
translocator protein (TSPO) using 18F-flutriciclamide and perfusion using 99mTc-
sestamibi allowed the identification of increased macrophages infiltration in the
hypoperfused region of the heart, verified by immunohistology and autoradiogra-
phy [46]. Such approaches have been directly translated to clinical application
[47, 48].
Image fusion is of additional benefit for analysis of multimodal imaging
probes. A 64Cu-labeled nanoparticle assessed macrophage distribution in rejected
cardiac allografts in mice, wherein heterotrophic abdominal grafts were imaged
using PET and iodine vascular contrast-enhanced CT to localize the transplanted
heart. Images were fused manually using Inveon Research Workplace and ana-
lyzed using OsiriX [49].
Viability and Necrosis. Definition of the infarct territory can be enhanced using
both viable myocardium- and scar-avid tracers. Wang et al. acquired SPECT and CT
images using a necrosis-avid agent 131I-rhein in rats after coronary artery occlusion.
Images were acquired sequentially, with fusion performed using Syntegra software.
Contrast-enhanced CT images provided the localization of the myocardium to focus
the diffuse 131I-rhein signal in the necrotic muscle [50]. A comparable study used
labeled platelets and CT to measure the gradual accumulation of 64Cu-labeled acti-
vated integrin GPIIb/IIIa sarcophagine cage complex, identifying focal platelet
accumulation in the ischemic territory in a mouse model of myocardial infarction.
The subsequent CT acquisition allowed definition of the focal PET signal to the
ischemic territory [51].

8.6.3 Ventricular Remodeling

In the remodeling heart, combination of targeted molecular imaging agents with

anatomic imaging can assist in ROI placement and analysis. Regional assessment
may be further enhanced with perfusion analysis or angiography. Jung et al.
employed a multimodality approach including 99mTc-RP805 for SPECT imaging of
matrix metalloproteinase in combination with CT angiography using iohexol
(Jung et al. 2015) to define the aortic valve and assess valvular calcification. Images
were co-registered automatically by the camera proprietary software (Gamma
Medica X-SPECT), and a ROI was defined at the level of the aortic valve identified
on the CT image. Voxels exceeding 130 Hounsfield units were defined as calcifica-
tion [52]. Matrix metalloproteinase activation was proportional to aortic valve cal-
cification but was detectable at an earlier timepoint, suggesting a capacity to predict
outcome. Such approaches underscore the complementary nature of multimodality
imaging and the importance of image fusion to understand the interconnection of
diverse pathophysiologic processes.
8 Preclinical Multimodality Imaging and Image Fusion in Cardiovascular Disease 175

8.6.4 Surgical Guidance

Image fusion is not limited to diagnostic techniques and has shown some potential
for the direct guidance of surgical procedures and interventions. Duckett et al.
applied a segmentation algorithm to overlay coronary sinus anatomy with real-time
acquired fluoroscopy for the accurate placement of ventricular leads in cardiac
resynchronization therapy. CT images were segmented using a 3D anatomic model
using fully automated Philips EP Planner software to give models of cardiac cham-
bers and great vessels. In a separate group of patients, segmentation was performed
semiautomatically using ITK-SNAP software, with the myocardial scar mapped
based on late gadolinium enhancement images. Overlay of the anatomic model on
the live x-ray fluoroscopy data was performed by an experienced operator, with the
heart isocentered in the x-ray field of view, then using implant electrodes for guid-
ance of the segmented anatomic model. For cardiac MR, the right atrium was local-
ized by a quadri-polar electrode catheter looped in the right atrium, allowing
localization of the segmented anatomic model. The co-registration enabled accurate
placement of the CRT leads in the left ventricle [53].
Fusion of PET data with electrophysiology maps provided insights into the con-
tribution of cardiac sympathetic innervation to the substrate of ventricular arrhyth-
mia. In a study of pigs after myocardial infarction, Sasano et al. fused a
11
C-epinephrine polar map denoting regional defects in sympathetic neurons with
an electrophysiology study showing the sites of activation of ventricular tachycar-
dia. The sites of earliest activation of arrhythmia were co-localized to 11C-epinephrine
uptake defects [54], suggesting contribution of sympathetic dysinnervation to the
tachycardia substrate. This concept has been translated clinically in a subsequent
study in which a 123I-MIBG SPECT scan was performed prior to ventricular abla-
tion, with the electrophysiology map superimposed onto the 3D SPECT innervation
map post hoc. The ablation sites were consistently located in denervated regions of
myocardium as defined by MIBG SPECT [55]. Precise overlay of nuclear and elec-
trophysiology data has translational potential for clinical and surgical refinement,
though further characterization is warranted.

8.6.5 Vascular Imaging

Critically, fusion imaging can overcome limitations of defining vasculature in stand-

alone PET or SPECT images. While such studies require the administration of con-
trast agent, definition of the regions of interest using the anatomic images can limit
operator bias and partial volume effects on vascular image interpretation in athero-
sclerosis or aortic aneurysm. As a proof of concept, Fricke et al. evaluated manual
registration of PET images with CT angiography from iomeprol contrast CT images.
Polar maps were generated by MATLAB-based software using a PET uptake image
for co-registration to the CT study and displayed in slices along the transverse, sag-
ittal, and coronal axes. Contour lines and alpha blending allowed for dataset fusion
with realignment by rotation and translation of the datasets. Manual co-registrations
176 J. T. Thackeray

were compared among five operators, with an average Euclidean variation of 1.9–
2.1 mm. The co-registration allows for clear visualization of culprit coronary artery
with perfusion abnormality on stress or rest PET images [56].
More recently, aortic atherosclerotic plaques were targeted using 111In-tilmanocept
directed at macrophage-expressed mannose receptor in a mouse model of athero-
sclerosis. Fusion of the CT image with the 111In-tilmanocept SPECT image enabled
the accurate localization of mannose receptor within atherosclerotic plaque [57].
Other studies have applied a similar localization concept [58, 59], moving toward
more standardized anatomic and physiologic assessment of plaque stability.
Clinical studies suggest the capability to translate these approaches to coronary
artery disease. The multicenter EVINCI study employed hybrid imaging of CT
angiography with perfusion scintigraphy to determine added value in multimodal
imaging in 252 patients with stable angina and intermediate pretest likelihood of
coronary artery disease. The authors emphasized the value of hybrid imaging to
more accurately co-localize myocardial perfusion defects with subtending coronary
arteries accounting for variability in individual coronary anatomy, with a direct
impact on clinical decision-making in 20% of patients [60].

8.6.6 Cell Tracking

Molecular imaging techniques can be valuable in tracing the distribution and reten-
tion of transplanted cells, but single modality approaches are frequently compli-
cated by the lack of an anatomic template to localize the cell-based signal.
Terrovitis et al. employed either SPECT-CT or PET-CT to localize transplanted
NIS reporter gene-expressing cardiac-derived cardiospheres in the ischemic terri-
tory of rats after myocardial infarction. 99mTc-pertechnetate and 201Tl were co-
administered for concurrent imaging using SPECT, providing perfectly fused
images of transplanted cells and perfusion, respectively. The combined SPECT
images were further co-registered to a fast CT image to better localize the activity
[61]. This approach was repeated using whole body PET imaging, using 124I to
image NIS-positive cells, followed by a short acquisition of a higher administered
dose of 13N-ammonia for perfusion. The image fusion was matched due to the iden-
tical position of the animal and further merged to a subsequently acquired CT image
for anatomic localization [61]. The combination of reporter gene imaging with a
perfusion or metabolic agent improves the precise localization of the transplanted
cell signal.
Using a similar approach for clinical application, SPECT-MR assessed CD34+
transplanted cells in the peri-infarct zone in revascularized patients after myocardial
infarction. Transplanted cells were labeled with 99mTc-exametazime to determine
the precise points of administration with subsequent perfusion imaging with 99mTc-
sestamibi to define the infarct territory. Images were fused manually using Siemens
proprietary software with the border zone automatically defined by interactive
thresholding of the perfusion images. Cardiac MR with late gadolinium enhance-
ment (LGE) confirmed the infarct localization, and images were co-registered to the
8 Preclinical Multimodality Imaging and Image Fusion in Cardiovascular Disease 177

99m
Tc-exametazime SPECT to appropriately localize the signal. The study identified
some variability in the definition of the infarct territory and border zone between
SPECT perfusion and LGE MRI, but the fusion of images provided appropriate
anatomic guideposts for the location of transplanted cells [62].
Multi-isotope approaches can improve the localization of the cell-based imaging
signal. Wang et al. performed dual isotope reporter gene imaging using
111
In-octreotide to target somatostatin receptor type 2 (SSTR2) in macrophages and
99m
Tc-pertechnetate to target sodium iodide symporter transduced in HT29 tumor
cells. The different isotope energies allowed for simultaneous acquisition of SPECT
images with perfect co-registration. Anatomic localization was validated using co-
registered CT, with tracer accumulation comparable between NIS- and SSTR2--
transduced cells [63]. The ability to precisely localize transplanted or homing cells
can be invaluable to the development of therapeutic strategies to enhance engraft-
ment, maximize recruitment, and optimize tissue repair.

8.7 Summary and Future Perspective

Considering the expansion of hybrid imaging technologies, it is important to take

advantage of the capability to fuse images acquired using multiple modalities with
complementary information. Indeed, clinical imaging frequently employs CT angi-
ography in combination with myocardial perfusion imaging, facilitating improved
visualization of culprit coronary arteries prior to revascularization [60, 64–67].
Moreover, localization of many nuclear imaging signals in the heart and vasculature
lacks sufficient background signal to clearly define the signal localization without a
fused guidance image, which may be acquired using CT, MR, or perfusion/viability
nuclear imaging. A wealth of software packages has improved the simplicity and
reproducibility of image fusion, such that three-dimensional co-registration is
largely automated, requiring only minimal user interaction to optimize the overlay.
Nonetheless, ideal cardiac image fusion in preclinical applications is obtained using
a multi-modality integrated camera system in which the geometry, physiologic sta-
tus, and positioning of the animal are identical between acquisitions. The combina-
tion of anatomic imaging with physiologic imaging can greatly improve the
localization of regions of interest for quantitative image analysis, particularly when
using a molecular targeted tracer that does not provide clear myocardial contours,
or for vascular imaging where the target can be poorly defined without anatomic
reference. It is therefore important to design the imaging protocol to take full advan-
tage of multimodality imaging capabilities.
The precision of automated and manual image fusion can be enhanced by trans-
ferrable animal beds, fiducial markers (ideally in close proximity to the animal),
dedicated transformation algorithms, and contrast agents or complementary radio-
tracers to define myocardial contours. Indeed, fusion of multiple isotope and diverse
physiologic nuclear images bears potential to localize molecular signals more pre-
cisely. Numerous third-party softwares including open-source packages will allow
continuous improvement in the precision and dissemination of these techniques.
178 J. T. Thackeray

Beyond mere signal localization, image fusion in preclinical cardiac imaging

enables the concurrent evaluation of tissue substrate, metabolic function, innerva-
tion, inflammation, necrosis, and remodeling processes in myocardium and vascu-
lature, aided by the anatomic positioning afforded by CT and/or MR imaging.
Multimodality imaging probes, which combine multiple functionalities within a
single targeted structure, will contribute to the expansion of image fusion tech-
niques. In practice, cardiovascular imaging in the clinical and preclinical arena has
not just approached the multimodality crossroad, but has effectively left it behind,
progressing forward along a unified, fused path toward the future.

References
1. Jaffer FA, Libby P, Weissleder R. Molecular imaging of cardiovascular disease. Circulation.
2007;116(9):1052–61.
2. Feher A, Sinusas AJ. Quantitative assessment of coronary microvascular function: dynamic sin-
gle-photon emission computed tomography, positron emission tomography, ultrasound, com-
puted tomography, and magnetic resonance imaging. Circ Cardiovasc Imaging. 2017;10(8).
3. Tragardh E, Hesse B, Knuuti J, Flotats A, Kaufmann PA, Kitsiou A, et al. Reporting nuclear
cardiology: a joint position paper by the European Association of Nuclear Medicine (EANM)
and the European Association of Cardiovascular Imaging (EACVI). Eur Heart J Cardiovasc
Imaging. 2015;16(3):272–9.
4. Loening AM, Gambhir SS. AMIDE: a free software tool for multimodality medical image
analysis. Mol Imaging. 2003;2(3):131–7.
5. Das NM, Hatsell S, Nannuru K, Huang L, Wen X, Wang L, et al. In vivo quantitative micro-
computed tomographic analysis of vasculature and organs in a normal and diseased mouse
model. PLoS One. 2016;11(2):e0150085.
6. Yan D, Zhang Z, Luo Q, Yang X. A novel mouse segmentation method based on dynamic
contrast enhanced micro-CT images. PLoS One. 2017;12(1):e0169424.
7. Rubeaux M, Joshi NV, Dweck MR, Fletcher A, Motwani M, Thomson LE, et al. Motion
correction of 18F-NaF PET for imaging coronary atherosclerotic plaques. J Nucl Med.
2016;57(1):54–9.
8. Karlsen OT, Verhagen R, Bovee WM. Parameter estimation from Rician-distributed data sets
using a maximum likelihood estimator: application to T1 and perfusion measurements. Magn
Reson Med. 1999;41(3):614–23.
9. Thackeray JT, Bankstahl JP, Wang Y, Korf-Klingebiel M, Walte A, Wittneben A, et al.
Targeting post-infarct inflammation by PET imaging: comparison of (68)Ga-citrate and
(68)Ga-DOTATATE with (18)F-FDG in a mouse model. Eur J Nucl Med Mol Imaging.
2015;42(2):317–27.
10. Thackeray JT, Bankstahl JP, Wang Y, Wollert KC, Bengel FM. Targeting amino acid metabo-
lism for molecular imaging of inflammation early after myocardial infarction. Theranostics.
2016;6(11):1768–79.
11. Lamoureux M, Thorn S, Dumouchel T, Renaud JM, Klein R, Mason S, et al. Uniformity and
repeatability of normal resting myocardial blood flow in rats using [13N]-ammonia and small
animal PET. Nucl Med Commun. 2012;33(9):917–25.
12. Thackeray JT, deKemp RA, Beanlands RS, DaSilva JN. Early diabetes treatment does not
prevent sympathetic dysinnervation in the streptozotocin diabetic rat heart. J Nucl Cardiol.
2014;21(4):829–41.
13. Shekhar R, Walimbe V, Raja S, Zagrodsky V, Kanvinde M, Wu G, et al. Automated 3-dimen-
sional elastic registration of whole-body PET and CT from separate or combined scanners. J
Nucl Med. 2005;46(9):1488–96.
8 Preclinical Multimodality Imaging and Image Fusion in Cardiovascular Disease 179

14. Suh JW, Kwon OK, Scheinost D, Sinusas AJ, Cline GW, Papademetris X. CT-PET weighted
image fusion for separately scanned whole body rat. Med Phys. 2012;39(1):533–42.
15. Khmelinskii A, Groen HC, Baiker M, de Jong M, Lelieveldt BP. Segmentation and visual
analysis of whole-body mouse skeleton microSPECT. PLoS One. 2012;7(11):e48976.
16. Lee WW, Marinelli B, van der Laan AM, Sena BF, Gorbatov R, Leuschner F, et al. PET/MRI
of inflammation in myocardial infarction. J Am Coll Cardiol. 2012;59(2):153–63.
17. Dogdas B, Stout D, Chatziioannou AF, Leahy RM. Digimouse: a 3D whole body mouse atlas
from CT and cryosection data. Phys Med Biol. 2007;52(3):577–87.
18. Thorn SL, deKemp RA, Dumouchel T, Klein R, Renaud JM, Wells RG, et al. Repeatable
noninvasive measurement of mouse myocardial glucose uptake with 18F-FDG: evaluation of
tracer kinetics in a type 1 diabetes model. J Nucl Med. 2013;54(9):1637–44.
19. Matsunari I, Miyazaki Y, Kobayashi M, Nishi K, Mizutani A, Kawai K, et al. Performance eval-
uation of the eXplore speCZT preclinical imaging system. Ann Nucl Med. 2014;28(5):484–97.
20. O'Neill AS, Terry SY, Brown K, Meader L, Wong AM, Cooper JD, et al. Non-invasive molecu-
lar imaging of inflammatory macrophages in allograft rejection. EJNMMI Res. 2015;5(1):69.
21. Pissarek M, Meyer-Kirchrath J, Hohlfeld T, Vollmar S, Oros-Peusquens AM, Flogel U, et al.
Targeting murine heart and brain: visualisation conditions for multi-pinhole SPECT with
(99m)Tc- and (123)I-labelled probes. Eur J Nucl Med Mol Imaging. 2009;36(9):1495–509.
22. Di Carli MF, Asgarzadie F, Schelbert HR, Brunken RC, Rokhsar S, Maddahi J. Relation
of myocardial perfusion at rest and during pharmacologic stress to the PET patterns of
tissue viability in patients with severe left ventricular dysfunction. J Nucl Cardiol.
1998;5(6):558–66.
23. Luisi AJ Jr, Suzuki G, Dekemp R, Haka MS, Toorongian SA, Canty JM Jr, et al. Regional
11C-hydroxyephedrine retention in hibernating myocardium: chronic inhomogeneity of sym-
pathetic innervation in the absence of infarction. J Nucl Med. 2005;46(8):1368–74.
24. Yukinaka M, Nomura M, Ito S, Nakaya Y. Mismatch between myocardial accumulation
of 123I-MIBG and 99mTc-MIBI and late ventricular potentials in patients after myocar-
dial infarction: association with the development of ventricular arrhythmias. Am Heart J.
1998;136(5):859–67.
25. Thackeray JT, Derlin T, Haghikia A, Napp LC, Wang Y, Ross TL, et al. Molecular Imaging
of the Chemokine Receptor CXCR4 After Acute Myocardial Infarction. JACC Cardiovasc
Imaging. 2015;8(12):1417–26.
26. Higuchi T, Anton M, Saraste A, Dumler K, Pelisek J, Nekolla SG, et al. Reporter gene PET for
monitoring survival of transplanted endothelial progenitor cells in the rat heart after pretreat-
ment with VEGF and atorvastatin. J Nucl Med. 2009;50(11):1881–6.
27. Martin EB, Williams A, Richey T, Stuckey A, Heidel RE, Kennel SJ, et al. Comparative evalu-
ation of p5+14 with SAP and peptide p5 by dual-energy SPECT imaging of mice with AA
amyloidosis. Sci Rep. 2016;6:22695.
28. Perin EC, Tian M, Marini FC 3rd, Silva GV, Zheng Y, Baimbridge F, et al. Imaging long-term
fate of intramyocardially implanted mesenchymal stem cells in a porcine myocardial infarc-
tion model. PLoS One. 2011;6(9):e22949.
29. Caobelli F, Wollenweber T, Bavendiek U, Kuhn C, Schutze C, Geworski L, et al. Simultaneous
dual-isotope solid-state detector SPECT for improved tracking of white blood cells in sus-
pected endocarditis. Eur Heart J. 2017;38(6):436–43.
30. Thackeray JT, Korf-Klingebiel M, Wang Y, Kustikova OS, Bankstahl JP, Wollert KC, et al.
Non-invasive tracking of endogenous bone marrow cell recruitment after myocardial infarc-
tion in mice (Abstract). Circulation. 2015;132:A17146.
31. Buscher K, Judenhofer MS, Kuhlmann MT, Hermann S, Wehrl HF, Schafers KP, et al.
Isochronous assessment of cardiac metabolism and function in mice using hybrid PET/MRI. J
Nucl Med. 2010;51(8):1277–84.
32. Weissler B, Gebhardt P, Dueppenbecker PM, Wehner J, Schug D, Lerche CW, et al. A digital
preclinical PET/MRI insert and initial results. IEEE T Med Imaging. 2015;34(11):2258–70.
33. Nieman BJ, Szulc KU, Turnbull DH. Three-dimensional, in vivo MRI with self-gating and
image coregistration in the mouse. Magn Reson Med. 2009;61(5):1148–57.
180 J. T. Thackeray

34. Hatt CR, Stanton D, Parthasarathy V, Jain AK, Raval AN. A method for measuring the
accuracy of multi-modal image fusion system for catheter-based cardiac interventions
using a novel motion enabled targeting phantom. Conf Proc IEEE Eng Med Biol Soc.
2011;2011:6260–4.
35. Jin H, Yang H, Liu H, Zhang Y, Zhang X, Rosenberg AJ, et al. A promising carbon-11-labeled
sphingosine-1-phosphate receptor 1-specific PET tracer for imaging vascular injury. J Nucl
Cardiol. 2017;24(2):558–70.
36. Majmudar MD, Keliher EJ, Heidt T, Leuschner F, Truelove J, Sena BF, et al. Monocyte-directed
RNAi targeting CCR2 improves infarct healing in atherosclerosis-prone mice. Circulation.
2013;127(20):2038–46.
37. Golestani R, Razavian M, Nie L, Zhang J, Jung JJ, Ye Y, et al. Imaging vessel wall biology to
predict outcome in abdominal aortic aneurysm. Circ Cardiovasc Imaging. 2015;8(1)
38. Hong X, Bu L, Wang Y, Xu J, Wu J, Huang Y, et al. Increases in the risk of cognitive impair-
ment and alterations of cerebral beta-amyloid metabolism in mouse model of heart failure.
PLoS One. 2013;8(5):e63829.
39. Cusso L, Vaquero JJ, Bacharach S, Desco M. Comparison of methods to reduce myocar-
dial 18F-FDG uptake in mice: calcium channel blockers versus high-fat diets. PLoS One.
2014;9(9):e107999.
40. D'Egidio G, Nichol G, Williams KA, Guo A, Garrard L, deKemp R, et al. Increasing benefit
from revascularization is associated with increasing amounts of myocardial hibernation: a sub-
study of the PARR-2 trial. JACC Cardiovasc Imaging. 2009;2(9):1060–8.
41. Thackeray JT, Bankstahl JP, Wang Y, Wollert KC, Bengel FM. Clinically relevant strategies for
lowering cardiomyocyte glucose uptake for 18F-FDG imaging of myocardial inflammation in
mice. Eur J Nucl Med Mol Imaging. 2015;42(5):771–80.
42. Thackeray JT, Bengel FM. PET imaging of the autonomic nervous system. Q J Nucl Med Mol
Imaging. 2016;60(4):362–82.
43. Fallavollita JA, Heavey BM, Luisi AJ Jr, Michalek SM, Baldwa S, Mashtare TL Jr, et al.
Regional myocardial sympathetic denervation predicts the risk of sudden cardiac arrest in
ischemic cardiomyopathy. J Am Coll Cardiol. 2014;63(2):141–9.
44. Thackeray JT, Renaud JM, Kordos M, Klein R, Dekemp RA, Beanlands RS, et al. Test-retest
repeatability of quantitative cardiac 11C-meta-hydroxyephedrine measurements in rats by
small animal positron emission tomography. Nucl Med Biol. 2013;40(5):676–81.
45. Lautamaki R, Sasano T, Higuchi T, Nekolla SG, Lardo AC, Holt DP, et al. Multiparametric
molecular imaging provides mechanistic insights into sympathetic innervation impairment in
the viable infarct border zone. J Nucl Med. 2015;56(3):457–63.
46. Thackeray JT, Hupe HC, Wang Y, Bankstahl JP, Berding G, Ross TL, et al. Myocardial
Inflammation Predicts Remodeling and Neuroinflammation After Myocardial Infarction. J Am
Coll Cardiol. 2018;71(3):263–75.
47. Wollenweber T, Roentgen P, Schafer A, Schatka I, Zwadlo C, Brunkhorst T, et al. Characterizing
the inflammatory tissue response to acute myocardial infarction by clinical multimodality non-
invasive imaging. Circ Cardiovasc Imaging. 2014;7(5):811–8.
48. Rischpler C, Dirschinger RJ, Nekolla SG, Kossmann H, Nicolosi S, Hanus F, et al. Prospective
evaluation of 18F-fluorodeoxyglucose uptake in postischemic myocardium by simultaneous
positron emission tomography/magnetic resonance imaging as a prognostic marker of func-
tional outcome. Circ Cardiovasc Imaging. 2016;9(4):e004316.
49. Ueno T, Dutta P, Keliher E, Leuschner F, Majmudar M, Marinelli B, et al. Nanoparticle
PET-CT detects rejection and immunomodulation in cardiac allografts. Circ Cardiovasc
Imaging. 2013;6(4):568–73.
50. Wang Q, Yang S, Jiang C, Li J, Wang C, Chen L, et al. Discovery of radioiodinated monomeric
anthraquinones as a novel class of necrosis avid agents for early imaging of necrotic myocar-
dium. Sci Rep. 2016;6:21341.
51. Ziegler M, Alt K, Paterson BM, Kanellakis P, Bobik A, Donnelly PS, et al. Highly sensitive
detection of minimal cardiac ischemia using positron emission tomography imaging of acti-
vated platelets. Sci Rep. 2016;6:38161.
8 Preclinical Multimodality Imaging and Image Fusion in Cardiovascular Disease 181

52. Jung JJ, Razavian M, Challa AA, Nie L, Golestani R, Zhang J, et al. Multimodality and molec-
ular imaging of matrix metalloproteinase activation in calcific aortic valve disease. J Nucl
Med. 2015;56(6):933–8.
53. Duckett SG, Ginks MR, Knowles BR, Ma Y, Shetty A, Bostock J, et al. Advanced image fusion
to overlay coronary sinus anatomy with real-time fluoroscopy to facilitate left ventricular lead
implantation in CRT. Pacing Clin Electrophysiol. 2011;34(2):226–34.
54. Sasano T, Abraham MR, Chang KC, Ashikaga H, Mills KJ, Holt DP, et al. Abnormal sym-
pathetic innervation of viable myocardium and the substrate of ventricular tachycardia after
myocardial infarction. J Am Coll Cardiol. 2008;51(23):2266–75.
55. Klein T, Abdulghani M, Smith M, Huang R, Asoglu R, Remo BF, et al. Three-dimensional
123I-meta-iodobenzylguanidine cardiac innervation maps to assess substrate and successful
ablation sites for ventricular tachycardia: feasibility study for a novel paradigm of innervation
imaging. Circ Arrhythm Electrophysiol. 2015;8(3):583–91.
56. Fricke H, Elsner A, Weise R, Bolte M, van den Hoff J, Burchert W, et al. Quantitative myocar-
dial perfusion PET combined with coronary anatomy derived from CT angiography: validation
of a new fusion and visualisation software. Z Med Phys. 2009;19(3):182–8.
57. Varasteh Z, Hyafil F, Anizan N, Diallo D, Aid-Launais R, Mohanta S, et al. Targeting mannose
receptor expression on macrophages in atherosclerotic plaques of apolipoprotein E-knockout
mice using 111In-tilmanocept. EJNMMI Res. 2017;7(1):40.
58. Luehmann HP, Detering L, Fors BP, Pressly ED, Woodard PK, Randolph GJ, et al. PET/CT
imaging of chemokine receptors in inflammatory atherosclerosis using targeted nanoparticles.
J Nucl Med. 2016;57(7):1124–9.
59. Hyafil F, Pelisek J, Laitinen I, Schottelius M, Mohring M, Doring Y, et al. Imaging the cytokine
receptor CXCR4 in atherosclerotic plaques with the radiotracer (68)Ga-pentixafor for PET. J
Nucl Med. 2017;58(3):499–506.
60. Liga R, Vontobel J, Rovai D, Marinelli M, Caselli C, Pietila M, et al. Multicentre multi-device
hybrid imaging study of coronary artery disease: results from the EValuation of INtegrated
Cardiac Imaging for the Detection and Characterization of Ischaemic Heart Disease (EVINCI)
hybrid imaging population. Eur Heart J Cardiovasc Imaging. 2016;17(9):951–60.
61. Terrovitis J, Kwok KF, Lautamaki R, Engles JM, Barth AS, Kizana E, et al. Ectopic expression
of the sodium-iodide symporter enables imaging of transplanted cardiac stem cells in vivo by
single-photon emission computed tomography or positron emission tomography. J Am Coll
Cardiol. 2008;52(20):1652–60.
62. Musialek P, Tekieli L, Kostkiewicz M, Miszalski-Jamka T, Klimeczek P, Mazur W, et al.
Infarct size determines myocardial uptake of CD34+ cells in the peri-infarct zone: results from
a study of (99m)Tc-extametazime-labeled cell visualization integrated with cardiac magnetic
resonance infarct imaging. Circ Cardiovasc Imaging. 2013;6(2):320–8.
63. Wang J, Arulanandam R, Wassenaar R, Falls T, Petryk J, Paget J, et al. Enhancing expression
of functional human sodium iodide symporter and somatostatin receptor in recombinant onco-
lytic vaccinia virus for in vivo imaging of tumors. J Nucl Med. 2017;58(2):221–7.
64. de Knegt MC, Fuchs A, Weeke P, Mogelvang R, Hassager C, Kofoed KF. Optimisation of
coronary vascular territorial 3D echocardiographic strain imaging using computed tomogra-
phy: a feasibility study using image fusion. Int J Cardiovasc Imaging. 2016;32(12):1715–23.
65. Gaemperli O, Saraste A, Knuuti J. Cardiac hybrid imaging. Eur Heart J Cardiovasc Imaging.
2012;13(1):51–60.
66. Grant EK, Faranesh AZ, Cross RR, Olivieri LJ, Hamann KS, O'Brien KJ, et al. Image fusion
guided device closure of left ventricle to right atrium Shunt. Circulation. 2015;132(14):1366–7.
67. Plank F, Mueller S, Uprimny C, Hangler H, Feuchtner G. Detection of bioprosthetic valve
infection by image fusion of (18)fluorodeoxyglucose-positron emission tomography and com-
puted tomography. Interact Cardiovasc Thorac Surg. 2012;14(3):364–6.
Dual-Energy SPECT Imaging
with Contrast-Enhanced CT: 9
A Case Study

Emily B. Martin, Alan Stuckey, Stephen J. Kennel,

and Jonathan S. Wall

9.1 Introduction to Imaging Modalities

9.1.1 SPECT Imaging

Single-photon emission computed tomography (SPECT) is the most common

molecular imaging modality in the clinic and requires the use of an injectable
tracer radiolabeled with isotopes that emit higher-energy (>140 keV) gamma ray
photons. The most commonly used SPECT radionuclides in the clinic are
iodine-123 (123I; 159 keV gamma) with a half-life of 13 h and technetium-99m
(99mTc; 140 keV gamma) with a half-life of 6.1 h. In preclinical studies, however,
the low-energy emitter iodine-125 (125I; ~30 keV) is a suitable alternative because
it has a much longer half-life (59 days) allowing for extended observations and,
despite the low energy of the emitted photons, is suitable for imaging small sub-
jects such as mice. SPECT imaging requires the detection of emitted photons that
are collimated prior to interaction with the detectors that, in small animal imaging
systems, either rotate around or surround the subject to generate a 3D image [1, 2].
Several versions of these scanners are commercially available. Imaging systems
with more than one detector have increased sensitivity. SPECT imaging offers sev-
eral advantages when compared to positron emission tomography (PET). The most
notable advantages are the ability to use a range of radionuclides with longer phys-
ical half-lives than those routinely used in PET protocols, which facilitate correla-
tive preclinical studies such as high-resolution microautoradiography, and the
ability to detect multiple radiotracers with different photon emission energies (such
as 125I and 99mTc) simultaneously in vivo [3].

E. B. Martin · A. Stuckey · S. J. Kennel · J. S. Wall (*)

Department of Medicine, University of Tennessee Medical Center, Knoxville, TN, USA
e-mail: [email protected]

© Springer Nature Switzerland AG 2019 183

C. Kuntner-Hannes, Y. Haemisch (eds.), Image Fusion in Preclinical
Applications, https://doi.org/10.1007/978-3-030-02973-9_9
184 E. B. Martin et al.

9.1.2 CT Imaging

X-ray computed tomography (CT) is a high-resolution anatomic imaging modality

capable of visualizing structures based on their relative attenuation of externally
applied X-rays [4]. Spatial resolution is dependent upon many factors; however, for
most preclinical imaging systems, 50–150 μm resolution is used for imaging stud-
ies. Three-dimensional images are generated from multiple X-ray projections that
are acquired as the X-ray source and detector rotate around the subject. As the
X-rays travel through diverse types of material (e.g., fat, tissue, bone, air), they
become variably attenuated, which generates a high-resolution attenuation map or
CT image of the subject in which many anatomic structures can be readily visual-
ized. While CT provides much more information than X-ray radiography, it exposes
individuals to more radiation [5], but to further address this concern, faster multi-
detector CT scanners with the automatic exposure control have been implemented
into many clinical and preclinical imaging platforms [5, 6]. Dual modality imaging
has become the norm, and CT is frequently coupled with SPECT and PET in a sin-
gle imaging system.
One limitation of small animal CT imaging is that organ boundaries, especially
within the abdominal cavity of mice, can be difficult to discern because the X-ray
attenuation is similar for those organs. To improve anatomical visualization, elec-
tron dense X-ray contrast agents, which appear white in CT images, can be admin-
istered. Contrast agents used in the clinic are concentrates of highly attenuating
elements commonly administered intravascularly (typically an iodine-based com-
pound, e.g., Iohexol™) or orally (commonly a barium-based compound) in the
clinical setting which can aid in the detection of vascular or intestinal abnormalities,
respectively. In the preclinical setting, when imaging mice or other small animals
with little visceral fat (since fat can provide a natural contrast to discern organs [7]),
reverse contrast techniques can be used, wherein intraperitoneal (i.p.) injection of
contrast agents, such as Iohexol, can assist in abdominal organ delineation [8]. This
technique can dramatically improve interpretation of the CT image and allow pre-
cise identification of abdominal organs that can be “segmented” in the image such
that volumetric analyses can be performed and whole organ volumes, or tumor vol-
umes, can be established (Fig. 9.1). Increased contrast between the soft abdominal
soft tissues may be achieved by acquiring CT data with decreased X-ray voltage;
however, image integrity may be compromised using this approach with little
increase in information. In the clinic, dual-energy (spectral) CT can now be used, in
conjunction with or without IV contrast agents, to generate image data based on the
tissues attenuation characteristics [9–11].

9.1.3 Dual Modality Imaging

In many instances, CT imaging can be informative as a singular modality; however,

as noted above, CT has been used in conjunction with SPECT (the focus of this
9 Dual-Energy SPECT Imaging with Contrast-Enhanced CT: A Case Study 185

a b

Fig. 9.1 Complete segmentation of mouse abdominothoracic organs using contrast-enhanced CT

images. (a) Representative 2D sagittal image of a contrast-enhanced mouse CT. The contrast
enables organ delineation and facilitates segmentation. (b) A 3D rendering which shows ventral
(left) and dorsal (right) views of fully segmented lungs, heart, liver, spleen, kidneys, stomach, and
intestines

chapter) and PET to provide enhanced capabilities including providing data for
scatter correction of the PET and SPECT images as well as, most importantly, pro-
viding an anatomic frame of reference to enhance interpretation of the PET and
SPECT data [12]. As stated by Mariani et al., “the main advantages of SPECT/CT
are represented by better attenuation correction, increased specificity and accurate
depiction of the localization of disease” [13]. Additionally, SPECT/CT imaging
improves visualization of functional and metabolic information that is provided
with nuclear medicine tests which monitor molecular changes that may not be
observed with anatomical tests alone [13–16]. SPECT/CT imaging platforms are
utilized extensively in the preclinical setting for research and drug development
purposes and have proven to be a valuable imaging tool for translational research
most commonly pertaining to oncology and neuroscience [14]. Moreover, as we
demonstrate in one example below, contrast-enhanced CT imaging may be used in
conjunction with SPECT imaging to better estimate appropriate anatomical place-
ment of regions of interest when performing radiotracer biodistribution analysis
from SPECT images (see Fig. 9.2).
186 E. B. Martin et al.

a b

L
St

Sp

Fig. 9.2 Contrast-enhanced CT imaging aids in placement of ROI. (a) Mouse CT image without
contrast limits the ability of delineate organ boundaries. (b) Mouse CT image enhanced with i.p.
contrast agent visualizes organ boundaries and facilitates proper placement of ROI in the heart (H),
liver (L), stomach (St), and spleen (Sp). The kidneys were also assessed but are not visible in this
CT slice

9.1.4 Dual-Energy SPECT/CT Imaging

A major advantage of SPECT imaging over PET imaging is the ability to assess the
comparative efficacy of two radiotracers simultaneously within a single subject by
employing dual-energy imaging. This is achieved by exploiting the energy proper-
ties of different radioisotopes such that tracers are radiolabeled with high- or low-
energy gamma photon-emitting nuclides or isotopes (i.e., 123I = 159 keV paired with
125
I = 20–35 keV [in the preclinical setting]). This capability allows direct compari-
son of two imaging agents under the same physiological conditions within the same
individual; for example, the gold standard agent can be compared to a novel reagent.
It also permits simultaneous visualization of different physical processes, such as
the measurement of blood flow metabolic rates or protein binding, when appropriate
radiotracers are available [17–23]. The ability to detect and acquire multiple mea-
surements within an individual subject is particularly useful in murine models of
complex disease where inherent dynamic biological variability perpetually exists.
Using dual-energy SPECT imaging circumvents the questionable reproducibility of
9 Dual-Energy SPECT Imaging with Contrast-Enhanced CT: A Case Study 187

in vivo conditions that may arise when large independent groups of animals are used
to compare the biodistribution of radiotracers in vivo. In this way, it also reduces the
number of animals that must be used to achieve statistically significant results.
Despite the simultaneous injection of the dual energy-emitting nuclides, the system
we use acquires the high- and low-energy data sequentially. In the original MicroCAT
II SPECT/CT system, simultaneous acquisition of high- and low-energy photons
was possible; however, in this instance, the system used independent detectors for
high- and low-energy emission. Given these approaches to data acquisition, param-
eters (such as detector voltage) can be optimized for each nuclide. In our system, the
major concern is spill-down of high-energy photons into the low-energy window,
but this is only a major concern when using 123I with 125I, and we have demonstrated
that appropriate corrections can be made [24]. The following case study utilizing
preclinical dual-energy SPECT/CT imaging of independent peptide radiotracers in
mice serves as a workable example that can be adapted to numerous investigational
studies.

9.2  ase Study: The Use of Dual-Energy SPECT/CT

C
to Compare the Biodistribution of Two Peptide
Radiotracers in Mice with Systemic Amyloidosis

9.2.1 I ntroduction to the Complex Disease Model: Murine

Systemic AA Amyloidosis

Systemic amyloidosis refers to a class of diseases characterized by the overproduc-

tion of misfolded proteins which ultimately leads to the formation and deposition of
proteinaceous fibrils, known as amyloid, in numerous and variable tissues and
organs. These deposits jeopardize the structural integrity and function of affected
organs, altering physiological and catabolic processes depending on the amount of
amyloid in each organ. The insidious accumulation of amyloid, notably in the heart
and kidney, can ultimately lead to death [25]. Therapeutic advancements have been
made in recent years with various immunotherapies under evaluation in clinical
trial, but the key to enhancing patient survival is early detection of amyloidosis. In
Europe, radiolabeled serum amyloid P component is used for imaging abdominal
amyloidosis [26]; however, despite radiotracers that can detect cardiac amyloidosis
[27–30], there are no approved imaging agents available in the USA for visualizing
whole-body amyloid load [18, 31]. To address this issue, we have been developing
and evaluating, in an iterative fashion, a series of synthetic peptides with different
bioactivities as potential agents for targeting amyloid and that, when radiolabeled,
can be used as imaging agents. We have used dual-energy SPECT/CT imaging to
directly compare the efficacy of our peptides in both healthy (wild type; WT) mice
and those with systemic multi-organ inflammation-associated (AA) amyloidosis as
a way to identify the optimal agent for translation to the clinic. Herein, we describe
techniques to quantitatively compare two amyloid-binding peptides in a single sub-
ject (mouse) with amyloid disease.
188 E. B. Martin et al.

9.2.2 Evaluation of Two Radiolabeled Peptides in Healthy Mice

The two polybasic peptide radiotracers analyzed in this study, known as aqap5 and
AQA
p5 (radiolabeled with 125I and 123I, respectively), have the same amino acid
sequence, but the N-terminal 3 amino acids of peptide aqap5 are d-enantiomers
(denoted with lowercase letters) as opposed to the all l form, peptide AQAp5.
Incorporation of d-amino acids prevents dehalogenation of radioiodinated aqap5
in vivo, while AQAp5 undergoes both renal and hepatic catabolism resulting in deha-
logenation and thus redistribution of the radioiodide [32]. The goal of this study is
to compare, by dual-energy SPECT/CT imaging, the distribution of radiolabeled
peptide (and free radioiodide) in the same mouse. When 125I-aqap5 and 123I-AQAp5
were administered intravenously (in the lateral tail vein) to healthy WT mice, the
biodistribution of the two radiotracers was discrete and readily visualized (Fig. 9.3).

Thyroid

Stomach

Liver

Kidneys

Fig. 9.3 Dual-energy SPECT imaging of 125I-aqap5 and 123I-AQAp5 peptides co-injected into a WT
mouse. Coronal and sagittal views at 2 h postinjection show the distribution of 125I-aqap5 (pseudo-
colored blue-red) is predominantly located in the liver and kidneys and the 123I-AQAp5 radiotracer
(pseudo-colored green-yellow) is confined to the stomach and thyroid
9 Dual-Energy SPECT Imaging with Contrast-Enhanced CT: A Case Study 189

At 2 h pi, SPECT imaging revealed the presence of aqap5 in the liver and kidneys
(blue-red color scale). In contrast, imaging of AQAp5 revealed radioisotope within the
stomach and thyroid (green-yellow color scale)—organs known to scavenge liber-
ated radioiodide when peptide radiotracers undergo catabolism [33]. The two dis-
tinct distributions demonstrate the different catabolic mechanisms of these two,
essentially identical, peptides in vivo and exemplify the power of dual-energy
SPECT imaging to enable their simultaneous evaluation.

9.2.3  valuation of Radiolabeled aqap5 and AQAp5 Peptides in AA

E
Amyloid Mice

The transgenic murine model of AA amyloidosis that we use in this case study con-
stitutively overexpresses the human interleukin-6 (hIL-6), which results in a chronic
inflammatory process that results in the systemic deposition of abdominothoracic
amyloid, most commonly seen in the liver, spleen, kidneys, heart, and intestines
[25]. Since the structurally similar aqap5 and AQAp5 are both amyloid-binding pep-
tides, we anticipated that a direct comparison of the two, as radiotracers, in indi-
vidual AA mice would result in similar distribution patterns. Dual-energy SPECT
with contrast-enhanced CT imaging at 2 h pi confirmed the expected similar distri-
bution in the amyloid-laden mouse with uptake of both radiotracers in the liver,
spleen, and kidneys (Fig. 9.4). However, subtle differences were readily evident—
notably, the lack of significant uptake of 123I-AQAp5 in the intestine (Fig. 9.4c, d) as
compared to 125I-aqap5 (Fig. 9.4a, b). These differences could only be documented in
a dual-energy experiment.
Despite the ability of whole-body SPECT imaging to demonstrate the uptake of
radiotracers in organs known to involve amyloid disease, these images do not dem-
onstrate specific retention of the peptide in amyloid deposits. This is a critical step
in the translational development of a novel radiotracer targeting a pathologic lesion
in an experimental animal model. Therefore, to ensure that the radioactivity within
these organs is associated with amyloid-specific retention of the radiotracer, we
routinely perform microautoradiography on samples of each tissue [34]. This type
of tissue analysis requires days of treatment making assessment of 123I-AQAp5 impos-
sible due to its 13-h half-life. However, a parallel imaging study can easily be per-
formed using 125I-AQAp5 and 125I-aqap5, to assess the former using this technique.
Analysis of tissues from the 125I-aqap5-injected mouse revealed amyloid-specific
binding of the radiotracer as evidenced by the co-localization of black silver grains
(indicative of the presence of radioiodinated peptide in the autoradiograph [ARG])
with green birefringent amyloid in a Congo red (CR)-stained consecutive tissue sec-
tion (Congo red is a definitive histological stain for amyloid, and green birefrin-
gence in stained tissues is pathognomonic for the pathology). Additionally, this
technique readily indicated no off-target retention of 125I-aqap5 in the amyloid-free
regions of the organs and provided evidence of the renal clearance of the peptide, as
expected (Fig. 9.5, WT mouse kidney sample).
Dual-energy SPECT imaging using contrast-enhanced CT in the abdomen
affords precise visualization of the radiotracers in abdominal organs and tissues;
190 E. B. Martin et al.

a c b

Int

Fig. 9.4 Dual-energy SPECT/CT imaging of 125I-aqap5 and 123I-AQAp5 peptides in an individual
mouse with AA amyloidosis. (a) 2D SPECT/CT (with contrast) of 125I-aqap5 (pseudo-colored red,
sagittal plane). (b) 3D SPECT representation of 125I-aqap5. (c) 2D SPECT/CT (with contrast) of
I- p5 (pseudo-colored blue). (d) 3D SPECT representation of 123I-AQAp5 in the same mouse.
123 AQA

When amyloid is present, the two radiotracers provide similar distribution; however, 125I-aqap5 also
revealed uptake in the intestines (Int)

indeed, without the contrast-enhanced CT for anatomical reference (as shown in

Fig. 9.4b, d), it would be challenging to discern organ boundaries, making analysis
of the images extremely difficult (see Fig. 9.2). Intraperitoneal injection of an iodin-
ated contrast agent, such as Iohexol™ (a 50% v:v solution), can be readily used for
this purpose. This technique employs a relatively inexpensive contrast medium that
can be injected in a volume of 500 μL and preserves the integrity of the tail vein that
may be required for injection of radiotracers. With guidance from the contrast-
enhanced CT, we were able to draw accurate regions of interest (ROIs) on organs
which were then mapped onto the SPECT image data allowing precise measure-
ments of tissue-associated radioactivity for both 125I-aqap5 and 123I-AQAp5. This tech-
nique allows quantification of the two radiotracers within the same ROI in an
individual mouse. These measurements (expressed in this case as mean voxel inten-
sities) can then be compared with standard biodistribution measurements obtained
by measuring radioactivity in tissues harvested at necropsy (often expressed as per-
cent injected dose per gram of tissue)—obviously a terminal step in any imaging
protocol. A comparison of the biodistribution measurements obtained from measur-
ing radioactivity in tissue samples with those obtained directly from the analysis of
9 Dual-Energy SPECT Imaging with Contrast-Enhanced CT: A Case Study 191

AA Mouse WT Mouse

CR ARG ARG

Liver

Spleen

Kidney

Fig. 9.5 Congo red staining and microautoradiography of mouse tissues are used to visualize
pathology and radiotracer within the tissues. Left panel shows green Congo red (CR) birefringence
indicating amyloid deposition in the liver, spleen, and kidney. The center panel shows microauto-
radiography (ARG) of 125I-aqap5 (black dots) counterstained with hematoxylin and eosin for tissue
visualization in a consecutive tissue slice. The peptide distribution in the ARG corresponds with
amyloid seen in the CR images (left panel). The right panel shows 125I-aqap5 in WT tissues; no
accumulation is seen in the liver or spleen, but the kidney reveals excretion of the peptide in
Bowman’s capsules and tubules

ROIs on the SPECT images revealed similar relative quantitation of each peptide
radiotracer in the AA mice, due to their comparable amyloid-targeting efficacy
(Figs. 9.6a, b). Similar analysis of the peptides in WT mice also corresponded with
the image data in that 125I-aqap5 was seen predominantly in the kidneys, and free
radioiodide liberated during catabolism of 123I-AQAp5 was detected in the stomach
(Fig. 9.6c, d). The thyroid, which also scavenges free iodide, is too small to dissect
from mice to obtain accurate biodistribution data. Given these results, dual-energy
SPECT/CT imaging can be considered a powerful tool for the quantitative compari-
son of two (or potentially more) independent radiotracers in individual mice, allow-
ing detailed comparative analysis studies to be performed using small numbers of
animals and accounting for the inherent biology variability of complex disease sys-
tems in animals.
192 E. B. Martin et al.

a b
15 175

Mean Voxel Intensity

150
% ID/gram

125
10
100
75
5 50
25
0 0
er en ey ey ch eart er en ey ey ch eart
Liv Sple Kidn Kidn oma H Liv Sple Kidn Kidn oma H
L R St L R St
c d

Mean Voxel Intensity

15 175
150
% ID/gram

10 125
100
75
5 50
25
0 0
er en ey ey ch eart er en ey ey ch eart
Liv Sple Kidn Kidn oma H Liv Sple Kidn Kidn oma H
L R S t L R S t

Fig. 9.6 Tissue biodistribution measurements and image analyses of both 125I-aqap5 and 123I-AQAp5
peptides in a single representative AA and WT mouse. Tissue biodistribution measurements of
I- p5 (light gray) and 123I-AQAp5 (dark gray) in an AA mouse (a) and WT mouse (c). Radiotracer
125 aqa

distribution was similar in AA mice and varied in WT mice. Image analyses of both radiotracers in
an AA (b) and WT (d) mouse

9.3 Materials and Methods

9.3.1 Radiotracer Preparation

Peptides aqap5 and AQAp5 were synthesized commercially (AnaSpec, Fremont,

CA), purified by reverse-phase HPLC and analyzed by mass spectrometry to
ensure their integrity [18]. Peptide aqap5 was radiolabeled with 125I using 20 μg
chloramine T followed by 20 μg sodium metabisulfite to quench the reaction.
Radioiodination of peptide AQAp5 with 123I was achieved using 1 mg/mL Iodogen
(Pierce Chemical, Rockford, IL) oxidation followed with 2 mg/mL ascorbic acid
reduction [33]. In both cases, radioiodine incorporated in the lone tyrosine amino
acid in the sequence. Both radiolabeled peptides were purified from free radioio-
dide by gel filtration using a Sephadex G-25 size exclusion matrix (PD10, GE
Healthcare). Fractions were collected under gravity, and gamma-counting was
used to identify peptide in fractions with the highest radioactivity, which were
then pooled. Radiochemical purity was assessed with SDS-PAGE analyzed by
phosphor imaging (Cyclone Storage Phosphor System, PerkinElmer, Shelton,
CT).
9 Dual-Energy SPECT Imaging with Contrast-Enhanced CT: A Case Study 193

9.3.2 SPECT/CT Data Acquisition

Dual-energy SPECT/CT experiments were performed in individual AA or WT mice

using an Inveon trimodality SPECT/PET/CT platform [35] (Siemens Preclinical
Solutions, Knoxville, TN) with Inveon Acquisition Workplace (v.2) image recon-
struction software. Although image acquisition parameters presented here are spe-
cific for this imaging platform, all preclinical SPECT multimodality imaging
systems are capable of capturing and analyzing dual-energy SPECT/CT data.
Detailed methods have been previously published [18]; however, briefly, AA (n = 2)
and WT (n = 2) mice were each administered both radiotracers (113 μCi 125I-aqap5
and 46 μCi 123I-AQAp5 mixed in a 200 μL volume) simultaneously in the lateral tail
vein, and at 2 h pi, mice received ~0.5 mL of 1:1 dilution of Iohexol™ in PBS i.p.
1 min before being euthanized by an isoflurane inhalation overdose. An approxi-
mate 2:1 ratio for 125I- and 123I-labeled radiotracer, respectively, reduces spillover
effects from high-energy photons emitted by 123I. SPECT data for the low- and high-
energy isotopes were obtained sequentially with 60, 16-second projections in 1.5
revolutions using a 1-mm-diameter aperture and a five-pinhole “whole mouse body”
collimator positioned 30 mm from the center of the field of rotation. Data were
reconstructed using a point spread function model with a 3D maximum a priori
algorithm (16 iterations, 6 subsets, β = 1) onto an 88 × 88 × 312 matrix with 0.5 mm
isotropic voxels with appropriate scatter and attenuation correction applied.
CT data were acquired after the SPECT images using an 80 kVp X-ray voltage
with a 500 μA anode current. Two bed positions were used with 240 ms exposure
and 361 projections over a 360° rotation with binning at 4. Reconstruction was
achieved with a Feldkamp-filtered cone beam algorithm onto a 256 × 256 × 603
matrix with 211.4 μm voxels.

9.3.3 Necropsy and Biodistribution

Organs (liver, spleen, kidneys, stomach, and heart) were harvested from each mouse
at necropsy, and a small volume of tissue was placed into tared vials and weighed.
Radioactivity was measured using an automated Wizard 3 gamma counter (1480
Wallac Gamma Counter, PerkinElmer) using low- and high-energy windows for 125I
and 123I detection, respectively. Crossover radioactivity detected by the gamma
counter (i.e., low-energy photons from 123I that appear in the 125I energy window)
was accounted for by manually applying a 41.5% crossover correction, and the 125I
values were reduced accordingly. Biodistribution of the peptides from this tech-
nique was expressed as percent injected dose per gram of tissue (%ID/g).

9.3.4 Congo Red Staining and Microautoradiography

Histological evaluation of tissue samples by Congo red (CR) and microautoradiog-

raphy was performed using 6-μm-thick tissue sections cut from formalin-fixed,
194 E. B. Martin et al.

paraffin-embedded tissues which were harvested from mice at necropsy. For CR,
tissues were placed on slides and stained with alkaline CR solution for 1 h at RT
followed by Mayer’s hematoxylin counterstain for 2 min. For microautoradiogra-
phy, slides were dipped in Kodak NTB-2 emulsion, stored in the dark and devel-
oped after a 96 h exposure before being counterstained with hematoxylin. Tissues
were examined using a Leica DM500 light microscope (Leica, Buffalo Grove, IL)
with cross-polarizing filters (for CR), and digital images were obtained using a
cooled CCD camera (SPOT RT-Slider; Diagnostic Instruments, Sterling Heights,
MI) [18, 31].

9.3.5 Image Analysis

SPECT and contrast-enhanced CT image data were co-registered using the

Inveon Research Workplace visualization software (Siemens Preclinical). From
the CT image, small, spherical three-dimensional regions of interest were drawn
over the liver, spleen, stomach, kidneys, and heart of amyloid-laden and WT
mice, taking care to avoid regions thought to contain large blood vessels (in the
liver). These volumes were then mapped to each SPECT dataset, and the mean
voxel intensity of each radiotracer within the ROI was recorded. A spillover cor-
rection factor (5.9%) was applied to the 125I image data to account for the pres-
ence of additional low-energy photons arising from the 123I isotope. This
correction factor for dual-energy image analyses was previously determined
[24]. Additionally, the image data were appropriately scaled for differences in
the injected dose.

9.4 Summary

Dual-energy, or indeed multi-energy, SPECT/CT imaging is a valuable technique

that can be used to directly compare two or more suitably labeled radiotracers
in vivo. Using both SPECT and contrast-enhanced CT modalities allows for precise
anatomical guidance and enhanced accuracy of image analysis. The strength of this
technique lies in the ability to perform comparative effectiveness studies of radio-
tracers in individual animals, thereby minimizing subject-to-subject variability
which is an inherent complication in complex biological systems, particularly when
disease models are used. In our exemplary study, we were able to quantitatively
compare two radioiodinated peptides which behave differently in WT mice but have
similar, although distinct, distribution patterns in mice with systemic AA amyloido-
sis. In the preclinical setting, dual-energy SPECT/CT provides a powerful tool in
the development of imaging agents and other reagents designed to target pathologi-
cal or physiological events, allowing for the direct comparison of reagents in vivo
and thereby providing a pathway to identifying the optimal reagents for translation
to clinical studies.
9 Dual-Energy SPECT Imaging with Contrast-Enhanced CT: A Case Study 195

References
1. Levin CS. Primer on molecular imaging technology. Eur J Nucl Med Mol Imaging.
2005;32(Suppl 2):S325–45. https://doi.org/10.1007/s00259-005-1973-y.
2. Ziegler S. PET and SPECT. In: Kiessling F, Pichler B, editors. Small animal imaging. Berlin:
Springer; 2011. p. 231–6.
3. Madsen MT. Recent advances in SPECT imaging. J Nucl Med. 2007;48(4):661–73.
4. Bazanez-Borgert M. Basics of SPECT, PET and PET/CT imaging: JASS; 2006.
5. Lee CH, Goo JM, Ye HJ, Ye SJ, Park CM, Chun EJ, Im JG. Radiation dose modulation tech-
niques in the multidetector CT era: from basics to practice. Radiographics. 2008;28(5):1451–
9. https://doi.org/10.1148/rg.285075075.
6. Keat N. Report 05016 CT scanner automatic exposure control systems. 2005. http://www.
impactscan.org/reports/Report05016. Accessed 2 Feb 2018.
7. Levine CD, Aizenstein O, Lehavi O, Blachar A. Why we miss the diagnosis of appendicitis
on abdominal CT: evaluation of imaging features of appendicitis incorrectly diagnosed on
CT. AJR Am J Roentgenol. 2005;184(3):855–9. https://doi.org/10.2214/ajr.184.3.01840855.
8. Aide N, Kinross K, Beauregard JM, Neels O, Potdevin T, Roselt P, Dorow D, Cullinane C,
Hicks RJ. A dual radiologic contrast agent protocol for 18F-FDG and 18F-FLT PET/CT imag-
ing of mice bearing abdominal tumors. Mol Imaging Biol. 2011;13(3):518–25. https://doi.
org/10.1007/s11307-010-0378-x.
9. Thaiss WM, Sauter AW, Bongers M, Horger M, Nikolaou K. Clinical applications for dual
energy CT versus dynamic contrast enhanced CT in oncology. Eur J Radiol. 2015;84(12):2368–
79. https://doi.org/10.1016/j.ejrad.2015.06.001.
10. Forghani R, De Man B, Gupta R. Dual-energy computed tomography: physical principles,
approaches to scanning, usage, and implementation: Part 2. Neuroimaging Clin N Am.
2017;27(3):385–400. https://doi.org/10.1016/j.nic.2017.03.003.
11. Silva AC, Morse BG, Hara AK, Paden RG, Hongo N, Pavlicek W. Dual-energy (spectral) CT:
applications in abdominal imaging. Radiographics. 2011;31(4):1031–46.; ; discussion 1047–
50. https://doi.org/10.1148/rg.314105159.
12. Histed SN, Lindenberg ML, Mena E, Turkbey B, Choyke PL, Kurdziel KA. Review of func-
tional/anatomical imaging in oncology. Nucl Med Commun. 2012;33(4):349–61. https://doi.
org/10.1097/MNM.0b013e32834ec8a5.
13. Mariani G, Bruselli L, Kuwert T, Kim EE, Flotats A, Israel O, Dondi M, Watanabe N. A review
on the clinical uses of SPECT/CT. Eur J Nucl Med Mol Imaging. 2010;37(10):1959–85.
https://doi.org/10.1007/s00259-010-1390-8.
14. Bernsen MR, Vaissier PE, Van Holen R, Booij J, Beekman FJ, de Jong M. The role of pre-
clinical SPECT in oncological and neurological research in combination with either CT or
MRI. Eur J Nucl Med Mol Imaging. 2014;41(Suppl 1):S36–49.
15. Buck AK, Nekolla S, Ziegler S, Beer A, Krause BJ, Herrmann K, Scheidhauer K, Wester HJ,
Rummeny EJ, Schwaiger M, Drzezga A. Spect/Ct. J Nucl Med. 2008;49(8):1305–19. https://
doi.org/10.2967/jnumed.107.050195.
16. Seo Y, Mari C, Hasegawa BH. Technological development and advances in single-photon
emission computed tomography/computed tomography. Semin Nucl Med. 2008;38(3):177–
98. https://doi.org/10.1053/j.semnuclmed.2008.01.001.
17. Ma KH, Huang WS, Chen CH, Lin SZ, Wey SP, Ting G, Wang SD, Liu HW, Liu JC. Dual
SPECT of dopamine system using [99mTc]TRODAT-1 and [123I]IBZM in normal and 6-OHDA-
lesioned formosan rock monkeys. Nucl Med Biol. 2002;29(5):561–7.
18. Martin EB, Williams A, Richey T, Stuckey A, Heidel RE, Kennel SJ, Wall JS. Comparative
evaluation of p5+14 with SAP and peptide p5 by dual-energy SPECT imaging of mice with
AA amyloidosis. Sci Rep. 2016;6:22695. https://doi.org/10.1038/srep22695.
19. Meikle SR, Kench P, Kassiou M, Banati RB. Small animal SPECT and its place in the
matrix of molecular imaging technologies. Phys Med Biol. 2005;50(22):R45–61. https://doi.
org/10.1088/0031-9155/50/22/R01.
196 E. B. Martin et al.

20. Nakazawa A, Ikeda K, Ito Y, Iwase M, Sato K, Ueda R, Dohi Y. Usefulness of dual 67Ga and
99m
Tc-sestamibi single-photon-emission CT scanning in the diagnosis of cardiac sarcoidosis.
Chest. 2004;126(4):1372–6. https://doi.org/10.1378/chest.126.4.1372.
21. Sanchez-Crespo A, Petersson J, Nyren S, Mure M, Glenny RW, Thorell JO, Jacobsson H,
Lindahl SG, Larsson SA. A novel quantitative dual-isotope method for simultaneous ventila-
tion and perfusion lung SPET. Eur J Nucl Med Mol Imaging. 2002;29(7):863–75. https://doi.
org/10.1007/s00259-002-0803-8.
22. Wall JS, Richey T, Williams A, Stuckey A, Osborne D, Martin E, Kennel SJ. Comparative
analysis of peptide p5 and serum amyloid P component for imaging AA amyloid in mice
using dual-isotope SPECT. Mol Imaging Biol. 2012;14(4):402–7. https://doi.org/10.1007/
s11307-011-0524-0.
23. Weinmann P, Faraggi M, Moretti JL, Hannequin P. Clinical validation of simultaneous dual-
isotope myocardial scintigraphy. Eur J Nucl Med Mol Imaging. 2003;30(1):25–31. https://doi.
org/10.1007/s00259-002-0995-y.
24. Lee S, Gregor J, Kennel SJ, Osborne DR, Wall J. GATE validation of standard dual energy cor-
rections in small animal SPECT-CT. PLoS One. 2015;10(4):e0122780. https://doi.org/10.1371/
journal.pone.0122780.
25. Wechalekar AD, Gillmore JD, Hawkins PN. Systemic amyloidosis. Lancet. 2015; https://doi.
org/10.1016/S0140-6736(15)01274-X.
26. Hazenberg BP, van Rijswijk MH, Piers DA, Lub-de Hooge MN, Vellenga E, Haagsma EB,
Hawkins PN, Jager PL. Diagnostic performance of 123I-labeled serum amyloid P component
scintigraphy in patients with amyloidosis. Am J Med. 2006;119(4):355 e315–24. https://doi.
org/10.1016/j.amjmed.2005.08.043.
27. Dorbala S, Vangala D, Semer J, Strader C, Bruyere JR Jr, Di Carli MF, Moore SC, Falk
RH. Imaging cardiac amyloidosis: a pilot study using (18)F-florbetapir positron emission
tomography. Eur J Nucl Med Mol Imaging. 2014;41(9):1652–62. https://doi.org/10.1007/
s00259-014-2787-6.
28. Law WP, Wang WY, Moore PT, Mollee PN, Ng AC. Cardiac amyloid imaging with
18
F-Florbetaben PET: a pilot study. J Nucl Med. 2016;57(11):1733–9. https://doi.org/10.2967/
jnumed.115.169870.
29. Osborne DR, Acuff SN, Stuckey A, Wall JS. A routine PET/CT protocol with streamlined
calculations for assessing cardiac amyloidosis using (18)F-Florbetapir. Front Cardiovasc Med.
2015;2:23. https://doi.org/10.3389/fcvm.2015.00023.
30. Rossi P, Tessonnier L, Frances Y, Mundler O, Granel B. 99mTc DPD is the preferential bone
tracer for diagnosis of cardiac transthyretin amyloidosis. Clin Nucl Med. 2012;37(8):e209–10.
https://doi.org/10.1097/RLU.0b013e318248512c.
31. Wall JS, Martin EB, Richey T, Stuckey AC, Macy S, Wooliver C, Williams A, Foster JS,
McWilliams-Koeppen P, Uberbacher E, Cheng X, Kennel SJ. Preclinical validation of the hep-
arin-reactive peptide p5+14 as a molecular imaging agent for visceral amyloidosis. Molecules.
2015;20(5):7657–82. https://doi.org/10.3390/molecules20057657.
32. Martin EB, Williams A, Richey T, Wooliver C, Stuckey A, Foster JS, Kennel SJ, Wall
JS. Evaluation of the effect of D-amino acid incorporation into amyloid-reactive peptides. J
Transl Med. 2017;15(1):247. https://doi.org/10.1186/s12967-017-1351-0.
33. Martin EB, Kennel SJ, Richey T, Wooliver C, Osborne D, Williams A, Stuckey A, Wall
JS. Dynamic PET and SPECT imaging with radioiodinated, amyloid-reactive peptide p5 in
mice: a positive role for peptide dehalogenation. Peptides. 2014;60:63–70. https://doi.
org/10.1016/j.peptides.2014.07.024.
34. Wall JS, Paulus MJ, Gleason S, Gregor J, Solomon A, Kennel SJ. Micro-imaging of amyloid in
mice. Methods Enzymol. 2006;412:161–82. https://doi.org/10.1016/S0076-6879(06)12011-X.
35. Magota K, Kubo N, Kuge Y, Nishijima K, Zhao S, Tamaki N. Performance characterization of
the Inveon preclinical small-animal PET/SPECT/CT system for multimodality imaging. Eur J
Nucl Med Mol Imaging. 2011;38(4):742–52. https://doi.org/10.1007/s00259-010-1683-y.
Multimodality Imaging in Small Animal
Radiotherapy 10
Christian Vanhove and Stefaan Vandenberghe

10.1 Introduction

Cancer is a devastating disease with a rapidly increasing incidence, mainly due to

the ageing population. In 2017, it was estimated that the predicted number of cancer
deaths in the European Union was 1,373,500, compared to 1,333,400 in 2012 (+3%)
[1]. Despite improvements in prevention and treatment, cancer remains one of the
most important causes of morbidity and mortality, leading to high economic costs.
Together with surgery and systemic treatment, radiotherapy is a key part of cancer
treatment, both in the curative and in the palliative setting. About half of cancer
patients will receive radiotherapy as part of their anticancer treatment. Multimodality
imaging is the cornerstone of radiation treatment planning and allows to deliver a
high radiation dose to the tumour, while at the same time keeping the dose in the
surrounding tissue sufficiently low to prevent side effects. In Fig. 10.1, a typical
workflow is shown when radiotherapy is involved during treatment. Multimodality
imaging is used at multiple time points: at the stage of disease diagnosis, for the
planning of the treatment, repeatedly during treatment, and after treatment to moni-
tor anticancer treatment response. In case of a multi-fraction treatment, imaging
during the course of treatment enables corrective action in a feedback loop, known
as adaptive radiotherapy. This also allows alteration of the treatment in tumours that
are not responding to radiotherapy.
The past 40-odd years have seen a great progress in imaging technologies. These
images enable to better localize the desired targets for radiation. It was the pioneer-
ing work of Sir Godfrey Hounsfield that resulted in the introduction of computed
tomography (CT) and allowed for the first time to non-invasively visualize a tumour

C. Vanhove (*) · S. Vandenberghe

Institute Biomedical Technology, Medical Imaging and Signal Processing,
University of Ghent, Ghent, Belgium
e-mail: [email protected]

© Springer Nature Switzerland AG 2019 197

C. Kuntner-Hannes, Y. Haemisch (eds.), Image Fusion in Preclinical
Applications, https://doi.org/10.1007/978-3-030-02973-9_10
198 C. Vanhove and S. Vandenberghe

Imaging:
Diagnosis
CT, MRI, PET, ...

Treatment

Treatment Imaging:
Radiotherapy
planning CT, MRI, PET, ...

Imaging:
Follow-up
CT, MRI, PET, ...

Fig. 10.1 Radiotherapy workflow indicating the role of imaging at multiple stages. A feedback
loop based on imaging information allows adaptation of the therapy to optimize the outcome

in the three spatial dimensions [2], opening the era of three-dimensional conformal
radiotherapy. In three-dimensional conformal radiotherapy, patient-specific target
selection can be performed based on image information, together with the identifi-
cation of sensitive tissues. Based on this one can develop treatment plans that con-
form to the target and avoid sensitive tissues as best as possible. At first, the creation
of treatment plans was a process of trial and error and required the optimization of
a number of parameters, such as the number of radiation beams, the beam size, the
beam weights, and the beam arrangement. Based on CT information, these param-
eters were used to accurately model the behaviour of the radiation beams as they
travel through the body [3–5]. Recently, three-dimensional conformal radiotherapy
has been replaced by intensity-modulated radiotherapy, where dynamic multi-leaf
collimators in combination with variable intensities of the radiation beam are used
to allow even greater control over the shape of the dose distribution [3]. Intensity-
modulated radiotherapy allows almost unlimited degrees of freedom to shape the
radiation beams. As a result, inverse planning is required to (automatically) calcu-
late the desired treatment plan. This process requires defining the target dose for
tumour and sensitive tissues, and a mathematical optimization algorithm is then
used to create a treatment plan that best matches the input criteria. This allows the
delivery of very complex treatment plans, only limited by the physics that governs
photons. As radiation treatment is a delicate balance between tumour control and
normal tissue toxicity, further treatment plan optimization can only be reached
when we obtain greater knowledge of the target and a better understanding of nor-
mal tissue complications. However, as mentioned by Hoffmann et al. [6], in current
10 Multimodality Imaging in Small Animal Radiotherapy 199

treatment plans, the compromise between the dose delivered to tumour and normal
tissue is ‘frozen’, based on what is considered as the best trade-off for a specific
patient population. Moreover, in the current clinical radiotherapy practice, the
tumours are generally irradiated with a spatially uniform dose. It is known, how-
ever, that some tumour regions need a higher dose to be destroyed than other regions
[7, 8] and the hypothesis is that a non-uniform dose distribution should be used to
improve treatment outcome. The information of intra-tumour inhomogeneity comes
from imaging, and it is hypothesized that advances in multimodality image-guided
radiotherapy will lead to improved cancer cure rates. Imaging modalities, such as
magnetic resonance imaging (MRI) and positron emission tomography (PET), can
visualize intra-tumoural biological heterogeneity [9] and have the potential to
improve current treatment strategies.

10.2 Small Animal Radiation Research

For many decades, animal radiation studies were mostly performed using fairly
crude experimental setups with radiation fields that did not conform only to the
desired target [10]. Commonly, these experiments were done on devices intended
for human patient use, and the radiation sources employed were often producing
megavolt (MV) X-rays that have several characteristics that are unsuitable for irra-
diating small targets in small animals [11]. A MV photon beam exhibits dose build-
up at the air-tissue interface in the entrance region of the beam. The extent of this
build-up region corresponds roughly to the order of the animal size itself. This
makes it very challenging to deliver a uniform dose to a tumour. Another issue is the
beam penumbra, which for MV photon beams may extend several millimetres
beyond the target, leading to unacceptable dose distributions in small structures.
However, to implement changes to the existing clinical standard of care, research
must be conducted to develop alternative treatment strategies. Therefore, a novel
approach in radiotherapy is the introduction of preclinical precision image-guided
radiation research. Tumour-bearing small animal models (mostly mice and rats) are
used to investigate the efficacy of complex radiation patterns, possibly combined
with other treatment agents (e.g. angiogenesis inhibitors or radiosensitizers) that
would otherwise be unethical to investigate on patients. Small animal radiation
research can also play an important role in assessing radiation response of normal
tissue and in investigating radioprotective agents. Moreover, preclinical radiation
research allows for studies under controlled experimental conditions using large
cohorts and delivering accelerated results due to the shorter lifespans of rodents.
Consequently, there is an emerging consensus that novel combinations of imaging
and therapy regimen should first be investigated in a preclinical research environ-
ment offering precision irradiation and multimodality imaging. The preclinical find-
ings should then be translatable into a clinical trial in a much faster and more
efficient way than in current practice.
As a result, precision image-guided small animal radiation research platforms
were developed [11, 12]. These platforms make use of kilo-voltage (kV) X-ray
200 C. Vanhove and S. Vandenberghe

sources to avoid dose build-up and to obtain extremely sharp penumbras. They typi-
cally integrate:

• A kV X-ray source that is used for imaging and radiation treatment.

• A computer-controlled stage for animal positioning.
• A rotational gantry assembly to allow radiation delivery from various angles.
• A collimating system to shape the radiation beam.

This is similar to modern human radiotherapy practice and enables a wide vari-
ety of preclinical experiments, such as the synergy of radiation with other therapies,
complex radiation schemes, and image-guided sub-target boost studies. In Fig. 10.2,
an example of a small animal radiation research platform is shown (SARRP,
XStrahl, Surrey, UK).
Treatment planning on these small animal radiation research platforms is based
on CT, which is equivalent to human planning systems and currently the gold stan-
dard for radiation planning [4, 5]. For CT imaging an on-board X-ray detector is
used in combination with the same kV X-ray tube that is used during treatment
(Fig. 10.2). CT imaging is preferred as it allows for accurate beam positioning and
provides electron density information necessary for individual radiation dose calcu-
lations. However, the CT systems installed on these research platforms are based on

kV X-ray tube

Collimator

Rotating gantry

Flat-panel detector

Computer-controlled stage

Fig. 10.2 Small animal radiation research platform integrating a kV X-ray tube, a rotating gantry,
a computer-controlled stage, a collimating system to shape the beam, and a flat-panel CT
detector
10 Multimodality Imaging in Small Animal Radiotherapy 201

the cone-beam (CB) geometry, instead of the spiral (slice-based) CT geometry used
in human systems. CB-CT is hampered by low soft tissue contrast and high back-
ground noise as a result of the large amount of scatter due to a high scatter-to-pri-
mary ratio when no anti-scatter grid is used on these systems [13]. Although many
investigators have shown that CB-CT can be extremely useful for guiding focal
irradiation [14–16], it remains challenging to localize soft tissue targets on CB-CT
images. To localize soft tissue targets more efficiently, CB-CT can be combined
with other imaging modalities, where the alternative imaging technology is often
used for target selection, and CB-CT is used for dose calculations and accurate
beam positioning.

10.3  ultimodality Imaging in Small Animal

M
Radiation Research

Currently, a large number of preclinical non-invasive in vivo imaging techniques are

available that have the potential to provide more detailed information compared to
CB-CT, such as optical imaging, MRI, and nuclear imaging techniques [17, 18].
Each of these modalities has intrinsic advantages and limitations when used in com-
bination with small animal precision image-guided radiation research platforms. In
the next three sections, these advantages and limitations will be discussed into more
detail.

10.3.1 CB-CT Combined with Bioluminescence Imaging

Bioluminescence imaging (BLI) enables to directly visualize cancer cells when they
are transfected with luciferase [19]. Related to image guidance, the major advantage
of BLI is that it provides excellent signal-to-background ratios due to the negligible
background signal. Moreover, it is a relatively inexpensive imaging technique, the
short acquisition times lead to high throughput, it allows non-invasive monitoring of
tumour progression, and the compact footprint enables to integrate BLI into a
micro-irradiator. However, BLI suffers from absorption and scattering of visible
light by tissue, which results in a nonlinear relationship between the true signal
strength and the measured optical signal at the animal’s surface, limiting the accu-
racy to localize a target in 3D. BLI is mostly available in planar mode, though bio-
luminescence tomography (BLT) is feasible, and is hampered by a limited spatial
resolution [17].
Tuli et al. [20] demonstrated the feasibility of BLI-guided irradiation in an ortho-
topic mouse model of pancreatic cancer using an offline optical imager. Targeting
accuracy was measured using a glass bulb, with 5 mm diameter and filled with
bioluminescent cells, which were orthotopically implanted into the tail of the pan-
creas. The centroid of this bulb measured on the CB-CT of the micro-irradiator was
used as a reference. Using planar optical images, a targeting accuracy of 3.5 mm
could be achieved, which indicates the deviation from the centroid of the implanted
202 C. Vanhove and S. Vandenberghe

bulb to a vertical line going through the maximum optical signal detected on the
animal’s surface. Accordingly, using only planar optical images, no accurate infor-
mation of the depth position of the target can be provided. Therefore, the same
group investigated the added value of offline BLT to guide irradiation [21]. Although
an overall accuracy of approximately 1 mm could be achieved with BLT, the authors
indicated that this was a best-case scenario because ex vivo mice were used during
the experiment, minimizing the repositioning error when moving the animals from
the optical imager to the CB-CT. They concluded that an integrated optical/CB-CT
instrument is necessary to further reduce the targeting error.
Weersink et al. [22] were the first to integrate BLI into a micro-irradiator and
concluded that a targeting accuracy of 1 mm can be achieved in rigid homogeneous
phantoms, using planar optical images. The uncertainty of the depth position of the
target was reduced by using a pair of parallel-opposed radiation beams.
Recently, a group from Johns Hopkins University [23, 24] introduced an inte-
grated BLT/CB-CT system (Fig. 10.3) in a small animal radiation research plat-
form. Using a novel reconstruction algorithm based on multispectral BLT [25, 26]
and incomplete variables truncated conjugate gradients [27], an overall targeting
accuracy of 1 mm could be obtained using phantoms and ex vivo mice
experiments.
Improvements of these integrated multimodality systems should be further inves-
tigated. For example, the systems described by Zhang et al. [24] (Fig. 10.3) require
manual docking of the BLT system, and the authors mention that efforts are in
progress for an automatic system that will result in better mechanical
reproducibility.

a b 3-mirror system
Filter wheel
3-mirror system
CCD
camera Lens

c 3-mirror system

CCD camera Filter wheel

Light enclosure

Fig. 10.3 Bioluminescence tomography integrated into a small animal radiation research plat-
form. Reprinted from [24] with permission from Elsevier
10 Multimodality Imaging in Small Animal Radiotherapy 203

10.3.2 CB-CT Combined with MRI

Compared to CB-CT and CT, MRI provides vastly superior soft tissue contrast. This
makes it much easier to visualize lesion boundaries that will result in a much better
delineation of the target volume, helping to better irradiate the lesion and avoid sur-
rounding tissue. An additional advantage is that MRI uses nonionizing radio-waves,
unlike CT that is using ionizing radiation. The major disadvantages of MRI are the
relatively long acquisition times, the significant investment in an MRI scanner and
high operational costs. Moreover, integrating an MRI into a radiation platform is far
from trivial, notwithstanding, clinical systems are currently under construction
[28–30].
Because MRI scans alone cannot be used for dose planning, as they do not pro-
vide the required electron density information, combining MRI with CT data is
increasingly used for radiotherapy planning in the clinic [31]. This combined CT/
MRI dataset contains both the information required for targeting (MRI-based vol-
umes) and for dose calculations (CT-based electron density). Obviously, correct
registration between MRI and CT is required to obtain accurate treatment
planning.
Preclinically, only a few studies have been published that are using MRI-based
radiotherapy. Bolcaen et al. [32] successfully applied a combined CB-CT/MRI
dataset for the irradiation of brain tumours in a F98 glioblastoma rat model [33]
using a micro-irradiator. Rigid-body transformations in combination with a multi-
modality bed were used for image registration between MRI and CB-CT. Contrast-
enhanced MR images were acquired to follow up tumour growth after orthotopic
inoculation, to monitor treatment response, and to delineate the target volume dur-
ing radiotherapy planning. Using three noncoplanar arcs, the prescribed dose could
be delivered to 90% of the target volume, while minimizing the dose to normal brain
tissue. The authors concluded that this combined CT/MRI-based workflow is a
major step forward in bridging the gap between preclinical and clinical radiotherapy
planning.
One of the more challenging aspects of CB-CT imaging relates to the radiation
dose received by the animals, where measurements showed typical radiation doses
in the range of 10–50 cGy for a single CB-CT [34]. The dose at which 50% of mice
die within 30 days (LD50/30) is roughly 5 Gy, which means that a single CB-CT
can represent as much as 10% of the LD50/30. This might become a very important
issue when the therapeutic dose has to be delivered in multiple fractions spaced over
time, where each individual irradiation requires a CB-CT for accurate animal posi-
tioning. Therefore, Gutierrez et al. [35] investigated the feasibility of a MRI-only-
based workflow for radiotherapy planning of the rat brain that enables both accurate
target delineation and accurate dose calculations using only MRI-based volumes.
Moreover, the image registration process between CB-CT and MRI would become
redundant using such an MRI-only-based workflow. Multiple MRI sequences were
used to generate synthetic CT images that could be used for dose calculations
(Fig. 10.4), because only one MRI volume was not sufficient to separate all major
tissue types (air, soft tissue, bone) in the rat head. The synthetic CT images were
204 C. Vanhove and S. Vandenberghe

Fig. 10.4 Coronal slices through the rat head using four different MRI sequences: zero echo time
(ZTE), ultra-short echo time (UTE), T1-weighted (T1w), and T2-weighted (T2w). These images
were used to generate a synthetic CT image for dose calculations

sufficiently similar to the segmented CB-CT images that are routinely used for
radiotherapy planning on preclinical radiation research platforms. No significant
differences were observed between CB-CT- and MRI-based dose calculations when
more complex beam configurations (multiple beams) were used in the dose plan.
The authors concluded that further research is required in the thoracic or abdominal
region of small animals, where more tissue classes will be required to allow for
accurate dose calculations compared to the rat head. Moreover, total MRI scan time
might become an issue because of animal anaesthesia and throughput, and the pro-
posed MRI-only-based workflow still requires the on-board CB-CT of the micro-
irradiation for accurate animal positioning. For the latter, a solution should be found
to ensure a common coordinate system between MR image space and micro-irradi-
ator space, which is a non-trivial issue without the on-board CB-CT information. A
possible solution is the use of digitally reconstructed radiographs (DRRs), extracted
from the acquired MRI volumes, which may provide sufficient information for the
purpose of image guidance.
10 Multimodality Imaging in Small Animal Radiotherapy 205

10.3.3 CB-CT Combined with Nuclear Imaging

An advantage of nuclear imaging techniques, such as positron emission tomography

(PET), is that these can be used to target metabolically highly active regions and
have the potential to visualize intra-tumoural biological heterogeneity [9]. In 2000,
Ling et al. [36] introduced the concept of biological target volume (BTV) by inte-
grating anatomical and functional imaging into the radiotherapy workflow, leading
to what they called multidimensional conformal radiotherapy. This offers the oppor-
tunity to improve dose targeting by delivering a non-uniform dose to a target region
using, for example, PET images. The most widely used PET tracer for tumour stag-
ing and to monitor treatment response is fluor-18 (18F)-labelled fluorodeoxyglucose
(FDG), which visualizes the glucose metabolism [37]. However, PET can also be
used to detect hypoxic tumour regions with tracers such as 18F-fluoromisonidazole
(FMISO) [38]. Because tumour hypoxia is a known radioresistance factor, such
information can be used to give an extra boost of radiation to hypoxic regions of the
tumour. Despite that, the implementation of PET for radiotherapy planning is still
under investigation in the clinic [39–41], and preclinical research might provide
new insights for combining PET with radiotherapy. However, only a limited number
of preclinical studies have been published. In 2007, Schütze et al. [42] investigated
whether tumour heterogeneity in FDG uptake relates to radiation response. They
showed that an increase of radiation dose had a greater effect on long-term recur-
rence-free survival in tumours with higher FDG uptake than in tumours with lower
FDG uptake, supporting the hypothesis that pretreatment FDG-PET may provide
useful information to deliver a non-uniform dose to the tumour. This hypothesis was
evaluated by Trani et al. [43], who assessed the therapeutic efficacy of two different
strategies to boost tumour sub-volumes with high FDG uptake: targeted dose esca-
lation and dose redistribution. For targeted dose escalation, a 40% higher dose was
delivered to a tumour region with high FDG uptake compared to the rest of the
tumour. In the dose redistribution experiments, a 40% or 60% higher dose was
delivered to a tumour region with high FDG, while a lower dose was delivered to the
rest of the tumour. In contrast to what they expected, their results indicated that a
decrease of radiation dose to tumour sub-volumes with low FDG uptake, while
increasing the dose to high FDG uptake sub-volumes in a dose redistribution
approach, might be detrimental for some dose levels. They concluded that their data
are consistent with a hypothesis that tumour response depends on a minimum intra-
tumoural dose. Thus further preclinical research is required before translating radio-
therapy based on metabolic active regions into clinical trials (Fig. 10.5).
Evidently, PET-based radiotherapy planning requires correct registration with
the planning CB-CT to obtain accurate treatment planning. This process can be
simplified by using a multimodality bed to move the animal from the PET device to
the micro-irradiator. In addition, the use of a hybrid PET-CT scanner can make the
co-registration less difficult by using the CB-CT and the CT acquired in combina-
tion with the PET for image fusion.
206 C. Vanhove and S. Vandenberghe

a b

Fig. 10.5 Example of PET-based treatment planning. (a) PET-CT image of the rat brain in a rat
model of glioblastoma where a heterogeneous uptake of fluor-18-labelled fluorodeoxyglucose
(FDG) can be observed in the tumour. (b) Treatment planning based on multiple beams to deliver
a heterogeneous dose to the tumour. In this example, a larger dose is delivered to the part of the
tumour with larger FDG uptake

Finally, compact preclinical PET scanners offering submillimetre spatial resolu-

tion are under development [44], and these devices might provide a very elegant
solution to integrate PET into a small animal radiation platform.

10.4 Conclusion

To enable more accurate irradiation in small animal research, precision image-

guided small animal radiation research platforms were developed. Similar to clini-
cal systems, treatment planning on these platforms is based on CT. However,
preclinical CT is hampered by very poor soft tissue contrast and high radiation
doses. Consequently, multimodality imaging is the next logical step in the field of
small animal precision image-guided radiotherapy.

References
1. Malvezzi M, Carioli G, Bertuccio P, Boffetta P, Levi F, La Vecchia C, et al. European can-
cer mortality predictions for the year 2017, with focus on lung cancer. Ann Oncol [Internet].
2017;28(5):1117–23. https://academic.oup.com/annonc/article-lookup/doi/10.1093/annonc/
mdx033.
2. Petrik V, Apok V, Britton JA, Bell BA, Papadopoulos MC. Godfrey Hounsfield and the dawn
of computed tomography. Neurosurgery. 2006;58(4):780–7.
3. Khan FM. The physics of radiation therapy. 4th ed. Philadelphia: Lippincott Williams and
Wilkins; 2010. 531 p.
4. Aird EGA, Conway J. CT simulation for radiotherapy treatment planning. Br J Radiol.
2002;75(900):937–49.
5. Baker GR. Localization: conventional and CT simulation. Br J Radiol. 2006;79(Spec. No.
1):S36–49.
10 Multimodality Imaging in Small Animal Radiotherapy 207

6. Hoffmann AL, Huizenga H, Kaanders JH. Employing the therapeutic operating characteris-
tic (TOC) graph for individualised dose prescription. Radiat Oncol [Internet]. 2013;8(1):55.
http://www.pubmedcentral.nih.gov/articlerender.fcgi?artid=3606307&tool=pmcentrez&rend
ertype=abstract.
7. Bentzen SM, Gregoire V. Molecular imaging-based dose painting: a novel paradigm for radia-
tion therapy prescription. Semin Radiat Oncol [Internet]. 2011;21(2):101–10. https://doi.
org/10.1016/j.semradonc.2010.10.001.
8. Diaz-Cano SJ. Tumor heterogeneity: mechanisms and bases for a reliable application of
molecular marker design. Int J Mol Sci. 2012;13(2):1951–2011.
9. Metz S, Ganter C, Lorenzen S, Van Marwick S, Holzapfel K, Herrmann K, et al. Multiparametric
MR and PET imaging of intratumoral biological heterogeneity in patients with metastatic lung
cancer using voxel-by-voxel analysis. PLoS One. 2015;10(7):1–14.
10. Hamstra DA, Rice DJ, Anthony P, Oyedijo D, Ross BD, Rehemtulla A. Combined radiation
and enzyme/prodrug treatment for head and neck cancer in an orthotopic animal model. Radiat
Res. 1999;152(5):499–507.
11. Verhaegen F, Granton P, Tryggestad E. Small animal radiotherapy research platforms. Phys
Med Biol. 2011;56(12):R55–83.
12. Butterworth KT, Prise KM, Verhaegen F. Small animal image-guided radiotherapy: status,
considerations and potential for translational impact. Br J Radiol. 2015;88(1045):4–6.
13. Siewerdsen JH, Moseley DJ, Bakhtiar B, Richard S, Jaffray DA. The influence of antiscatter
grids on soft-tissue detectability in cone-beam computed tomography with flat-panel detectors.
Med Phys. 2004;31(12):3506–20.
14. Bolcaen J, Descamps B, Deblaere K, Boterberg T, De Vos F, Kalala Okito J-P, et al. F18-
fluoromethylcholine (FCho), F18-fluoroethyltyrosine (FET) and F18-fluorodeoxyglucose
(FDG) for the discrimination between high-grade glioma (HGG) and radiation necrosis (RN):
a μPET study. Soc Nucl Med Annu Meet Abstr [Internet]. 2014;55(Suppl. 1):1379. http://
jnumedmtg.snmjournals.org/cgi/content/meeting_abstract/55/1_MeetingAbstracts/1379.
15. Clarkson R, Lindsay PE, Ansell S, Wilson G, Jelveh S, Hill RP, et al. Characterization of image
quality and image-guidance performance of a preclinical microirradiator. Med Phys [Internet].
2011;38(2):845–56. http://www.pubmedcentral.nih.gov/articlerender.fcgi?artid=3188651&to
ol=pmcentrez&rendertype=abstract
16. Baumann BC, Benci JL, Santoiemma PP, Chandrasekaran S, Hollander AB, Kao GD, et al. An
integrated method for reproducible and accurate image-guided stereotactic cranial irradiation
of brain tumors using the small animal radiation research platform. Transl Oncol [Internet].
2012;5(4):230–7. http://www.pubmedcentral.nih.gov/articlerender.fcgi?artid=3431032&tool
=pmcentrez&rendertype=abstract.
17. James ML, Gambhir SS. A molecular imaging primer: modalities, imaging agents, and appli-
cations. Physiol Rev. 2012;92(2):897–965.
18. Cunha L, Horvath I, Ferreira S, Lemos J, Costa P, Vieira D, et al. Preclinical imaging: an essen-
tial ally in modern biosciences. Mol Diagn Ther. 2014;18(2):153–73.
19. Sato A, Klaunberg B, Tolwani R. In vivo bioluminescence imaging BLI: an overview. Comp
Med. 2004;54(6):631–4.
20. Tuli R, Surmak A, Reyes J, Hacker-Prietz A, Armour M, Leubner A, et al. Development of a
novel preclinical pancreatic cancer research model: bioluminescence image-guided focal irra-
diation and tumor monitoring of orthotopic xenografts. Transl Oncol [Internet]. 2012;5(2):77–
84. https://doi.org/10.1593/tlo.11316%5Cnpapers2://publication/doi/10.1593/tlo.11316
21. Tuli R, Armour M, Surmak A, Reyes J, Iordachita I, Patterson M, et al. Accuracy of off-
line bioluminescence imaging to localize targets in preclinical radiation research. Radiat Res.
2013;179(4):416–21.
22. Weersink RA, Ansell S, Wang A, Wilson G, Shah D, Lindsay PE, et al. Integration of optical
imaging with a small animal irradiator. Med Phys [Internet]. 2014;41(10):102701. http://www.
ncbi.nlm.nih.gov/pubmed/25281980.
23. Yang Y, Wang KK-H, Eslami S, Iordachita II, Patterson MS, Wong JW. Systematic calibra-
tion of an integrated X-ray and optical tomography system for preclinical radiation research.
208 C. Vanhove and S. Vandenberghe

Med Phys [Internet]. 2015;42:1710–20. http://scitation.aip.org/content/aapm/journal/

medphys/42/4/10.1118/1.4914860.
24. Zhang B, Wang KK-H, Yu J, Eslami S, Tran PT, Iordachita I, et al. Bioluminescence
tomography guided radiation therapy for preclinical research. Radiat Oncol Biol [Internet].
2016;94(5):1–27. https://doi.org/10.1016/j.ijrobp.2015.11.039.
25. Dehghani H, Davis SC, Jiang SD, Pogue BW, Paulsen KD, Patterson MS. Spectrally resolved
bioluminescence optical tomography. Opt Lett. 2006;31(3):365–7.
26. Dehghani H, Davis SC, Pogue BW. Spectrally resolved bioluminescence tomography using
the reciprocity approach. Med Phys [Internet]. 2008;35(11):4863–71. http://www.pubmedcen-
tral.nih.gov/articlerender.fcgi?artid=2737244&tool=pmcentrez&rendertype=abstract.
27. He X, Liang J, Wang X, Yu J, Qu X, Wang X, et al. Sparse reconstruction for quantitative
bioluminescence tomography based on the incomplete variables truncated conjugate gradient
method. Opt Express [Internet]. 2010;18(24):24825–41. https://doi.org/10.1364/oe.18.024825
28. Wooten HO, Green O, Yang M, DeWees T, Kashani R, Olsen J, et al. Quality of intensity
modulated radiation therapy treatment plans using a (60)Co magnetic resonance image guid-
ance radiation therapy system. Int J Radiat Oncol Biol Phys [Internet]. 2015;92(4):771–8.
http://www.sciencedirect.com/science/article/pii/S0360301615003004.
29. Wooten HO, Rodriguez V, Green O, Kashani R, Santanam L, Tanderup K, et al. Benchmark
IMRT evaluation of a Co-60 MRI-guided radiation therapy system. Radiother Oncol [Internet].
2015;114(3):402–5. https://doi.org/10.1016/j.radonc.2015.01.015.
30. Lagendijk JJW, Raaymakers BW, Raaijmakers AJE, Overweg J, Brown KJ, Kerkhof EM, et al.
MRI/linac integration. Radiother Oncol. 2008;86(1):25–9.
31. Schmidt MA, Payne GS. Radiotherapy planning using MRI. Phys Med Biol [Internet].
2015;60(22):R323–61. http://stacks.iop.org/0031-9155/60/i=22/a=R323?key=crossref.73d6c
6caed0c84bc13887e1da0f28de1.
32. Bolcaen J, Descamps B, Deblaere K, Boterberg T, Hallaert G, Van den Broecke C, et al. MRI-
guided 3D conformal arc micro-irradiation of a F98 glioblastoma rat model using the Small
Animal Radiation Research Platform (SARRP). J Neuro-Oncol. 2014;120(2):257–66.
33. Bryant MJ, Chuah TL, Luff J, Lavin MF, Walker DG. A novel rat model for glioblastoma
multiforme using a bioluminescent F98 cell line. J Clin Neurosci. 2008;15(5):545–51.
34. Cherry SR. In vivo molecular and genomic imaging: new challenges for imaging physics. Phys
Med Biol. 2004;49(3):R13–48.
35. Gutierrez S, Descamps B, Vanhove C. MRI-only based radiotherapy treatment planning for
the rat brain on a Small Animal Radiation Research Platform (SARRP). PLoS One [Internet].
2015;10(12):e0143821. http://dx.plos.org/10.1371/journal.pone.0143821.
36. Ling CC, Humm J, Larson S, Amols H, Fuks Z, Leibel S, et al. Towards multidimensional
radiotherapy (MD-CRT):biological imaging and biological conformality. Int J Radiat Oncol
Biol Phys. 2000;47(3):551–60.
37. Wahl RL, Jacene H, Kasamon Y, Lodge MA. From RECIST to PERCIST: evolving con-
siderations for PET response criteria in solid tumors. J Nucl Med [Internet]. 2009;50 Suppl
1(5):122S–50S. http://www.pubmedcentral.nih.gov/articlerender.fcgi?artid=2755245&tool=p
mcentrez&rendertype=abstract.
38. Lopci E, Grassi I, Chiti A, Nanni C, Cicoria G, Toschi L, et al. PET radiopharmaceuticals for
imaging of tumor hypoxia: a review of the evidence. Am J Nucl Med Mol Imaging [Internet].
2014;4(4):365–84. http://www.ncbi.nlm.nih.gov/pubmed/24982822.
39. Gregoire V, Chiti A. PET in radiotherapy planning: particularly exquisite test or pending and
experimental tool? Radiother Oncol. 2010;96(3):275–6.
40. Chiti A, Kirienko M, Grégoire V. Clinical use of PET-CT data for radiotherapy planning: what
are we looking for? Radiother Oncol [Internet]. 2010;96(3):277–9. https://doi.org/10.1016/j.
radonc.2010.07.021
41. Delikgoz Soykut E, Ozsahin EM, Yukselen Guney Y, Aytac Arslan S, Derinalp Or O, Altundag
MB, et al. The use of PET/CT in radiotherapy planning: contribution of deformable registra-
tion. Front Oncol [Internet]. 2013;3(April):33. http://www.pubmedcentral.nih.gov/articleren-
der.fcgi?artid=3624079&tool=pmcentrez&rendertype=abstract.
10 Multimodality Imaging in Small Animal Radiotherapy 209

42. Schütze C, Bergmann R, Yaromina A, Hessel F, Kotzerke J, Steinbach J, et al. Effect of increase
of radiation dose on local control relates to pre-treatment FDG uptake in FaDu tumours in nude
mice. Radiother Oncol. 2007;83(3):311–5.
43. Trani D, Yaromina A, Dubois L, Granzier M, Peeters SGJA, Biemans R, et al. Preclinical
assessment of efficacy of radiation dose painting based on intratumoral FDG-PET uptake.
Clin Cancer Res [Internet]. 2015;21:5511–9. http://clincancerres.aacrjournals.org/cgi/
doi/10.1158/1078-0432.CCR-15-0290.
44. España S, Marcinkowski R, Keereman V, Vandenberghe S, Van Holen R. DigiPET: sub-mil-
limeter spatial resolution small-animal PET imaging using thin monolithic scintillators. Phys
Med Biol [Internet]. 2014;59(13):3405. http://stacks.iop.org/0031-9155/59/i=13/a=3405.