Preclinical MRI - Methods and Protocols (PDFDrive)

Methods in
Molecular Biology 1718
María Luisa García-Martín

Pilar López-Larrubia Editors
Preclinical
MRI
Methods and Protocols
METHODS IN MOLECULAR BIOLOGY
Series Editor
John M. Walker
School of Life and Medical Sciences
University of Hertfordshire
Hatfield, Hertfordshire, AL10 9AB, UK
For further volumes:

http://www.springer.com/series/7651
Preclinical MRI
Methods and Protocols
Edited by
María Luisa García-Martín

BIONAND, Andalusian Centre for Nanomedicine and Biotechnology, Junta de Andalucía, Universidad de
Málaga, Málaga, Spain; Networking Research Center on Bioengineering, Biomaterials and Nanomedicine,
CIBER-BBN, Málaga, Spain
Pilar López-Larrubia
Instituto de Investigaciones Biomédicas “Alberto Sols”, CSIC/UAM, Madrid, Spain
Editors
Marı́a Luisa Garcı́a-Martı́n Pilar López-Larrubia
BIONAND, Andalusian Centre Instituto de Investigaciones Biomédicas “Alberto Sols”
for Nanomedicine and Biotechnology CSIC/UAM
Junta de Andalucı́a Madrid, Spain
Universidad de Málaga
Málaga, Spain
Networking Research Center on Bioengineering
Biomaterials and Nanomedicine, CIBER-BBN
Málaga, Spain
ISSN 1064-3745 ISSN 1940-6029 (electronic)

Methods in Molecular Biology
ISBN 978-1-4939-7530-3 ISBN 978-1-4939-7531-0 (eBook)
https://doi.org/10.1007/978-1-4939-7531-0
Library of Congress Control Number: 2017962639
© Springer Science+Business Media, LLC 2018

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.
Printed on acid-free paper
This Humana Press imprint is published by Springer Nature

The registered company is Springer Science+Business Media, LLC
The registered company address is: 233 Spring Street, New York, NY 10013, U.S.A.
Preface
Magnetic resonance imaging (MRI), the most versatile of all in vivo imaging modalities, was
born in 1973, when the Nobel Prize awardees, Paul C. Lauterbur and Peter Mansfield, at
the State University of New York and the University of Nottingham, published their
pioneering works on the use of magnetic field gradients to spatially localize the NMR signal.
Lauterbur obtained images of two water filled tubes using magnetic field gradients and
backprojection. He named this new imaging technique “zeugmatography,” derived from
the Greek word zeugma, meaning “that which is used for joining,” in reference to the joint
action of magnetic field gradients and radiofrequency to generate the image. In the same
year, Mansfield published his work demonstrating how a linear field gradient, along with the
Fourier transform, could be used to localize the NMR signal from different layers within a
solid sample, which is the basis of the slice selection used nowadays. Later, in 1977,
Mansfield and Maudsley obtained the first image of a part of the human body, a finger.
Since then, MRI has experienced a tremendous evolution thanks to the joint effort of
scientists from many different fields. Today, MRI is undoubtedly the leading technique in
diagnostic imaging. It has attracted a great deal of interest because of its unique combination
of qualities. MRI uses non-ionizing radiation, which is harmless to human tissue; offers very
high image quality, providing excellent anatomical detail; and additionally is also capable of
providing functional and metabolic information. On the negative side, it was conventionally
argued that MRI suffered from low sensitivity compared to other imaging modalities.
However, a new generation of contrast agents based on nanotechnology is making it
possible to overcome this limitation and bring MRI into the molecular imaging category.
Consequently, interest in MRI continues to grow and gain new adepts from different
fields who see MRI as a very powerful tool capable of answering many of their scientific
questions. Thus, in addition to the unquestionable growth of MRI use in clinical diagnosis,
its applications in basic and translational research have also increased enormously in recent
years, resulting in the creation of a multitude of preclinical imaging units worldwide.
This book was conceived with the idea of providing an update on a wide variety of
preclinical MRI methods and protocols to help technicians and researchers interested in this
technology to perform studies that have already been implemented by recognized experts in
the field.
The book is organized in seven parts:
Part I covers the basics of MRI physics, relaxation, image contrast, and main acquisition
sequences.
Part II describes updated methodology and protocols for diffusion, perfusion, and
functional imaging.
Part III is dedicated to in vivo spectroscopy, covering both proton and heteronuclear
spectroscopy, as well as spectroscopic imaging.
Part IV is intended to include some less common advanced techniques that we thought
might be of high interest to the readers of this book.
Parts V and VI illustrate some applications of the methods described above.
Part VII includes theoretical chapters aimed at providing relevant information on
anesthesia and contrast agents.
v
vi Preface
Finally, we would like to thank the collaboration of all the excellent experts who have
generously contributed with their chapters to the elaboration of this book.
Málaga, Spain Marı́a Luisa Garcı́a-Martı́n

Madrid, Spain Pilar Lopez-Larrubia
Contents
Preface . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . v
Contributors. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xi
PART I MRI BASICS

1 Introduction to MRI Physics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
Gary V. Martinez
2 Basic Pulse Sequences in Magnetic Resonance Imaging . . . . . . . . . . . . . . . . . . . . . . 21
Daniel Calle and Teresa Navarro
PART II PERFUSION, DIFFUSION AND FUNCTIONAL MRI
3 Dynamic Susceptibility Contrast MRI in Small Animals. . . . . . . . . . . . . . . . . . . . . . 41

Pilar Lopez-Larrubia
4 Preclinical Arterial Spin Labeling Measurement of Cerebral Blood Flow . . . . . . . 59
Eric R. Muir
5 Dynamic Contrast-Enhanced MRI . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 71
Jennifer Moroz and Stefan A. Reinsberg
6 Diffusion-Weighted Magnetic Resonance Imaging . . . . . . . . . . . . . . . . . . . . . . . . . . 89
Irene Guadilla, Daniel Calle, and Pilar Lopez-Larrubia
7 Diffusion Tensor Imaging (DTI) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 103
Silvia Lope-Piedrafita
8 Mapping Functional Connectivity in the Rodent Brain Using
Electric-Stimulation fMRI . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 117
Laura Pérez-Cervera, José Marı́a Caramés, Luis Miguel Fernández-Mollá,
Andrea Moreno, Begoña Fernández, Elena Pérez-Montoyo, David Moratal,
Santiago Canals, and Jesús Pacheco-Torres
9 Functional Diffusion Magnetic Resonance Imaging . . . . . . . . . . . . . . . . . . . . . . . . . 135
Rita Maria Rocha Oliveira, Irene Guadilla, and Pilar Lo pez-Larrubia
PART III IN VIVO SPECTROSCOPY

10 In Vivo 1H Magnetic Resonance Spectroscopy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 151
M. Carmen Muñoz-Hernández and Marı́a Luisa Garcı́a-Martı́n
11 In Vivo Heteronuclear Magnetic Resonance Spectroscopy . . . . . . . . . . . . . . . . . . . 169
Blanca Lizarbe, Antoine Cherix, and Rolf Gruetter
1
12 H Spectroscopic Imaging of the Rodent Brain. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 189
Rui V. Simões, Emma Muñoz-Moreno, Raúl Tudela, and Guadalupe Soria
vii
viii Contents
PART IV SPECIAL MRI TECHNIQUES
13 Susceptibility Weighted MRI in Rodents at 9.4 T . . . . . . . . . . . . . . . . . . . . . . . . . . . 205

Ferdinand Schweser, Marilena Preda, and Robert Zivadinov
14 Biomedical 19F MRI Using Perfluorocarbons . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 235
Tuba G€u den-Silber, Sebastian Temme, Christoph Jacoby, and Ulrich Flögel
15 Rodent Abdominal Adipose Tissue Imaging by MR. . . . . . . . . . . . . . . . . . . . . . . . . 259
Bhanu Prakash KN, Jadegoud Yaligar, Sanjay K. Verma,
Venkatesh Gopalan, and S. Sendhil Velan
16 Cardiac MRI in Small Animals. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 269
Min-Chi Ku, Till Huelnhagen, Thoralf Niendorf,
and Andreas Pohlmann
17 In Utero MRI of Mouse Embryos . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 285
Jiangyang Zhang, Dan Wu, and Daniel H. Turnbull
18 Oxygenation Imaging by Nuclear Magnetic Resonance Methods . . . . . . . . . . . . . 297
Heling Zhou, Nuria Arias-Ramos, Pilar Lopez-Larrubia,
Ralph P. Mason, Sebastián Cerdán, and Jesús Pacheco-Torres
19 Molecular Magnetic Resonance Imaging (mMRI) . . . . . . . . . . . . . . . . . . . . . . . . . . 315
Maxime Gauberti, Antoine P. Fournier, Denis Vivien,
and Sara Martinez de Lizarrondo
PART V MRI AND MRS IN ANIMAL MODELS OF DISEASE
20 Magnetic Resonance Spectroscopy Studies of Mouse Models of Cancer . . . . . . . 331

Menglin Cheng and Kristine Glunde
21 MRI in the Study of Animal Models of Neurodegenerative Diseases . . . . . . . . . . 347
Nyoman D. Kurniawan
22 MRI in the Study of Animal Models of Stroke. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 377
Pedro Ramos-Cabrer and Daniel Padro
PART VI OTHER APPLICATIONS

23 Assessment of Blood Brain Barrier Leakage with
Gadolinium-Enhanced MRI. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 395
Min-Chi Ku, Sonia Waiczies, Thoralf Niendorf,
and Andreas Pohlmann
24 In Vivo Pharmacokinetics of Magnetic Nanoparticles. . . . . . . . . . . . . . . . . . . . . . . . 409
Carlos Caro, M. Carmen Muñoz-Hernández, Manuel Pernia Leal,
and Marı́a Luisa Garcı́a-Martı́n
PART VII ANESTHESIA AND ADVANCED CONTRAST AGENTS
25 Anesthesia and Monitoring of Animals During MRI Studies . . . . . . . . . . . . . . . . . 423

Jordi L. Tremoleda, Sven Macholl, and Jane K. Sosabowski
Contents ix
26 Advanced Contrast Agents for Multimodal Biomedical Imaging

Based on Nanotechnology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 441
Daniel Calle, Paloma Ballesteros, and Sebastián Cerdán
Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 459
Contributors
NURIA ARIAS-RAMOS Departament de Bioquı́mica i Biologia Molecular, Unitat de

Bioquı́mica de Biociències, Edifici Cs, Universitat Autònoma de Barcelona, Cerdanyola del
Vallès, Spain
PALOMA BALLESTEROS Facultad de Ciencias, Universidad Nacional de Educacion a
Distancia UNED, Madrid, Spain
BHANU PRAKASH KN Signal and Image Processing, Singapore Bioimaging Consortium,
Agency for Science, Technology and Research, Biopolis Way, Singapore
DANIEL CALLE Instituto de Investigaciones Biomédicas “Alberto Sols”, CSIC/UAM,
Madrid, Spain
SANTIAGO CANALS Instituto de Neurociencias, Consejo Superior de Investigaciones
Cientı́ficas, Universidad Miguel Hernández, Sant Joan d’Alacant, Spain
JOSÉ MARÍA CARAMÉS Instituto de Neurociencias, Consejo Superior de Investigaciones
CARLOS CARO BIONAND, Andalusian Centre for Nanomedicine and Biotechnology,
Junta de Andalucı́a, Universidad de Málaga, Málaga, Spain
SEBASTIÁN CERDÁN Instituto de Investigaciones Biomédicas “Alberto Sols”, CSIC/UAM,
Madrid, Spain
MENGLIN CHENG Division of Cancer Imaging Research, Russell H. Morgan Department
of Radiology and Radiological Science, Johns Hopkins University School of Medicine,
Baltimore, MD, USA; Sidney Kimmel Comprehensive Cancer Center, Johns Hopkins
University School of Medicine, Baltimore, MD, USA
ANTOINE CHERIX Laboratory of Functional and Metabolic Imaging (LIFMET), École
Polytechnique Fédérale de Lausanne, Lausanne, Switzerland
BEGOÑA FERNÁNDEZ Instituto de Neurociencias, Consejo Superior de Investigaciones
LUIS MIGUEL FERNÁNDEZ-MOLLÁ Center for Biomaterials and Tissue Engineering,
Universitat Politècnica de València, Valencia, Spain
ULRICH FLÖGEL Experimental Cardiovascular Imaging, Department of Molecular
Cardiology, Heinrich Heine University, D€ usseldorf, Germany
ANTOINE P. FOURNIER Normandie Univ, UNICAEN, INSERM, INSERM UMR-S
U1237, PhIND, Physiopathology and Imaging of Neurological Disorders, Cyceron, Caen,
France
MARÍA LUISA GARCÍA-MARTÍN BIONAND, Andalusian Centre for Nanomedicine and
Biotechnology, Junta de Andalucı́a, Universidad de Málaga, Málaga, Spain; Networking
Research Center on Bioengineering, Biomaterials and Nanomedicine, CIBER-BBN,
Málaga, Spain
MAXIME GAUBERTI Normandie Univ, UNICAEN, INSERM, INSERM UMR-S U1237,
PhIND, Physiopathology and Imaging of Neurological Disorders, Cyceron, Caen, France;
Department of Diagnostic Imaging and Interventional Radiology, CHU Caen, Caen,
France
xi
xii Contributors
KRISTINE GLUNDE Division of Cancer Imaging Research, Russell H. Morgan Department

of Radiology and Radiological Science, Johns Hopkins University School of Medicine,
Baltimore, MD, USA; Sidney Kimmel Comprehensive Cancer Center, Johns Hopkins
University School of Medicine, Baltimore, MD, USA
VENKATESH GOPALAN Signal and Image Processing, Singapore Bioimaging Consortium,
ROLF GRUETTER Laboratory of Functional and Metabolic Imaging (LIFMET), École
Polytechnique Fédérale de Lausanne, Lausanne, Switzerland; Department of Radiology,
University of Geneva, Geneva, Switzerland; Department of Radiology, University
of Lausanne, Lausanne, Switzerland
IRENE GUADILLA Instituto de Investigaciones Biomédicas “Alberto Sols”, CSIC/UAM,
Madrid, Spain
TUBA G€ uDEN-SILBER Experimental Cardiovascular Imaging, Department of Molecular
TILL HUELNHAGEN Berlin Ultrahigh Field Facility (B.U.F.F.), Max Delbr€ uck Center for
Molecular Medicine in the Helmholtz Association (MDC), Berlin, Germany
CHRISTOPH JACOBY Experimental Cardiovascular Imaging, Department of Molecular
MIN-CHI KU Berlin Ultrahigh Field Facility (B.U.F.F.), Max Delbr€ uck Center for
NYOMAN D. KURNIAWAN Centre for Advanced Imaging, The University of Queensland, St.
Lucia, QLD, Australia
BLANCA LIZARBE Laboratory of Functional and Metabolic Imaging (LIFMET), École
Polytechnique Fédérale de Lausanne, Lausanne, Switzerland
SILVIA LOPE-PIEDRAFITA Servei de Ressonància Magnètica Nuclear, Universitat Autònoma
de Barcelona, Cerdanyola del Vallès, Spain; Networking Research Center on
Bioengineering, Biomaterials and Nanomedicine (CIBER-BBN), Universitat Autònoma
de Barcelona, Cerdanyola del Vallès, Spain
PILAR LÓPEZ-LARRUBIA Instituto de Investigaciones Biomédicas “Alberto Sols”, CSIC/
UAM, Madrid, Spain
SVEN MACHOLL Centre for Molecular Oncology, Barts Cancer Institute, Queen Mary
University of London, London, UK
GARY V. MARTINEZ Department of Cancer Imaging and Metabolism, H. Lee Moffitt
Cancer Center & Research Institute, Tampa, FL, USA
SARA MARTINEZ DE LIZARRONDO Normandie Univ, UNICAEN, INSERM, INSERM
UMR-S U1237, PhIND, Physiopathology and Imaging of Neurological Disorders, Cyceron,
Caen, France
RALPH P. MASON Prognostic Imaging Research Laboratory, Department of Radiology, UT
Southwestern Medical Center, Dallas, TX, USA
DAVID MORATAL Center for Biomaterials and Tissue Engineering, Universitat Politècnica
de València, Valencia, Spain
ANDREA MORENO Instituto de Neurociencias, Consejo Superior de Investigaciones
Cientı́ficas, Universidad Miguel Hernández, Sant Joan d’Alacant, Spain; Center for
Biomaterials and Tissue Engineering, Universitat Politècnica de València, Valencia, Spain
JENNIFER MOROZ Department of Physics and Astronomy, The University of British
Columbia, Vancouver, BC, Canada
ERIC R. MUIR Department of Ophthalmology, Research Imaging Institute, University of
Texas Health Science Center at San Antonio, San Antonio, TX, USA
Contributors xiii
M. CARMEN MUÑOZ-HERNÁNDEZ BIONAND, Andalusian Centre for Nanomedicine and

Biotechnology, Junta de Andalucı́a, Universidad de Málaga, Málaga, Spain
EMMA MUÑOZ-MORENO Experimental 7T MRI Unit, IDIBAPS, Barcelona, Spain
TERESA NAVARRO Instituto de Investigaciones Biomédicas “Alberto Sols”, CSIC/UAM,
Madrid, Spain
THORALF NIENDORF Berlin Ultrahigh Field Facility (B.U.F.F.), Max Delbr€ uck Center for
Molecular Medicine in the Helmholtz Association (MDC), Berlin, Germany; DZHK
(German Centre for Cardiovascular Research), Berlin, Germany
JESÚS PACHECO-TORRES Instituto de Neurociencias, Consejo Superior de Investigaciones
DANIEL PADRO Molecular Imaging Unit, CIC biomaGUNE, Donostia-San Sebastián,
Spain
LAURA PÉREZ-CERVERA Instituto de Neurociencias, Consejo Superior de Investigaciones
ELENA PÉREZ-MONTOYO Instituto de Neurociencias, Consejo Superior de Investigaciones
MANUEL PERNIA LEAL Departamento de Quı́mica Orgánica y Farmacéutica, Universidad
de Sevilla, Sevilla, Spain
ANDREAS POHLMANN Berlin Ultrahigh Field Facility (B.U.F.F.), Max Delbr€ uck Center for
MARILENA PREDA Buffalo Neuroimaging Analysis Center, Department of Neurology, Jacobs
School of Medicine and Biomedical Sciences, University at Buffalo, The State University of
New York, Buffalo, NY, USA; Center for Biomedical Imaging, Clinical and Translational
Science Institute, University at Buffalo, The State University of New York, Buffalo, NY,
USA
PEDRO RAMOS-CABRER Molecular Imaging Unit, CIC biomaGUNE, Donostia-San
Sebastián, Spain; Ikerbasque, Basque Foundation for Science, Bilbao, Spain
STEFAN A. REINSBERG Department of Physics and Astronomy, The University of British
Columbia, Vancouver, BC, Canada
RITA MARIA ROCHA OLIVEIRA Instituto de Investigaciones Biomédicas “Alberto Sols”, CSIC/
UAM, Madrid, Spain
FERDINAND SCHWESER Buffalo Neuroimaging Analysis Center, Department of Neurology,
Jacobs School of Medicine and Biomedical Sciences, University at Buffalo, The State
University of New York, Buffalo, NY, USA; Center for Biomedical Imaging, Clinical and
Translational Science Institute, University at Buffalo, The State University of New York,
Buffalo, NY, USA
RUI V. SIMÕES Fetal i+D Fetal Medicine Research Center, BCNatal-Barcelona Center for
Maternal-Fetal and Neonatal Medicine (Hospital Clı́nic and Hospital Sant Joan de Déu),
Institut Clı́nic de Ginecologia, Obstetricia i Neonatologia, Institut d’Investigacions
Biomèdiques August Pi i Sunyer (IDIBAPS), Universitat de Barcelona, and Centre for
Biomedical Research on Rare Diseases (CIBER-ER), Barcelona, Spain; Champalimaud
Foundation, Lisbon, Portugal
GUADALUPE SORIA Experimental 7T MRI Unit, IDIBAPS, Barcelona, Spain; Networking
Research Center on Bioengineering, Biomaterials and Nanomedicine (CIBER-BBN),
Barcelona, Spain
JANE K. SOSABOWSKI Centre for Molecular Oncology, Barts Cancer Institute, Queen Mary
xiv Contributors
SEBASTIAN TEMME Experimental Cardiovascular Imaging, Department of Molecular

JORDI L. TREMOLEDA Centre for Trauma Sciences, Blizard Institute, Queen Mary
RAÚL TUDELA Networking Research Center on Bioengineering, Biomaterials and
Nanomedicine (CIBER-BBN), Barcelona, Spain
DANIEL H. TURNBULL Department of Radiology, Bernard and Irene Schwartz Center for
Biomedical Imaging, New York University (NYU) School of Medicine, New York, NY,
USA; Department of Pathology, NYU School of Medicine, New York, NY, USA; Kimmel
Center for Biology and Medicine at the Skirball Institute of Biomolecular Medicine, NYU
School of Medicine, New York, NY, USA
S. SENDHIL VELAN Metabolic Imaging Group, Singapore Bioimaging Consortium, Agency
for Science, Technology and Research, Biopolis Way, Singapore
SANJAY K. VERMA Signal and Image Processing, Singapore Bioimaging Consortium, Agency
for Science, Technology and Research, Biopolis Way, Singapore
DENIS VIVIEN Normandie Univ, UNICAEN, INSERM, INSERM UMR-S U1237,
PhIND, Physiopathology and Imaging of Neurological Disorders, Cyceron, Caen, France;
Clinical Research Department, CHU Caen, Caen, France
SONIA WAICZIES Berlin Ultrahigh Field Facility (B.U.F.F.), Max Delbr€ uck Center for
DAN WU Department of Radiology, Johns Hopkins University School of Medicine,
Baltimore, MD, USA
JADEGOUD YALIGAR Signal and Image Processing, Singapore Bioimaging Consortium,
JIANGYANG ZHANG Department of Radiology, Bernard and Irene Schwartz Center for
Biomedical Imaging, New York University (NYU) School of Medicine, New York, NY,
USA
HELING ZHOU Prognostic Imaging Research Laboratory, Department of Radiology, UT
Southwestern Medical Center, Dallas, TX, USA
ROBERT ZIVADINOV Buffalo Neuroimaging Analysis Center, Department of Neurology,
Jacobs School of Medicine and Biomedical Sciences, University at Buffalo, The State
University of New York, Buffalo, NY, USA; Center for Biomedical Imaging, Clinical and
Translational Science Institute, University at Buffalo, The State University of New York,
Buffalo, NY, USA
Part I
MRI Basics
Chapter 1
Introduction to MRI Physics

Gary V. Martinez
Abstract
Magnetic resonance imaging (MRI) is an imaging technique derived from radiofrequency (RF) signals of
proton that are magnetized by a strong magnetic field. These protons typically originate from water, fat, or
metabolites. The application of RF pulses is used to excite the magnetization, whereas pulsed magnetic field
gradients are used to provide spatial localization. This chapter describes the fundamental principles giving
rise to MR images. Furthermore, the connection between relaxation and image contrast is discussed.
Key words Pulse sequences, Encoding, k-Space, Relaxation, Contrast
1 Introduction
Magnetic resonance imaging (MRI) is an enormously rich tech-

nique with broad applications in preclinical biomedical imaging
research and in clinical practice. MRI has many advantages as an
imaging modality in that it provides the ability to exploit multiple
types of soft tissue contrast, while having excellent penetration
throughout the body. A vast array of contrast mechanisms provide
invaluable information due to multiple biophysical mechanisms at
play that, upon analysis and interpretation, can yield rich insight
into biological, biochemical, and medical phenomena. These pro-
vide the capability to scan and potentially detect afflicted tissues.
Currently, there are MRI tools that can be used to acquire an
image that has a particular type of contrast. These scans can poten-
tially detect multiple maladies that potentially can offer diagnostic
images with exquisite sensitivity. However, to do so requires a clear
understanding of the underlying tools, and how to best apply them
to extract the most diagnostic information possible.
In basic research and clinical practice, there are several key MR
parameters that provide informative contrast, and are the most
typical scans. These include, for example, T1 (longitudinal), T2
(transverse), and apparent diffusion coefficient (ADC). In addition,
less commonly used scans include: T2* (transverse), chemical
Marı́a Luisa Garcı́a-Martı́n and Pilar López-Larrubia (eds.), Preclinical MRI: Methods and Protocols, Methods in Molecular Biology,
vol. 1718, https://doi.org/10.1007/978-1-4939-7531-0_1, © Springer Science+Business Media, LLC 2018
3
4 Gary V. Martinez
exchange saturation transfer (CEST), and others. MRI pulse-

sequences are designed to exploit contrast mechanisms, which
occur on the molecular scale, to the mesoscopic scale and provide
rich noninvasive biophysical insight into tumor status. The most
useful parameters can also be evaluated using extensive post-
processing. The resulting images can be physically interpreted
with regard to physiological consequence.
The longitudinal (spin lattice) relaxation time constant T1, is
widely used in conjunction with an exogenous T1-shortening con-
trast agent (CA) to provide positive enhancement. Numerous T1-
shortening CAs have been developed and are routinely used in the
clinic. They vary in size and in their innate ability to shorten T1 at a
given concentration, which is quantified by its spin lattice relaxivity
(r1). The most widely used clinical agents are of the small molecule
variety, which cause signal enhancement in a T1-weighted image.
The rate at which a CA is delivered to tissue can be used to quantify
perfusion. Typically a time-series of images is collected during bolus
administration, and analyzed to show pixels that are enhanced and
the rate at which they do so. These are dependent on (1) flow of
blood, and agent, to a specific region and hence the presence of
vasculature, and (2) the permeability of that vasculature. Hence,
chaotic and permeable vasculature are depicted through the appli-
cation of T1-sensitive pulse sequence along with a suitable T1
shortening CA. Injectable agents vary in size, and biophysical
mechanism, and can be applied to various diseases.
Shortening of T2 is characterized by decreasing transverse
coherence, resulting in an attenuated signal. Thus, T2-shortening
is classified as negative contrast. T2 sequences provide insight into
endogenous molecular contrast, which result from a multitude of
sources ranging from self relaxation (molecular motions and chem-
ical properties) to intermolecular relaxation. T2 is also affected by
variations in temperature and viscosity. For example, in tumors are
often characterized by larger T2 values resulting in increased bright-
ness in a T2-weighted image. Peritumoral edema presents as
increased T2. In preclinical images, regions that experience greater
hemorrhagic necrosis may have hypo-intense pixels, whereas they
may show hyper-intensity in clinical images.
The biophysical mechanisms that give rise to ADC include
translational diffusion (affected by temperature and viscosity) and
the extent of restriction to water molecule displacement caused by
the variations in the local cellular density in a given voxel. Regions
of tissue necrosis are marked by increases in ADC.
It has been said that the Achilles heel of MRI is that it suffers
from poor sensitivity. Although true, there are large numbers of
MRI sensitive protons (water) that are present in the body. In
addition, strategies exist to help make up this shortcoming, as
there are clearly numerous technical capabilities possible with
MRI. The limits of applicability are imposed by the current state
MRI Physics 5
of the art, and there exist many difficult challenges to overcome,

and also exciting opportunities. Spectroscopic applications are
often aimed at metabolites in the mM concentration range, and
poor sensitivity makes many existing techniques challenging but
possible for routine use. A good example of an up and coming
approach is through hyperpolarization to increase sensitivity
[1, 2]. Beyond this, there are a number of other imaging techniques
that are seeing increased active research such as the application of
frequency specific pulses that are capable of discerning chemical
moieties with exchangeable protons, which may be exploited to
provide information on the exchange occurring by measuring sol-
vent saturation effects on the solvent pool.
In this chapter, the aim is to provide insight into the principles
of MRI, to provide a conceptual framework for the myriad of recent
advances in MRI applied to biomedical research and medicine.
2 Basic MRI Pulse Sequences
MRI is a form of Nuclear Magnetic Resonance (NMR), which has

been a source of a number of fundamental discoveries in physics,
chemistry, and medicine [3, 4] (Rabi, Bloch, Lauterbur, etc.) and
has lead to multiple Nobel prizes. The field has experienced a
continual advancement, from the initial discovery of NMR by
Rabi, the initial NMR measurements in tumors by Damadian [5],
the image of Lauterbur [6], and up to the present. MRI from its
onset has been distinguished by its ability to detect disease in vivo.
The underlying physical principles are widely applicable, and given
the precedent of previous innovations in the field, it is likely that
many excellent future innovations will follow.
The MR pulse-sequence is the instruction set that allows one to
select how an image scans the anatomy, or can be tuned to obtain
functional and molecular information. It consists of radio fre-
quency (RF) and gradient pulses that can be incorporated into a
greater coordinated scheme. The design of the sequence is critical
for detecting information that is embedded in the physicochemical
environment. In combination with the appropriate hardware, it can
also provide information on specific metabolic pathways. In princi-
ple, these can be related to the physiological status of the patient
being imaged.
From a practical perspective, it is essential to understand what
effects various MR parameters have on an image in order to opti-
mize them to achieve a high quality image. As a first step, it is
helpful to describe how raw data are acquired, and how these are
transformed to produce an image. These are described in this
section, and will provide an introduction into how MR images
depict the underlying physical phenomena.
6 Gary V. Martinez
2.1 The Mechanics The acronym MRI is used for imaging and MRS refers to the
of MR Image spectroscopy of in vivo systems (which will be used here instead
Acquisition of NMR). MR images are made from H2O molecules in the body.
Due to their abundance, they compensate for low sensitivity.
Because the initial magnetic resonance experiments were per-
formed to probe the spectroscopic principles involved in spin phys-
ics, only a magnetic field and RF electronics were needed.
Historically, MR was developed without applied gradients, result-
ing in a non-localized spectrum. A critical distinction is that captur-
ing an MR image additionally requires the application of temporal
magnetic field gradients along three axes.
Because of the historical development, and because MR spec-
troscopy is an active and ongoing area of research, we will begin
with the explanation of an MR spectrum and proceed to the more
complicated MRI scan. In short, they both give spectroscopic
(frequency) information, but the image results from the signals
emanating from distinct and separate physical locations, whereas a
spectrum comes from the entire object. Wherever 1H density is
present, the localized signals can be displayed in terms of distances,
or coordinates on a Cartesian grid, hence forming an image. Fur-
thermore, localized images can be obtained providing spectro-
scopic (MRS) noninvasive insight into the interior of an object.
For example, common applications include images in the interior of
the knee, the brain, and in the diagnosis and therapy of cancer.
In addition to these, a hybrid type of acquisition can be per-
formed, giving rise to spectroscopic imaging (MRSI). In MRSI, it is
typically run with two spatial dimensions and 1 spectroscopic
dimension, although 3D spatial is possible. The primary examples
of these are shift imaging (CSI) [7], and echo planar spectroscopic
imaging (EPSI) [8, 9]. This chapter will not go in depth into
MRSI, but rather focus on the commonality of the basic principles
of spectroscopy and imaging.
2.1.1 Larmor Frequency When placed in a magnetic field, spin ½ MR active nuclei experi-
and RF Excitation ence a separation into two energy levels. The separation in energy
between the two levels results in a frequency difference. In MRI,
the peak center frequency is where the 1H in water resonates and is
given by:
ω0 ¼ γ H B 0 ð1Þ
where ω0 is the angular frequency, γ H is the gyromagnetic ratio for

proton, and B0 is the static magnetic field. It is interesting to note
that the MRI frequency for a particular nucleus (typically proton) is
on a Hz scale, which is ν0 ¼ ω0/2π, and is referred as the Larmor
frequency.
MRI Physics 7
Fig. 1 Fourier transform of single doubly exponential decaying signals: (a) with no sinusoidal modulation: the
on-resonance condition. (b) Fourier transform of (a). (c) Signal with two components, both shifted and hence
modulated creating an interference pattern. (d) Fourier transform of (c), showing a spectrum that is far more
simplistic than the superposition time-series signal in (a)
An RF pulse is a critical building block in magnetic resonance

pulse sequences. Each one is applied to achieve a specific flip angle,
which for a hard pulse is defined as:
α ¼ 2 π τ γH B1 ð2Þ
Here τ is the duration of the pulse, and B1 is the amplitude of

the applied pulse, which is much smaller relative to the strength of
the static field. When multiple pulses are present in the pulse
sequence, a combination of flip angles and delays may be applied.
The simplest combinations include a single pulse followed by data
acquisition. This results in a free induction decay (FID) (Fig. 1a).
The FID does not contain explicit spatial information, but rather is
a reporter on the spectroscopic details of the water molecule, which
include chemical shift and linewidth. In order to rationalize the
obtained signal, either in spectroscopy or imaging, a mathematical
tool is needed, and there is one that is beautifully suited to the task:
the Fourier Transform (FT).
2.1.2 Fourier Transform The MR signal is a periodic function of time, which is quite com-
plicated and is not readily understandable by the human eye. How-
ever, there exists a transformation of these time-series data to a
8 Gary V. Martinez
more accessible representation. This operation is the fulcrum of

MRI and MRS: the FT. It performs its magic by decomposing a
complex signal into separate frequency components:
ð1
F ðνÞ ¼ f ðt Þe i2πνt dt ð3Þ
1
where t is time, f(t) is the sinusoidal signal, and ν is the frequency.

The result is a spectrum if the evolution of the time-series data does
not contain position dependent phase or frequency modulations
from gradients. If these modulations are present, then the FT
operation results in an image, assuming 2D data and Cartesian
acquisition.
This transform goes in two directions, and the original signal
can be recovered from its inverse transform:
ð1
f ðtÞ ¼ F ðvÞe i2πνt dv ð4Þ
1
It is particularly useful for making a sum of multiple decaying

signals understandable, given that the original sinusoid is a sum of
two or more frequency components. The signal can be transformed
to a simplified form that shows two peaks at two distinct frequen-
cies (Fig. 1a, b). The amplitudes of these peaks are also retained
based on the weighted contributions of each component to the
original signal. In short, we see a complicated signal that is trans-
formed into a form that may be readily understood, whether it is a
1D spectrum or an image.
2.1.3 MR Spectrum After FT, an MRS spectrum is achieved in the absence of magnetic
field gradients that encode after application of an RF pulse or
during acquisition of signal. It displays the spectroscopic character-
istics of a water molecule, such as the breadth of the peak, which is
known as the linewidth. The linewidth of a peak has an inverse
relationship with the rate of decay in the acquisition data. The FID
and spectrum are related through Fourier transformation, as shown
in Fig. 1a, b. However, due to heterogeneity in an image or
differences in the chemical shift, protons may resonate at different
frequencies, resulting in a superposed FID that is modulated by
different frequencies, which Fourier transforms into a spectrum of
broadened and/or multiple peaks. There are many applications to
understanding spectroscopic peak locations. For example, there is a
wide array of characteristic metabolite frequencies. Also, water in
tissue may experience different magnetic susceptibilities, which
could inform on the content and proximity of structural interfaces,
such as those due to bone, air, or regions of high iron content.
It is of interest to systematically introduce local magnetic field
changes so that signal from a particular position can be
MRI Physics 9
Fig. 2 Effect on resonance frequency due to a magnetic field gradient along the
z direction: there is a linear dependence of frequency and spatial position with
magnetic field
distinguished. This amounts to labeling and/or encoding a specific

location in three-dimensional space based on its frequency or phase.
The application of magnetic field gradients allow precise control
over this process and open up a wide range of strategies to produce
images. The next section describes the fundamental building blocks
and basic sequences to achieve this.
2.1.4 Magnetic Field A magnetic field gradient is defined as the derivative of the mag-
Gradient netic field as a function of spatial coordinate (Gz ¼ dBz/dx). A linear
gradient is transiently applied in a particular direction in the labo-
ratory frame (Fig. 2). If one considers a tube of pure water, it is
quite homogenous and produces a single MR peak, and if well
shimmed, results in narrow breadth (linewidth). However, in the
presence of a magnetic field gradient, a distribution of frequencies
results.
2.1.5 MRI: The Effect All of the water protons resonate at the Larmor frequency, yet the
of Magnetic Field Gradients frequency changes when turning on a magnetic field gradient. This
change is spatially dependent. If one assumes that a gradient is
applied in the z direction, then the resonance condition is given by:
ωðz Þ ¼ γ H þ γ H z G z ð5Þ
where z is the axis in which Gz is applied. This equation is one of the

most important ones in MRI as it describes the basis of how MR
images are generated. Spatial information is derived from this rela-
tion in different ways, depending on the details of a specific pulse
sequence design. Equation (5) describes where spins are excited
with slice selection, and how the other directions are encoded either
through frequency and/or phase. It provides explanation for how
the major image formation techniques in MRI work. The gradients
are not limited to a single direction and may also be applied in the
10 Gary V. Martinez
other laboratory axes, x, y, or in some combination of directions, to

encode 3D information. Once encoded, a mathematical transfor-
mation (e.g., FT) relates the signal to an image. An overview of the
entire process of how images are achieved is now given.
2.1.6 Slice Selection Once again considering an object with uniform 1H density, the
application of a gradient, along the z-direction, Gz, generates a
range of resonance frequencies across the object in that direction.
Linear change in the magnetic field as a function of position gives
rise to linear change in frequency. If one had a way to simulta-
neously excite multiple frequencies at once, then one could satisfy
the resonance condition (Eq. 1) for each of those frequencies, and
this would result in multiple locations being excited. This is where
the significance of the shape of an RF pulse comes into play.
There is indeed a relationship between the temporal domain
and the spectral domain. The shorter the RF pulse is, the broader
its excitation profile is, and the wider the range of frequencies that
are excited. The excitation pulse shape will dictate the shape of the
band of spins that is excited. In order to end up with a slab or slice
that is approximately a rectangular cuboid, RF pulses must have a
shape in the time domain. For example, a low flip angle sinc shaped
RF pulse has an approximately rectangular spectral profile, and
hence, excites a rectangular shaped band of spins in the object
that is being imaged (Fig. 3).
The strength of a signal is dictated in part by how many spins
were excited at that particular location, and other factors such as
slice thickness and pixel resolution (a signal equation and para-
meters used for acquisition), and these are influenced by the details
of how the RF pulse and the gradient were applied. At a given
gradient strength, the RF pulse excitation profile describes how
thick the slice is and the frequency of the pulse determines the
spatial location of the slice (Eq. 5) and Fig. 3). In particular, the
thickness of the slice can be determined from:
Δν ¼ γ H G z Δz ð6Þ
which shows that the pulse width Δν relates to the slice thickness Δz,
based on the gradient strength and gyromagnetic ratio. In a multi-
slice experiment, additional bands of spins, typically referred to as
slices, can be excited by changing the frequency of the pulse, which
serves to alter the location in the direction of Gz (Fig. 3). Once the
spins in a slice are excited, it is of interest to encode the other
directions.
2.1.7 Frequency When an RF pulse is applied at the Larmor frequency, it tips

Encoding magnetization down into the transverse plane after which it relaxes
in the transverse plane through T2 processes. To generate an image,
a scheme is needed to encode the other dimensions. The most
MRI Physics 11
Fig. 3 Schematic depiction of: (a) Spectroscopy and (b and c) 1D imaging (profile). In the spectroscopic case, a
slice-selective RF pulse excite a band of spins but is not spatially encoded after the initial RF pulse. In (b), the
spins are frequency encoded as indicated by the frequency encoding gradient being on during acquisition of
data. (c) A phase encode gradient precedes, and is turned off before, the acquisition of data. For the purposes
of comparison, the same direction is encoded in (c) as in (b). The differences between the spectroscopy and
imaging scenarios highlight the effects of encoding by a gradient and its ability to spatially localize signal
common way to achieve this is through a frequency-encoding

gradient. It is schematically depicted in Fig. 3, where data points
are collected while the gradient is on. Because different spins in the
object are situated at different spatial locations, they will experience
different fields, and according to Eq. (5), they will produce signals
with different resonance frequencies.
An interesting aspect of frequency-encoding is that the fre-
quencies are actually different because the gradient is on while
sampling of data is occurring. Sampling in this way is occurring
through direct detection of signal, and not subject to folding or
aliasing of signal outside of the sampled spectrum, which distin-
guishes it from indirect (i.e., phase) encoding.
Consider a single direction, say Gy, where we know from
Eq. (5) that the resonance frequency is dependent on the distance
from the isocenter multiplied by the gradient strength. Thus, each
position of a frequency-encoding gradient is distance and gradient
strength dependent. One important implication of this is that while
the gradient is on, a linear phase is accumulating. This phase is
12 Gary V. Martinez
dependent on the spatial position of a pixel and the strength of the

gradient. A strategy for collecting the entire echo is to precede the
read (frequency encoding) gradient with a negative gradient of half
the area of the read gradient. The shifts and phases are position
dependent, except at the center of the gradient, where all “reso-
nances” overlap and have zero phase, and is where maximal super-
position, and hence signal, occurs.
2.1.8 Phase-Encoding A phase-encoding gradient (Fig. 3c) imparts a distance and gradi-
ent dependent phase shift, where sampling is not performed simul-
taneously, which distinguishes it from frequency encoding. The end
result is the same as in a frequency-encoding gradient, but in the
phase-encoding case, there is a temporary frequency change while
gradient is turned on. After it is shut off, one could almost imagine
that the frequency information is lost; yet, there is memory of the
phase that is accumulated during the time that the gradient was on,
and this is retained by the spins in a position dependent manner.
In many pulse sequences, a phase-encode gradient is typically
applied orthogonal to a read gradient and is able to indirectly
encode over the course of the entire acquisition of multiple repeti-
tion time (TR) periods, rather than in a single shot (echo planar
imaging excepted).
The matrix size and gradient strength determine the in-plane
resolution of the image. In phase-encoding, the number of phase-
encode steps is the number of samples in this direction, and as such
dictates the resolution. Because phase-encoding occurs over the
course of the indirect dimension, it can be affected by changes
that may occur on this time-scale. Such changes include motion
and/or flow. Another limitation of the phase-encode gradient is
that sampling occurs without any hardware filtering and is suscep-
tible to aliasing artifacts. In such cases, one must either have a
sufficiently large field of view, or combine it with saturating
bands, to eliminate or reduce the amount of wrap around, which
can corrupt the image.
2.2 Reconstruction The k-space formalism is a convenient means of describing the

spatial frequencies that are present in the object being imaged.
2.2.1 K-Space
However, it is important to re-emphasize that these frequencies
do not exist without the application of a magnetic field gradient,
which gives rise to spatial frequencies, rather than the spectroscopic
frequencies that would exist without a gradient. This suggests that
k-space and image play a role analogous to the Fourier relationship
between FID and spectrum. Amplitude, frequency, and phase are
characteristics that are present in k-space signals.
What is particularly convenient about the k-space formalism is
that it allows a compact description of how a particular pulse
sequence can be designed to cover positive and negative values of
k-space. For example, the description above discusses dephasing
MRI Physics 13
such that both sides of an echo can be obtained. However, in the k-

space view, the dephasing gradient is used to traverse out to a
negative value of k-space, so that the effect of a read gradient is to
cause a gradual increase to positive k value. This traversal is captured
because sampling of signal is typically happening simultaneously.
Therefore, the execution of a read gradient, say Gx, causes a rapid
passage in k-values from –kx to kx, whereas increments in the phase-
encode gradient happen stepwise throughout the scan, and so
passage from –ky to ky, requires the entire scan to complete. In
this way, it is possible to raster through k-space to get complete
coverage. In more advanced forms of imaging, that are beyond the
scope of discussion, it is possible to simultaneously apply gradients
in more than one read direction, resulting in non-Cartesian sam-
pling of k-space. Such k-space trajectories as spiral, radial, propeller,
have their challenges to implement, but offer useful advantages.
The utility of the k-space paradigm is in its facility in describing all
of these scanning schemes.
2.3 Basic Sequences The tools for MRI, described above, form the foundation for more
advanced sequences. They are an integral part of both gradient
echo or spin echo sequences. As such, most of the advanced
sequences in MRI are some variant of these. We will therefore
discuss these sequences to demonstrate the basic principles of
how to create an MR image. After a discussion of relaxation in
following sections, we shall describe the basic acquisition para-
meters that can be adjusted to determine relaxation and contrast
characteristics of the image, and impact how they can be applied to
provide physiological insight.
2.3.1 Gradient Echo A gradient echo (GE) sequence is shown in Fig. 4. Considering first
the RF pulse, it is composed of a single pulse, a phase encode gradi-
ent, and then acquisition of signal during the frequency encoding
gradient. If one is interested in the spectral characteristics, we
would acquire data across time. However, our immediate concern
is the formation of an image, which means that we hope to sample
k-space long before the FID has completed. We thus apply an RF
pulse in the presence of a slice-select gradient, and execute a phase-
encoding gradient, and a frequency-encoding gradient as part of a
single unit, a repetition. Within each frequency-encoding gradient,
there are data points acquired while the gradient is on (NFE). This
TR period gets repeated multiple times (NPE), where the phase-
encoding gradient gets varied each time. After scanning is com-
plete, raw k-space data contain a matrix of dimension NFE NPE,
which, together with the field of view (FOV), defines the in-plane
resolution of the scan. The through plane resolution is defined by
the slice thickness (described above).
It should be noted that a GE sequence cannot refocus the
effects of magnetic field inhomogeneity.
14 Gary V. Martinez
Fig. 4 A gradient echo sequence. The RF pulse is applied during the first slice
selection gradient, and is immediately followed by a negative refocusing lobe of
half the slice select area. A single element of the phase-encode gradient is
shown. Throughout the entire acquisition of a scan, the phase-encode gradient
iterates to different values that range from positive to negative. The read gradient
is preceded by a dephasing lobe such that both sides of the echo may be
captured during the read. During the read gradient, the signal is
simultaneously being sampled. For each RF excitation, the echo time is the
time from the middle of the RF pulse to the center of k-space, which is the
maximum of the signal. The repetition time (TR) is the duration for the entire unit.
Note that the gradients are depicted as rectangles, but in reality are closer to
trapezoids
2.3.2 Spin Echo The spin echo (SE) was first described by Hahn [10]. A spin echo
imaging sequence is shown in Fig. 5. The main difference between
the SE is that it can refocus the effects of magnetic field inhomoge-
neity. Unlike the GE, the spin echo sequence decays according to
T2, which is often much longer than T2*, meaning that it can be
used to increase the TE to large times (40–90 ms), which might
otherwise not be feasible. In fact, differences in T2 values at much
longer TE values can be interrogated with T2-weighted imaging,
which provides rich contrast. A weakness of it is that it is not
sensitive to short echo times as the time required to execute an
additional RF pulse and gradient makes the minimum TE signifi-
cantly longer than that for a GE sequence.
3 Relaxation and Image Contrast
In MR imaging, the underlying biophysical mechanisms that mod-

ulate signal are leveraged to report on state of disease. There is an
emerging area of research that focuses on extracting information
from these images: Quantitative Imaging Biomarkers (QIB). QIB
are reproducible measures of normal physiological process, a
MRI Physics 15
Fig. 5 A spin echo sequence. The overall design is similar to the gradient echo
pulse sequence; however, in brief there are some important distinctions. There
are two RF pulse that make up a spin echo, which is an RF phenomenon. The
first RF pulse is a 90 pulse followed by a delay (TE/2) and then a 180 pulse.
Acquisition of data is timed such that the center of k-space, a gradient induced
phenomenon, coincides with the spin echo maximum at TE. The time to repeat
this entire unit is the repetition time, TR
diseased state, or a response to therapeutic intervention. To

increase the range of imaging features that may qualify as useful
QIB, it is helpful to obtain images that are sensitive to different
physiological phenomena. There are many ways of doing this, yet,
we will focus on the most commonly applied techniques for achiev-
ing contrast.
Each imaging sequence and set of acquisition parameters con-
fers a specific type of contrast mechanism. Each contrast mechan-
isms is sensitive to and reports on physiological and physical
processes that may be interpretable in terms of disease progression
and/or response to therapy. Usage of multiple pulse-sequences,
and experimental acquisition strategies, gives rise to unique infor-
mation, and if used in combination with one another can provide
complementary information that, as a whole, is quite powerful.
3.1 Relaxation Types Relaxation is a phenomenon inherent to MR that can be used to

bias various imaging sequence parameters in such a way as to
leverage biophysical phenomena in order to provide rich and
detailed information about the tissue being imaged. The principles
that govern relaxation can be considered a large and sometimes
challenging subject. However, the basic relaxation rates are given
by time constants T1 and T2. T1 relaxation is governed by exchange
of energy of a nuclear spin with its environment; hence it is given
the name spin-lattice relaxation, or longitudinal relaxation. After
magnetization has been tipped into the transverse plane, T2
16 Gary V. Martinez
relaxation is the rate constant that describes signal decay in that

plane. This relaxation describes the loss of coherence in the trans-
verse component, and does not necessarily require exchange of
energy with the lattice.
Observable MR signal is possible because MR active nuclei are
present in the tissue. The primary nucleus of interest is 1H, which is
present at all levels from water and fat to metabolites. Each MR
active nucleus has a magnetic moment, which experiences a torque
when placed in a magnetic field. This torque causes the magnetic
moment to precess about the magnetic field, but on average is
observed as an alignment with the magnetic field. Because of the
inherent insensitivity of MR, we are relegated to observing magne-
tization, which is a macroscopic phenomenon. This means that in
order to observe a signal in MR, it is necessary to sense a sufficiently
large collection of the magnetic moments that have been summed
together, and describe how a system returns to thermal equilibrium
after being perturbed. The phenomenological Bloch equations
describe how magnetization behaves in the presence of a magnetic
field. They are coupled differential equations that take into account
the precession that occurs in the magnetic field and by time con-
stants that describe how longitudinal and transverse magnetization
return to equilibrium.
3.1.1 Longitudinal Magnetization is aligned with the main static magnetic field B0, and
Relaxation by convention is set to point along the z-axis. After excitation on
resonance with an RF pulse, the return to longitudinal thermal
equilibrium is well described by a relaxation time constant T1, in
Eq. (6).
d 1
M z ¼ ðM z M 0 Þ ð6Þ
dt T1
where Mz is the magnetization in the z direction and M0 is its
equilibrium value. Assuming a 90 pulse, the solution to this is
given by Eq. (7).

M z ðt Þ ¼ M 0 1 e t=T 1 ð7Þ
Experimentally, there is a litany of methods available to measure

T1. Of these, the gold standard method is using an inversion
recovery sequence. One of the drawbacks of this is that it can be
somewhat time consuming, and often progressive saturation is
used, where TR is gradually decreased, causing the signal to be
saturated with each decrement in TR. Due to the time required for
these techniques, other one-shot techniques can be used (Look and
Locker 1970, Crawley and Henkelman 1988). This decaying func-
tion can be fit to the data using nonlinear least squares fitting using
a slightly modified form in Eq. (8):
MRI Physics 17

M z ðt Þ ¼ M 0 1 e TR=T 1 ð8Þ
Here TR is used in place of t. Alternatively, linear regression of the

logarithms of the data and the equation can be performed; how-
ever, the error from this approach may be non-Gaussian. Spin
lattice relaxation is sensitive to various factors that also influence
transverse relaxation. Some tissues have intrinsically shorter T1
values, such as fat for example, which can be made use of to
distinguish tissue regsions without exogenous contrast.
Nonetheless, exogenous contrast agents can be injected that
are quite powerful and provide significant diagnostic value. A typi-
cal lanthanide-containing contrast agent (CA) is capable of
shortening the T1 of water that comes into contact with it, and
has water and/or protons exchanging into close proximity to the
metal center. The efficiency, that CAs are measured by, is shown in
terms of the relaxation rate, which is the inverse of the relaxation
time (R1 ¼ 1/T1). It describes how much the relaxation rate
increases due a known concentration of CA. This is denoted by a
lower case r:

r 1 ¼ R1 R1, 0 =½CA ð9Þ
where R1 is the relaxation rate constant after injection of contrast

agent, and R1,0 is the relaxation rate before injection of contrast.
The concentration of agent is denoted by [CA].
3.1.2 Transverse After excitation by a 90 pulse, relaxation in the transverse plane is
Relaxation governed by the equations:
d M xy
M xy ðt Þ ¼ ð10Þ
dt T2
This results in transverse signal given by:
M xy ¼ M 0 e t=T 2 ð11Þ
T2 relaxation describes the phase coherence of magnetization in the

transverse plane. Multiple factors give rise to variation in magnetic
field across a given tissue, and these variations lead to differences in
T2. A fundamental aspect of T2 is that it is limited by T1, and in fact
T2 T1. T2-weighted images rely upon the fact that subtle differ-
ences in T2, in different organs and parts of organs, can provide
excellent discrimination. These differences are quantified by the
contrast changes. In addition, the fact that diseased tissue will
possess vastly different T2 values can be of great significance in the
diagnosis, progression, and therapy response of a number of
diseases.
T2 can be measured in a number of ways. The simplest method
includes running an array of spin echo experiments where each
18 Gary V. Martinez
array element (image) corresponds to an increase in TE. After

several increments, the TE has increased sufficiently that much of
the signal intensity across the image has decayed. On a pixel-by-
pixel basis, each point in the image can be fit to a decaying expo-
nential. After iterating through every pixel in the image, the result is
a T2 map.
T2* is a measure of traverse relaxation using only gradient echo
(single RF pulse) measurements. It is highly affected by variations
in field across the image as the helpful refocusing effects of a 180
pulse are not present. This results in a time that can be significantly
shorter than T2 if the magnetic field changes (B0) across the images
are significant.
A more efficient and accurate method for determining T2 is to
use a multiple echo approach, where instead of just a spin echo, a
single 180 pulse is substituted with a train of 180 pulses, and
results in a train of echoes. This approach limits the amount of
susceptibility related error, which provides a better measurement of
T2 as there is limited contamination by T2*.
3.2 Acquisition The acquisition parameters are critical for the appearance of the MR
Parameters image, and the kind of contrast information that can be extracted
from it. To adequately set acquisition parameters in a useful fashion,
it helps to be aware of the relaxation behavior of the body parts that
are being imaged, and how these relate to the pulse sequence that is
being used. There are some basic principles required to design an
imaging experiment, and here we summarize a few.
In Figs. 4 and 5, the pulse sequence timing diagrams indicate
the highly relevant acquisition parameters TR and TE. These have
profound influence on the contrast that is achieved. TR dictates
how frequently the sequence is executed and has profound influ-
ence on how the T1 of each region affects the image. In fact, varying
the value of TR is how one achieves either a T1-weighted image, or
a proton density (PD) image. Assuming a small value of TE, a small
TR results in a T1-weighted image, and large TR yields a proton
density. When a T1-weighting sequence is acquired, those protons
with a long T1 become saturated and are weak, whereas those with a
short T1 recover much more quickly and appear bright. This results
in variations in intensity across the image based on the value of T1.
For example, after a CA has been administered, those regions that
have high concentrations will have a short T1, and hence will appear
bright.
In contrast, when a TR is chosen to be much longer than the
T1-value of the longest T1 region (typically 5T1), then the result is a
proton density image. Although the signal is strong, the distinction
between T1 is lost, and these images typically display poor contrast.
In addition, the large TR results in a scan time that is relatively long.
On the other hand, choosing a large value of TE attenuates the
signal of regions that have short T2. Therefore, with knowledge of
MRI Physics 19
how T2 varies across an image, one can choose a value of TE that is

long for many image regions. This results in variations in image
intensity based on T2 of a given region, and is called a T2-weighted
image. These images are very sensitive to variations in T2 and
provide excellent anatomical contrast throughout the body without
a need for exogenous CA.
In addition to these primary means of setting the contrast,
there are additional parameters that provide limits on these para-
meters. The matrix size dictates the number of direct dimension
(frequency encoding) points that must be acquired. How quickly
these are collected depends on this matrix size and the acquisition
bandwidth, which describes how fast these data points are sampled.
Before frequency encoding can occur, a gradient must turn on, and
so the rise time is typically a hardware characteristic that also
dictates acquisition readout times.
References
1. Ardenkjær-Larsen JH, Fridlund B, Gram A, 6. Lauterbur PC (1973) Image formation by
Hansson G, Hansson L, Lerche MH, induced local interactions. Examples employ-
Servin R, Thaning M, Golman K (2003) ing nuclear magnetic resonance. Nature
Increase in signal-to-noise ratio of >10,000 242:190–191
times in liquid-state NMR. Proc Natl Acad Sci 7. Pykett I, Rosen B (1983) Nuclear magnetic
100(18):10158–10163 resonance: in vivo proton chemical shift imag-
2. Golman K, Lerche M, Pehrson R, Ardenkjaer- ing. Work in progress. Radiology 149
Larsen JH (2006) Metabolic imaging by hyper- (1):197–201
polarized 13C magnetic resonance imaging for 8. Mansfield P (1984) Spatial mapping of the
in vivo tumor diagnosis. Cancer Res 66 chemical shift in NMR. Magn Reson Med 1
(22):10855–10860 (3):370–386
3. Rabi II, Zacharias JR, Millman S, Kusch P 9. Posse S, DeCarli C, Le Bihan D (1994) Three-
(1938) A new method of measuring nuclear dimensional echo-planar MR spectroscopic
magnetic moment. Phys Rev 53(4):318 imaging at short echo times in the human
4. Bloch F (1946) Nuclear induction. Phys Rev brain. Radiology 192(3):733–738
70(7–8):460 10. Hahn EL (1950) Spin echoes. Phys Rev 80
5. Damadian R (1971) Tumor detection by (4):580
nuclear magnetic resonance. Science 171
(3976):1151–1153
Chapter 2
Basic Pulse Sequences in Magnetic Resonance Imaging

Daniel Calle and Teresa Navarro
Abstract
Magnetic resonance images are obtained by a combination of different radiofrequency pulses and gradient
waveforms applied to the subject inside a magnetic field. There are multiple pulse sequences used in clinical
and preclinical studies adjusted to whatever physician or researches want to analyze, from basic anatomic
images to accurate diagnostic techniques as diffusion, perfusion, or functional imaging. In this chapter, we
present the most used radiofrequency pulse combinations of the two groups of sequences available in
magnetic resonance imaging: spin-echo and gradient-echo sequences.
Key words Magnetic resonance imaging, Pulse sequences, Spin echo, Gradient echo
1 Introduction
Magnetic resonance imaging (MRI) is a noninvasive imaging tech-

nique which gives important anatomical and physiological informa-
tion of the subject of study to the physician or researcher. Briefly,
this technique consists on modifying the magnetization vector
induced when the subject is inside a magnetic field and recording
the echo signal produced when the magnetization vector recovers
its original value. Several modifications can be done to the magne-
tization vector depending on what is under study. A pulse sequence
is a combination of radiofrequency (RF) pulses and gradient wave-
forms in order to modify the inclination of the magnetization
vector and to acquire the echo signal controlling the time between
the RF excitation pulses and the acquisition time of the echo. So,
the main components of a sequence are a RF excitation pulse to
create a transversal magnetization, gradients in order to encode the
transversal magnetization, and a signal reading to acquire the data
that will generate the magnetic resonance (MR) image.
Usually, sequences are represented schematically by a basic
pulse sequence diagram, which is a description of the signal and
gradient intensities in the time line. Figure 1 shows one of these
diagrams as an example. The top line shows the radiofrequency
21
22 Daniel Calle and Teresa Navarro
Fig. 1 Scheme of a basic pulse sequence diagram. First trace corresponds to the
radiofrequency pulse train (α) and the lower ones to the slice selection gradient
(second), the variable phase encoding gradient (third), the readout gradient of the
echo (forth) and the echo signal (fifth trace). TE is the time between the excitation
pulse (α) and the echo signal
pulse which modifies the magnetization vector an angle α. Pulse

height indicates the amplitude of the RF pulse and, therefore, the
rotation angle of the magnetization; the greater the pulse is, the
higher is the rotation angle. The other three lines are the gradient
magnetic fields (slice select, phase encode and readout) positioned
to form a three-dimensional coordinate system that encode the
transversal magnetization. The width of the boxes represents the
time that the gradient is on and the high is its intensity.
The first step on a basic pulse sequence is to turn on the RF
pulse. In this step we obtain the nutation of the longitudinal
magnetization an angle α. Simultaneously, the slice select gradient
must be on to select the slice of interest. Next, the phase encoding
gradient is turned on at the same time that the frequency encoding
gradient (also known as the readout gradient) to spatially localize
the NMR signal. This process is repeated for a series of phase
encoding steps to complete the entire image.
To find the best compromise between contrast, spatial resolu-
tion and speed is necessary to adjust some of the main parameters of
a pulse sequence, like: Repetition time (TR), is the time between
excitation RF pulse and the start of the next RF pulse, usually
measured in milliseconds; echo time (TE), is the time between
the excitation pulse and the receipt of the echo, also measured in
milliseconds; inversion time (TI), is the time between inversion
(when it is used) and excitation pulses, measured in milliseconds;
flip angle (α), nutation of the magnetization caused by the RF
excitation pulse, measured in degrees; field of view (FOV),
Basic MRI Sequences 23
measured in mm (or cm), matrix size (Mtx), is the phase steps plus
the points of the echo signal and corresponds with the number of
pixels of the image; number of averages (Nex), etc. The final
duration of a spin echo (SE) or gradient echo (GE) sequence, in
absence of any acceleration factor, corresponds with the following
equation:
Acquisition time ¼ T R∙Phase Steps ∙N ex ð1Þ
Specific Absorption Rate (SAR) has to be considered in the

pulse sequence design. SAR is a measurement of the energy that is
absorbed by biological tissue exposed to electromagnetic radiation,
which is generally associated with an increase in local temperature.
SAR, measured in watts per kilogram (W/kg) is defined as follows:
ð
σ ðr ÞjE ðr Þj2
SAR ¼ dr ð2Þ
ρðr Þ
where E is the electric field and σ and ρ are the electric conductivity
and density of the biological tissue. The magnetic field strength and
the frequency of the RF pulses influence directly in the SAR
increase. Clinical magnetic fields entails low SAR, but it is necessary
to do a good pulse sequence design if short TR or high flip angles
are used because of possible temperature increases which can even-
tually produce burns to the patients. In general, commercial MRI
scanners calculate the SAR for each sequence design preventing the
critical values, which depends of region of the body to be studied,
and are defined by international committees (FDA or IEC) which
established a maximum SAR value of 4 W/kg for the whole body.
Depending on the RF sequence parameters, the image
obtained can be weighted in T1 (longitudinal relaxation time), T2
(transverse relaxation time), proton density (PD), or T2*. T1 is
defined as the time that the longitudinal magnetization takes to
recover the original value. A sequence with short TR and short TE
results in a T1-weighted image. T2 represents the time that the
transversal magnetization takes to disappear. A T2-weighted image
is obtained with long TR and TE. A proton density image repre-
sents the water content (water hydrogens) and is obtained with
long TR and short TE. T2* is related to T2. It represents the loss of
phase coherence due to magnetic inhomogeneities and produces a
faster decay of T2. T2*-weighted images can be obtained using
Gradient Echo sequences. Tissues have different magnetic relaxa-
tion properties, so, depending on the image sequence, they will
have different intensities. As an example, in a T1-weighted image,
liquid (cerebrospinal fluid, edemas, etc.) is darker than other tissues
because it has not restrictions to recover the longitudinal magneti-
zation, it has a long T1, while in a T2-weighted image liquid is
brighter than other tissues, it has a long T2 (Fig. 2).
Fig. 2 T1- and T2-weighted images of mice brain. Note brighter zones (long T2) in the T2-weighted image
corresponding to cerebrospinal fluid (red arrows). The same area is dark in the T1-weighted image because its
T1 value is long
1.1 Contrast Agents The magnetic relaxation properties of a tissue (T1, T2, T2*, proton
density) can be altered through the use of pharmacochemical com-
pounds known as contrast agents (CAs). MR contrast agent acts
reducing the relaxation times of tissue water (T1 or T2). Two types
of MRI contrast agents have been developed in based on the effect
on the signal intensity. They can be positives (signal hyperintensity
or T1 enhancement) or negatives (signal hypointensity or T2
enhancement), also called positive and negative contrast agents,
respectively. CAs for MRI, involve mainly Gd(III) chelates, able
to enhance water relaxation in those tissues where they accumulate
[1]. Gd(III) is used because it has seven unpaired, slow relaxing
electrons, and depicts the largest magnetic moment among the rare
earth series. The ligands most frequently used are linear chelates
derived from diethylenetriaminepentaacetic acid (DTPA) or cyclic
chelates derived from the tetraaza macrocycles or cyclen derivatives
(DOTA). In all these cases, the ligand provides eight binding sites
anchoring the Gd(III), leaving free one the nine chelating sites of
the metal, for water contact. The contact between water molecules
in the tissue with the Gd(III), and the fast exchange of this water
molecule with the bulk solution, reduces very significantly the
relaxation times of the tissue, resulting in clearly enhanced image
intensity in those regions containing the chelate [1]. It is very useful
to distinguish between tissues with similar relaxation times like
intracranial tumors and cerebral parenchyma. Gadolinium chelates
do not cross the intact blood–brain barrier (BBB), but when a
tumor is growing in the brain, usually a disruption of the BBB
take place allowing the extravasation of the CA that causes a signal
increase in the MR image becoming easily to distinguish tumor
from edema or healthy tissue (Fig. 3).
The use of other lanthanides as Dy(III) may transform the same
chelates in T2 enhancing probes due to the inherent T2 relaxing
properties of this element [2].
Fig. 3 T1-weighted images of a glioblastoma rat model before (left) and after (right) the intravenous
administration of a contrast agent (gadolinium chelate). It can be appreciated the areas where the blood
brain barrier is disrupted in the tumor and the contrast agent is captured reducing the T1 value and therefore
increasing the intensity of the damaged brain region
MRI includes additionally a large variety of molecules able to

enhance image intensity using other mechanisms including mainly
magnetization transfer methods [3]. These agents are known as
diamagnetic Chemical Exchange Saturation Transfer (CEST) or
Paramagnetic Chemical Exchange Saturation Transfer (PARA-
CEST) agents. These molecules can be customized to reveal impor-
tant aspects of the lesions including properties of the
microenvironment as pH [4], monovalent or divalent ion concen-
tration [5], or temperature [6], among others. Superparamagnetic
iron oxide nanoparticles have been implemented more recently to
increase the relaxing capacity of the paramagnetic chelates
[7–9]. These particles contain a magnetite (Fe3O4) core, covered
most frequently by a dextran or lipid coat. The particles are
prepared by alkaline precipitation of mixtures of Fe3+ and Fe2+ in
the presence of stabilizing agents as dextran or oleic acid. These
depict enormous relaxivity values, as compared to the Gd(III)
chelates, allowing for a significant increase in the sensitivity for
MRI detection. This is due to the fact that the cooperative align-
ment of the magnetic moments from the iron ions in the super-
paramagnetic nanoparticles results in significantly larger magnetic
moments than the additive alignment of the paramagnetic Gd(III)
moments. Superparamagnetic behavior results mainly in T2
enhancement, in contrast with the paramagnetic T1 enhancement
of the Gd(III) chelates.
1.2 MRI Sequences There are many pulse sequence classifications in MRI. In this
chapter, we are going to talk about the two main groups of MRI
pulse sequences depending on how the echo signal is acquired: spin
echo and gradient echo sequences. Considering both of them,
there are numerous variations that include (or not) modifications
to accelerate the image acquisition and reduce the acquisition time.
Table 1
Equivalent MR manufacturer’s abbreviations with the corresponding type of sequence
Sequence Philips Siemens GE Hitachi Toshiba Bruker

Spin echo SE SE SE SE SE SE
(SE)
Fast SE Turbo SE Turbo SE Fast SE Fast SE Fast SE
Ultra-fast SSH_TSE SSTSE SS-FSE FSE-ADA (Super)
SE UFSE HASTE FASE DIET
Gradient FFE GRE GRE GE FE GE
echo (GE)
STIR STIR STIR STIR STIR STIR
STIR TSE Turbo STIR Fast STIR Fast STIR Fast STIR
FLAIR FLAIR FLAIR FLAIR FLAIR FLAIR
FLAIR TSE Turbo FLAIR Fast FLAIR Fast FLAIR Fast FLAIR
Hybrid echo GRASE TGSE Hybrid EPI
STIR STIR STIR STIR STIR STIR
STIR TSE Turbo STIR Fast STIR Fast STIR Fast STIR
Some of them are fast spin echo sequences, inversion recovery

sequences, ultrafast gradient echo, spoiled gradient echo, and also
sequences that are a hybrid between spin echo and gradient echo
(e.g., GRASE). It is important to keep in mind that manufactures
choose their own acronyms to denominate these sequences
(Table 1). There are around 100 different sequences and there is
no standard denomination for each sequence, so it is important to
consult the manufacturer manual.
2 Spin Echo Sequences
The spin echo pulse sequence is one of the most commonly used in
MRI and historically it was the first sequence to be used in this
imaging technique. It was described in 1950 by Erwin Hahn [10],
so a spin echo is sometimes also referred as a Hahn echo. A spin
echo consists in an echo formed after application of two pulses, an
excitation pulse and a refocusing pulse, separated by suitable time
delays. This sequence, like the Carr Purcell Meiboom Gill sequence
that will be described later in this section, was first designed for
NMR in the 1950s and adapted for MRI later in the 1970s.
In the pulse sequence diagram (Fig. 4), a 90 (π/2) excitation
RF pulse is applied followed by one (or more) 180 (π) refocusing
RF pulse to generate signal echoes when the spins are again in
phase. The 90 pulse rotates the nuclear magnetization down into
Fig. 4 Scheme of a spin echo (SE) sequence. Top trace corresponds to the
radiofrequency pulse train (π/2, π) and the lower ones correspond to the slice
selection gradient, the variable phase encoding gradient, the readout gradient of
the echo and the echo signal. TE is the time between the π/2 pulse and the echo
the transversal plane (XY plane). Then, the transverse nuclear mag-
netization begins to dephase due to the T2* dephasing (some spins
slow down and others speed up due to the field inhomogeneities).
A 180 pulse is then applied to rephase the nuclear magnetization
(partially) and, after a time delay equal to TE/2, produces a signal
called echo (the 180 pulse is exactly located in the half way
between the echo and the 90 pulse). Both pulses are applied in
conjunction with the slice selection gradient. The phase encoding
gradient is applied between the 90 and 180 pulses. Finally, the
frequency encoding gradient is applied during the readout of the
echo at time TE and the first raw of the K-space is filling. In order to
generate an image we have to repeat this process changing the
strength in the phase encoding gradient to fill another line in the
K-space. Normally, this process starts with large, negative phase
encoding gradient amplitudes going through zero phase encoding
gradient amplitude and finally to a large, positive phase encoding
gradient amplitude.
Other pulses besides a 90–180 combination can produce a
spin echo. Hahn employed a 90 –90 RF pulses in his original
description of spin echo. The 90–180 pulses are most commonly
used because with this combination it is obtained the higher echo
signal.
A spin echo pulse sequence produces an echo whose formation
can be compared with a horse race. First all horses are lined up at
the starting lines. Once the race starts (90 pulse), the faster horses
separate from the slower horses (dephasing). After a while (in the
spin-echo sequence, TE/2, and corresponding to the 180 pulse)
the horses are transposed (refocuses), and now the faster horses are
behind the slower ones. Nevertheless, because the last horses are
the faster than the first ones, at the end of the race all horses reach
the finishing line at the same time (echo signal).
The main disadvantage of conventional spin echo sequence is
the relatively long acquisition time, especially when long TR are
required. The Carr Purcell Meiboom Gill (CPMG) sequence, first
designed by Carr and Purcell in 1954 [11] and completed by
Meiboom and Gill in 1958 [12], is a variety of spin echo pulse
sequence which is particularly useful for measuring T2. It consists of
a 90 RF pulse followed by a train of evenly spaced 180 RF pulses
applied alternately along the XY plane to acquire several echo
signals. Each echo is encoded with a phase encoding gradient.
This pack of echoes is known as an echo train, and the total number
of 180 RF pulses and echoes is called an echo train length. This is a
fast SE pulse sequence and the acquisition time is greatly reduced
with respect to a conventional spin echo sequence.
Important parameters in spin echo sequences are the TR and
the TE, which can be varied to control the contrast and, therefore,
to obtain T1-weighted, T2-weighted, or proton density images
(Fig. 5). Typical values for a 7 T equipment are:
– A short TR and short TE will give a T1-weighted image (Fig. 5,
T1 SE, TE ¼ 12 ms, TR ¼ 500 ms), a TR < 700 ms increases the
effect of T1 on image contrast and a TE < 20 ms minimizes T2
contrast. In T1-weighted images short T1 tissues appears bright
and long T1 tissues appears dark (hyposignal).
– A long TR and a short TE will give a proton density image
(Fig. 5, PD SE, TE ¼ 12 ms, TR ¼ 2000 ms). The contrast
Fig. 5 T1, T2 and proton density (PD) weighted axial images of a rat brain using spin echo sequences (MSME
and RARE) in a 7 T spectrometer. The TE/TR can be varied to control contrast in spin echo imaging
obtained depends on the density of the hydrogen nuclei in the

different tissues.
– A long TR and long TE will give a T2-weighted image (Fig. 5, T2
FAST SE, TE ¼ 40 ms, TR ¼ 3000 ms). In T2-weighted images
long T2 tissues appear as a hypersignal, and short T2 tissues as a
hyposignal.
2.1 Inversion This sequence is a variant of a spin echo sequence with the advan-
Recovery Sequence tage to null or suppress the signal from any tissue based on its
longitudinal relaxation value. It can be chosen a specific time
(inversion time, TI) value to null the signal from fat, white matter,
gray matter, cerebral spinal fluid, etc. This value it is known as the
“null point” or “zero crossing point.”
First a 180 pulse is applied, which inverts 180 the longitudi-
nal magnetization vector to the –Z axis. After the 180 pulse, the
spinning nuclei begin to relax and the net magnetization vector
passes to the transverse plane (the null point) at a time TI. At this
moment, a combination of 90 and 180 pulses are applied to
obtain an image without the signal of a specific tissue. The time
between the 180 and the 90 pulse is known as time to inversion
(TI). Figure 6 shows the inversion recovery pulse sequence.
Fig. 6 Inversion-recovery (IR) pulse sequence diagram. The IR sequence is

similar to a spin echo sequence but with a 180 inverting pulse which inverts
the magnetization vector. During the TI the magnetization returns to the
equilibrium state and at certain point during recovery a 90 and 180 pulses
are applied and the signal is measured
However, there are several disadvantages for this sequence:

– The multiple 180 pulses result in tissue heating because the
specific absorption rate (SAR) is higher.
– Conventional inversion recovery (IR) sequences result in longer
scan times. It can be solved using fast inversion recovery (com-
bination of inversion recovery and turbo spin echo).
To calculate the TInull, there are two equations depending on
the sequence employed, for a conventional spin echo:
h i
TI NULL ¼ T 1 ∙ ln 2 ln 1 þ e TR=T 1 ð3Þ
for a fast spin echo
h i
TI NULL ¼ T 1 ∙ ln 2 ln 1 þ e ðTRTE last Þ=T 1 ð4Þ
where TElast in a FSE sequence is the time of the last echo. In both
equations when TR> > T1, the equations can be reduced to:
TI NULL ¼ T 1 ∙ln 2 ð5Þ

and finally,
TI NULL ¼ T 1 ∙0:963 ð6Þ
The contrast in the image will be depended on the length of the

TI, TR, and TE. If TE is short in regarding to the T2 value, the T2
weighting of the IR sequence is minimized. In IR sequences, the
TR does not control T1 weighting, it is only necessary to make sure
that it is long enough to acquire all slices, so only TI value controls
T1 weighting.
Some variations of the inversion recovery sequence are STIR
(Short TI Inversion Recovery) and FLAIR (Fluid Attenuated
Inversion Recovery). Short T1 tissue signal is suppressed with a
short inversion time. STIR sequence is, therefore, effective to fat
signal suppression due to the T1 relaxation rate of adipose tissue
which is shorter than the T1 of water. Using a FLAIR sequence, it
can be suppressed long T1 tissue signal of fluids as cerebrospinal
fluid or urine with a long inversion time.
2.1.1 STIR (Short TI The clinical evaluation, the differential diagnosis and the preclinical
Inversion Recovery) research, can be complicated by the amount of fat in the tissue. To
avoid this problem, it is essential to suppress the signal from adipose
tissue. The use of STIR sequences entails the null of the MRI signal
from fat in a T1-weighted image, and they are based on the differ-
ence in T1 between water and lipids hydrogens. Not only the fat
signal can be suppressed on a STIR sequence, other tissues with a
short T1 also will be nulled, such as proteinaceous fluid or
hemorrhage.
After the 180 inversion pulse, the longitudinal magnetization

(M0) vector from fat in the –Z axis will start recovering faster (TI
short) than M0 from water. When the null point of fat (the point at
which M0 has a null value) is reached, the 90 pulse will avoid the
fat but not the water signal of tissues. The following 180 pulse will
generate the echo.
2.1.2 FLAIR (Fluid In clinical environment, the use of spin echo pulse sequences in
Attenuated Inversion some anatomy regions such as spinal cord or periventricular lesions
Recovery) in which the bright pathologic areas are close to the equally bright
areas is problematic due to the high signal from cerebrospinal fluid
(CSF), which makes difficult to delineate the cord outline and
reduce the image quality due to the motion of CSF. Similar pro-
blems can take place in preclinical research. In this line, FLAIR
sequences can be used to null the fluid signal and the artifacts can
be reduced in the images to get a more accurate study.
FLAIR sequences use a long inversion time to null or reduce
the signals from fluids (like CSF, edema, urine. . .) and a long echo
time to produce a very heavy T2-weighted image. The problem is
that this can result in a very long acquisition time. Nowadays, the
combination of the FLAIR and the FSE avoid using this protocol
for routine imaging acquisition. In clinic, FLAIR pulse sequence is
very useful to detect lesions in the brain not easily distinguishable
within the subarachnoid space and brain parenchyma.
2.2 FAST (TURBO) In 1986 Henning et al. [13] developed the RARE sequence (Rapid
Spin Echo Acquisition with Relaxation Enhancement) to allow a more rapid
data acquisition with the same resolution than an image acquired
with a conventional spin echo. Nowadays it’s the most commonly
used sequence in MRI. The terms fast spin echo (FSE) or turbo
spin echo (TSE) mean the same, it depends on the manufacturer.
These are conventional spin echo sequences that use a series of
180 refocusing pulses after just one 90 excitation pulse to gener-
ate a train of echoes. The number of echoes acquired during the TR
is known as the echo train length or turbo factor (generally from
2 to 32 for routine imaging). Each echo has a different phase
encoding value (in a conventional multi-echo sequence all echoes
are collected with the same phase encoding) and this allows to
acquire multiple lines of K-space within a given repetition time.
The main advantage of this sequence is the speed. K-space is
filled faster and scan time is decreased. This technique allows
obtaining heavy T2-weighted imaging and a contrast to noise
ratio good due to the long TR (around twofold than in a conven-
tional spin echo). In this sequence, the acquisition time is equal to
the repetition time multiplied by the number of encoding steps in
the phase direction and divided by the turbo factor/echo train
length.
3 Gradient Echo Sequences
The two main pulse sequence groups in magnetic resonance imag-

ing are spin echo and gradient echo. In the early 1980s, the spin
echo sequence was the main magnetic resonance sequence in clini-
cal imaging. Later, around the middle of the decade, appears the
gradient echo sequence proposed by Mansfield and Maudsley [14],
being in 1990 an essential technique and a good alternative to spin
echo sequences.
A gradient echo pulse sequence (GE) is based on the applica-
tion of an excitation pulse between 10 and 90 degrees (flip angle).
Larger flip angles give more T1 weighting to the images while
smaller flip angle give more T2* weighting to the images. The low
flip angle excitation produces a faster recovery of longitudinal
magnetization which allows faster image acquisition, therefore the
TR and TE can be reduced. So, the first difference with spin echo
sequence is that the excitation pulse can be lower than 90 . The
second difference is the utilization of bipolar readout gradient
instead of a 180 pulse that generates transverse magnetization to
obtain a gradient echo (GE).
These sequences are basic to see T2* relaxation that is caused by
a combination of spin-spin relaxation and magnetic field inhomo-
geneities, not detected in spin echo sequences because are elimi-
nated by the 180 pulse. In the case of GE sequences the transverse
relaxation is a combination of the “true” T2 relaxation and the
relaxation caused by magnetic field inhomogeneities (T2* relaxa-
tion). This value is shorter than T2 and the following equation can
be show the relationship between them:
1=T ∗
2 ¼ 1=T 2 þ γ∙ΔB inhom ð7Þ
where γ is the gyromagnetic ratio y ΔBinhom the magnetic field

inhomogeneity across a voxel.
In the pulse sequence diagram (Fig. 7) firstly it is applied the RF
pulse with a characteristic flip angle (α) and simultaneously the slice
select gradient. After the application of the excitation pulse is the
simultaneous application of the phase encode and the read dephas-
ing gradients, followed by the read rephasing gradient.
The main disadvantages of this sequence are the sensitive to B0
inhomogeneities and to magnetic susceptibility effects because the
gradient does not compensate for magnetic field inhomogenities.
For that reason, gradient echo sequence with long TE are T2*
weighted.
The contrast in the image can be varied with changes in some
parameters of the sequence. In gradient echo sequences, TE is
related to T2* weighting; when TE is large, this term dominates
while T2* contribution is lower when TE is small. The flip angle
Fig. 7 Gradient echo pulse (GE) sequence diagram. In this sequence, the RF pulse
angle is variable depending on the type of image weighting to be obtained. Here,
the echo signal is acquired using a dephasing readout gradient
Fig. 8 T1, T2*, and proton density (PD) weighted axial images of a rat brain using gradient echo sequences
(FLASH and GEFC) in a 7 T spectrometer. Varying the flip angle and TE/TR different MRI weighting can be
obtained
and TR are related to T1 weighting; a small flip angle minimizes T1

weighting and T2* effects predominate while a short TR increases
T1 weighting (long TR values minimized T1 effects in contrast).
This sequence can be used, therefore, to acquire T2*, T1, and
proton density weighting (Fig. 8). Typical values for a 7 T magnetic
field can be seen in Table 2:
The choice of flip angle value in GE sequences is important in
T1-weighted imaging and it depends on the T1 value of the tissue. A
Table 2
Typical values of Flip Angle, TE, and TR for T1, T2*, and proton density weighting using a GE sequence
T1 weighting T2* weighting Proton density weighting

Flip angle 60 10 5
TE (ms) 3 5 3
TR (ms) 65 500 500
short T1 value is related to a larger flip angle. Therefore, we can

obtain the optimum tissue contrast if we use the optimum flip
angle. Richard R. Ernst [15] described this relationship and for
that the flip angle is known as the Ernst Angle:
αErnst ¼ arccose TR=T 1 ð8Þ
There are two types of gradient echo sequences: incoherent

gradient echo or gradient spoiled (spoiled residual transverse mag-
netization) and coherent gradient echo (refocused transverse mag-
netization). In the incoherent gradient echo sequence, the
transverse magnetization is eradicated by a magnetic field gradient
or a spoiler RF pulse. This gradient spoiling occurs after each echo
by using strong gradients in the slice select direction. Some of the
commercial names for this sequence are FLASH, SPGR, or
RF-FAST. The coherent gradient echo incorporates a residual
transverse magnetization directly into the longitudinal steady
state, using a refocusing gradient in the phase encoding direction
to rephase the T2* magnetization while it is being dephased.
Gradient echo sequences are very sensitive to local magnetic
field inhomogeneity (local variation in B0). These inhomogeneities
appear at the interface between entities with different magnetic
susceptibilities (tissue and air, bone and air). This feature of gradi-
ent echo sequence is used to different MRI applications such as
susceptibility-weighted imaging, perfusion MR imaging, and func-
tional MR imaging, and can be used, for example, for detecting
hemorrhage (iron in hemoglobin produces local magnetic field), in
blood oxygenation level-dependent (BOLD) imaging (the amount
of deoxyhemoglobin in the blood changes the image intensity),
calcification or iron deposition.
4 Preclinical Applications
MRI is the best choice for the study of diseases located in the
structures of the nervous and musculoskeletal systems because
provides excellent contrast between soft tissues and has a high
spatial resolution. There are specific protocols to obtain images
from different regions of the subject. The two main Magnetic

Resonance Imaging sequences described in this chapter are used
in different preclinical applications. Nowadays, gradient echoes
sequences are the fast MR imaging techniques predominantly
used for signal generation.
In the case of spin echo sequence, the fast spin echo sequence
has practically substituted the conventional spin echo. Both
sequences have similar quality, good signal intensity from gray-
white matter, fat, CSE, vascular. . . although the examination time
is dramatically reduced in FSE. T1- and T2-weighted images are
useful to study anatomy. Tissues with some diseases could appear
edematous and/or vascularized, and due to the higher water con-
tent in these process, there is a strong signal on T2-weighted images
which are used to differentiate normal tissue from the pathological
ones. The inflammatory process in tissue lesion causes changes in
the relaxation times T1 and T2 because the swelling causes an
increase of water in the tissues, and due to this it increases the
signal intensity on T2-weighted images. The same effect can be
observed with T1-weighted images but in this case it is necessary
to apply a contrast agent to reduce the relaxation time. The fast spin
echo sequence has been widely used in anatomic structure studies
of central nervous, pelvis, and musculoskeletal systems. FLAIR
sequence can be used in the study of central nervous system dis-
orders, such as lacunar infarction, multiple sclerosis plaques, sub-
arachnoid hemorrhage, head trauma, meningitis, and other
leptomeningeal diseases. STIR sequences are commonly used in
the study of musculoskeletal diseases (multiple sclerosis of the
cervical spinal cord, in bone marrow abnormalities of foot and
ankle, spinal cord injury. . .).
Gradient echo sequences are highly used for many magnetic
resonance applications, such as susceptibility-weighted imaging
(SW), perfusion MR imaging, and functional MR imaging. These
sequences are a useful tool in preclinical studies. In contrast to spin
echo sequence which acquires a spin echo signal to obtain T2-
weighted images, with echo gradient sequence a gradient echo
signal is acquired obtaining a T2*-weighted image which can detect
the smallest changes in uniformity in the magnetic field and can
improve the rate of small lesion detection (paramagnetic substance
deposition) that can be used in the study of cerebral microbleeds
and iron deposition. Some diseases with iron deposition are chronic
liver disease [16, 17], chronic anemia [18], Alzheimer disease [19],
and atherosclerosis [20]. By the other hand the cerebral micro-
bleeds lesions can be visually confirmed with gradient echo T2*-
weighted imaging due to the paramagnetic effects of the deoxyhe-
moglobin, the magnetic susceptibility effect results in a signal loss
and can be detected with this technique [21]. The SW imaging is
based on the magnetic susceptibility differences of the blood, and
of iron and calcification in tissues and lesions. Neurodegenerative
disorders with increased iron deposition such as Parkinson disease,

Hungtinton disease, or Alzheimer disease are depicted with SW
imaging better than conventional GE sequence. Evaluation of
stroke, trauma, vasculitis, epilepsy and characterization of brain
tumor are some other clinical applications of SW imaging.
Perfusion MR imaging is based on the measurement of signal
intensity decreases during the passage of a magnetic resonance
contrast agent through the microvasculature.
Functional MR imaging uses the hemodynamic and metabolic
changes associated with brain functions, which affect the deoxyhe-
moglobin contents in the tissue and this generates a contrast that
can be detected using gradient echo sequences. Functional mag-
netic resonance imaging is used for studying regional brain function
that is associated with sensory, motor, and cognitive tasks in healthy
individuals and to understand neurobehavioral disorders (Alzhei-
mer’s disease, epilepsy, brain tumors, stroke, traumatic brain injury,
multiple sclerosis. . .).
References
1. Caravan P, Ellison JJ, McMurry TJ, Lauffer RB 8. Bonnemain B (1998) Superparamagnetic

(1999) Gadolinium (III) chelates as MRI con- agents in magnetic resonance imaging: physi-
trast agents: structure, dynamics, and applica- cochemical characteristics and clinical applica-
tions. Chem Rev 99(9):2293–2352 tions. A review. J Drug Target 6(3):167–174.
2. Aime S, Barge A, Delli Castelli D, Fedeli F, https://doi.org/10.3109/
Mortillaro A, Nielsen FU, Terreno E (2002) 10611869808997890
Paramagnetic lanthanide (III) complexes as 9. Liu F, Laurent S, Fattahi H, Vander Elst L,
pH-sensitive chemical exchange saturation Muller RN (2011) Superparamagnetic nano-
transfer (CEST) contrast agents for MRI appli- systems based on iron oxide nanoparticles for
cations. Magn Reson Med 47(4):639–648 biomedical imaging. Nanomedicine (Lond) 6
3. Hancu I, Dixon WT, Woods M, Vinogradov E, (3):519–528. https://doi.org/10.2217/nnm.
Sherry AD, Lenkinski RE (2010) CEST and 11.16
PARACEST MR contrast agents. Acta Radiol 10. Hahn EL (1950) Spin echoes. Phys Rev 80
51(8):910–923. https://doi.org/10.3109/ (4):580
02841851.2010.502126 11. Carr HY, Purcell EM (1954) Effects of diffu-
4. Sun PZ, Sorensen AG (2008) Imaging pH sion on free precession in nuclear magnetic
using the chemical exchange saturation transfer resonance experiments. Phys Rev 94(3):630
(CEST) MRI: correction of concomitant RF 12. Meiboom S, Gill D (1958) Modified spin-echo
irradiation effects to quantify CEST MRI for method for measuring nuclear relaxation times.
chemical exchange rate and pH. Magn Reson Rev Sci Instrum 29(8):688–691
Med 60(2):390–397 13. Hennig J, Nauerth A, Friedburg H (1986)
5. McRae R, Bagchi P, Sumalekshmy S, Fahrni CJ RARE imaging: a fast imaging method for clin-
(2009) In situ imaging of metals in cells and ical MR. Magn Reson Med 3(6):823–833
tissues. Chem Rev 109(10):4780–4827. 14. Mansfield P, Maudsley A (1976) Planar and
https://doi.org/10.1021/cr900223a line-scan spin imaging by NMR. In: Proc.
6. Zhang S, Malloy CR, Sherry AD (2005) MRI XIXth Congress Amperen, Heidelberg, pp
thermometry based on PARACEST agents. J 247–52
Am Chem Soc 127(50):17572–17573. 15. Ernst R, Anderson W (1966) Application of
https://doi.org/10.1021/ja053799t Fourier transform spectroscopy to magnetic
7. Na HB, Song IC, Hyeon T (2009) Inorganic resonance. Rev Sci Instrum 37(1):93–102
nanoparticles for MRI contrast agents. Adv
Mater 21(21):2133–2148
16. Henninger B, Kremser C, Rauch S, Eder R, 19. Qin Y, Zhu W, Zhan C, Zhao L, Wang J,
Judmaier W, Zoller H, Michaely H, Schocke Tian Q, Wang W (2011) Investigation on pos-
M (2013) Evaluation of liver fat in the presence itive correlation of increased brain iron deposi-
of iron with MRI using T2* correction: a tion with cognitive impairment in Alzheimer
clinical approach. Eur Radiol 23 disease by using quantitative MR R2’ mapping.
(6):1643–1649. https://doi.org/10.1007/ J Huazhong Univ Sci Technol Med Sci 31
s00330-012-2745-2 (4):578–585. https://doi.org/10.1007/
17. Kolnagou A, Natsiopoulos K, Kleanthous M, s11596-011-0493-1
Ioannou A, Kontoghiorghes GJ (2013) Liver 20. Mihai G, He X, Zhang X, McCarthy B, Tran T,
iron and serum ferritin levels are misleading for Pennell M, Blank J, Simonetti OP, Jackson RD,
estimating cardiac, pancreatic, splenic and total Raman SV (2011) Design and rationale for the
body iron load in thalassemia patients: factors study of changes in iron and atherosclerosis risk
influencing the heterogenic distribution of in Perimenopause. J Clin Exp Cardiol 2:152.
excess storage iron in organs as identified by https://doi.org/10.4172/2155-9880.
MRI T2*. Toxicol Mech Methods 23 1000152
(1):48–56. https://doi.org/10.3109/ 21. Greenberg SM, Vernooij MW, Cordonnier C,
15376516.2012.727198 Viswanathan A, Salman RA-S, Warach S, Lau-
18. Barzin M, Kowsarian M, Akhlaghpoor S, ner LJ, Van Buchem MA, Breteler MM, Group
Jalalian R, Taremi M (2012) Correlation of MS (2009) Cerebral microbleeds: a guide to
cardiac MRI T2* with echocardiography in detection and interpretation. Lancet Neurol 8
thalassemia major. Eur Rev Med Pharmacol (2):165–174
Sci 16(2):254–260
Part II
Perfusion, Diffusion and Functional MRI

Chapter 3
Dynamic Susceptibility Contrast MRI in Small Animals

Pilar López-Larrubia
Abstract
The use of magnetic resonance imaging (MRI) for studying the cerebral perfusion mechanisms is well
proved and contrasted in the clinical and research setups. This methodology is a promising tool in assessing
numerous brain diseases like intracranial tumors, neurodegeneration processes, mental disorders, injuries
and so on. In the preclinical environment, perfusion MRI offers a powerful resource for characterizing
pathological models and specially identifying biomarkers to monitor the illness and validate the efficacy of
therapeutical approaches. This chapter presents the theoretical bases and experimental protocols of dynamic
susceptibility contrast MRI acquisitions for developing perfusion MRI studies in small animals.
Key words Preclinical MRI, Brain perfusion, Dynamic susceptibility contrast, Bolus tracking, Cere-
bral blood flow, Cerebral blood volume, Mean transit time, Animal model
1 Introduction
Magnetic resonance imaging (MRI) has been proven to be one of

the most powerful tools for studying the coherent (perfusion) and
incoherent (diffusion) motion of water molecules in tissues
[1–3]. In fact, MRI approaches that take advantage of these move-
ments to improve the information of images have experienced a
huge development in the last years [4]. The basis of the determina-
tion of the physiological flow are known from more than a century
ago, and the most of the magnetic resonance methodologies for
measuring perfusion are based on those principles [5]. These MRI
methodologies are generically named Perfusion-Weighted Imaging
(PWI) and have spectacularly evolved reaching nowadays a great
impact both in the clinical and preclinical imaging, especially in the
neuroimaging field [6]. In fact, clinical perfusion MRI measure-
ments are currently recognized as potent approaches to assess the
microvascularization in the healthy and diseased brain, and also to
monitor the responses to therapies through the evaluation of
hemodynamic parameters.
41
42 Pilar López-Larrubia
The perfusion processes in an organism are associated to a

coherent displacement of water molecules describing the blood
flow that feed a volume element in an organ or a tissue. This
concept must be distinguished from the venous or arterial flow,
and therefore from the angiographic methodologies [7], although
like those, perfusion also can be analyzed and measured with dedi-
cated MRI techniques with an adequate temporal and spatial
resolution.
The ability of obtaining relevant and precise information from
these imaging studies also depends on the development of adequate
mathematical models and signal fitting algorithms to those models.
1.1 Techniques MRI methodologies to measure perfusion can be divided in two

and Methods broad classes: those monitoring tissue signal changes using an
exogenous relaxation contrast agent (CA) and those that use
endogenous contrast. The latter is based on the changes in tissues
caused by the level of oxygenation in blood that are detected in
T2*-weighted images—this contrast is called blood oxygenation
level dependent (BOLD) [8]. Regarding on the former techniques,
exogenous contrast media include: (1) ferromagnetic and super-
paramagnetic agents that shorten the T2 and T2* values from water
molecules around them; and (2) paramagnetic compounds that
shorten the T1 values of hydrogens in close contact with them.
1.2 “Bolus Tracking” The use of exogenous contrast agents allows obtaining unique
Perfusion MRI information of tissular perfusion and so, improving the edge of
regions with a deficient blood flow in diseased conditions. Com-
pounds that include gadolinium (Gd), dysprosium (Dy), or iron
(Fe) provoke a regional decrease of the signal intensity in T2- and
T2*-weighted images. This phenomenon can be exploited to indi-
rectly valuate the perfusion profiles in tissues and calculate hemo-
dynamic parameters to be used as biomarkers of diseases. The MRI
methodology to do that is generally called bolus tracking technique
[9], or hemodynamic-weighted magnetic resonance imaging, and
takes part of the dynamic susceptibility contrast techniques
(DSC) [10].
To map perfusion entails the ability of visualizing alterations in
the MRI signal during the transit of a bolus of CA—quickly
injected—with ultra-fast image acquisition sensitive to T2*. The
method relies in the ability of detecting magnetic susceptibility
changes in the microvasculature employing very rapid imaging to
capture the circulation of a bolus of CA rapidly injected. In the
brain, the first-pass extraction of the agent is zero when the blood–-
brain barrier (BBB) is intact, and the intravascular compartmentali-
zation of the contrast creates strong, microscopic susceptibility
gradients. These microscopic gradients cause dephasing of spins
as the CA diffuse among them, and in terms of signal loss, all
water molecules in the proximity of the vessels will be affected
Dynamic Susceptibility Contrast MRI 43
Fig. 1 Representative scheme of a bolus tracking perfusion MRI study. A pixel-by-pixel representation of the
signal intensity with time, assesses the passage of the contrast agent through the cerebral microvasculature.
Basing on the relationship between T2* (or relaxation R2*) and the CA concentration (Ct), the data can be
converted into a concentration-time curve that fitted to a gamma function allows the measurement of
hemodynamic parameters like CBV (that corresponds to the area under the curve) at any pixel or region of
interest
because the magnetic susceptibility effects spread beyond the

immediate vicinity of the agent itself. So, pulse sequences without
full refocusing of static field inhomogeneities—as gradient-echo
(GE) are—will suffer a general signal loss due to the presence of
microscopic field disturbances in the microvasculature. Figure 1
depicts a representative scheme of the process.
In the setup showed, the bolus of CA can be tracked by the
acquisition of a multislice series of T2*-sensitive images. In prac-
tice, PWI based on the tracking of a bolus of contrast medium
requires the use of ultra-fast acquisition sequences—echo-planar
imaging (EPI)—able to obtain images in milliseconds to achieve a
good temporal resolution—approx. 6–12 slices in 1–2 s—and at the
same time with an adequate spatial resolution (Fig. 2).
From images, a pixel-by-pixel representation of the relative
signal intensity with time shows the effect of the bolus transit
through every slice. It exists an approximate linear relationship
between tissue contrast agent concentration and change in T2*
relaxation rate, and from the signal-time course the
concentration-time course of the contrast medium can be calcu-
lated. Bolus behavior is then fitted to a gamma-variate function to
correct for tracer recirculation. Assuming uniform arterial
Fig. 2 EPI gradient-echo sequence diagram. An RF exciting pulse is applied in

presence of a slice selection gradient and multiple echoes of different phase
steps are acquired using rephasing gradients. This is accomplished by rapidly
reversing the frequency-encoding gradient
concentration profiles in all arterial inputs, relative CBV measure-

ments are determined by integration of the area under the
concentration-time curve that represents the volume of the effect.
Other parameters to be measured are the time to peak (TTP)—time
that the CA takes since the intravenous injection to the maximum
of the curve—and mean transit time (MTT). On these grounds,
hemodynamic parameters can be calculated by employing the
appropriate equations and mathematical models [9, 11]. However,
to perform an absolute quantification of data, the temporal data of
arterial flow has to be known to perform a deconvolution of the
tissue concentration time curves. For that, it is mandatory to obtain
an arterial input function (AIF) by measuring the signal changes
around or inside of a big vessel [12].
1.3 Cerebral Several parametric images of hemodynamic parameters can be

Measurable Perfusion obtained with bolus tracking MRI [11]:
Parameters
1. Cerebral blood volume (CBV)
2. Cerebral blow flood (CBF)
3. Mean transit time (MTT)
The regional CBV (rCBV) is defined as the blood volume in a
voxel of cerebral tissue divided by the mass of that voxel. In general,
this term only incudes microcirculation—arterioles, capillaries, and
venules—and is expressed in units of mL/100 g (or μL/g). The
CBV is frequently considered as a fraction of a volume and
represented as percentage. The regional CBF (rCBF) is defined as

the net blood flow through the voxel divided by the mass of the
voxel and is commonly quantified with the unit mL/100 g/
s (or μL/g/s). Lastly, the regional MTT (rMTT) describes the
average amount of time that it takes any water molecule or particle
of contrast agent to pass through the voxel vasculature and is
generally expressed in seconds.
While the measurement of CBF and MTT necessarily relies in
the sensitive of the technique to the movement, there are other
options for determining CBV.
1.4 Applications Perfusion-weighted MRI has numerous and important applications

not only in the clinical but also in the research field [20]. DSC
imaging provides unique and valuable information relative to the
cerebral hemodynamic and it is also an inestimable tool in experi-
mental preclinical studies. These approaches offer more and more
exciting possibilities in the characterization and evaluation of the
healthy and damaged cerebral function. PWI has also prominent
applications in the validation of therapies of pathologies like neu-
rodegeneration, dementia, tumors [13, 14], brain injuries [15],
glial lesions, and any other disorder that entails alterations in the
vascular flow.
One of the main applications found with bolus tracking meth-
odologies has been the study of the cerebral ischemia, particularly
in the acute stroke context. The concept of perfusion-diffusion
mismatch (PDM)—areas with abnormal perfusion values but with
diffusion characteristics apparently normal—has gained a great
interest in the last years. In fact, the spreading of the initial lesion
detected by diffusion MRI takes place in such a way that the final
infarcted area includes regions of tissue that were located in this
PDM zone during the hyperacute phase of ischemic stroke [16].
2 Materials
2.1 MRI Equipment 1. MR high-field horizontal magnet.

2. Gradient coils >200 mT/m with ramp times <200 μs.
3. RF coils. The two basic MRI coil setup that can be used are:
(a) Transceiver coil setup. The excitation and reception of the
MR signal are performed with the same volume coil of
around 35–40 mm inner diameter for rats and 25–30
inner diameter for mice.
(b) Active decoupled transmit and receive coil setup. Excita-
tion and reception of the signal are typically made by a
volume transmit coil and a brain surface coil (rat and/or
mouse dedicated). In that case, coils must be actively
detuned from each other.
2.2 Software 1. Acquisition: Paravision 5.1 (Bruker Biospin) running in a con-

sole Bruker AVANCE III operated on a Linux platform.
2. Post-processing: Software to perform a pixel-by-pixel fitting of
the MRI signal to the appropriate equations in each case yield-
ing the corresponding parametric images CBF, CBV, and MTT
(see Subheading 3.4). The mathematical approach is based on
the fact that changes in transversal relaxation are proportional
at the concentration of contrast medium at any point according
to Eq. 1:

S ðt Þ
ΔR∗
2 ðt Þ ¼ k ln ð1Þ
S 0 ðt Þ
where S(t) represents the dynamics of the signal intensity, S0(t)

denotes the pre-contrast basal levels, k is a proportionality
constant, and ΔR∗ 2 is a magnitude proportional to the concen-
tration of the contrast agent in the tissue (corresponds to the
increment in 1/T2*). The parametric images are calculated
from the expressions:
Z
CBV ¼ ΔR∗ 2 ðτÞdτ ð2Þ
R
τ ΔR∗2 ðτ Þdτ
MTT ¼ R ∗ ð3Þ
ΔR2 ðτÞdτ
CBV
CBF ¼ ð4Þ
MTT
Later on, sections of interest are manually selected in the maps
and data statistically analyzed, with image analysis free software like
Image J and data statistically analyzed with adequate programs (see
Subheading 3.4).
2.3 Other Equipment, 1. Monitoring system. Physiological parameters are controlled

Devices, and Materials during the whole study by using a small animal monitoring
and gating system with a rectal sensor and a pneumatic pillow.
This equipment continuously reports data to monitor the state
of the animal inside the magnet and control the level of
anesthesia.
2. Anesthesia equipment. Animals are anesthetized during the
MRI acquisitions. The combination of a gas carrier and the
liquid anesthetic has been used in combination with the suit-
able equipment that includes vaporizer, flowmeter, circulating
tubes, nose cap or mask, and a filter to optimize halogenated
gas absorption. Also, an appropriate poly-methyl methacrylate
(PMM) chamber is necessary to put the awake animal and
initiate the anesthesia.
3. Animal holders. Suitable animal beds of PMM (or any other

MR compatible material) are used to slide the rat or mouse
inside the magnet and localize the region of interest (ROI) in
the magnet isocenter.
4. Warm blanket (or some other heating system).
5. Infusion pump to administrate the contrast agent at a fixed
infusion rate. Also are necessary: syringe, catheter needle/nee-
dle, tubing, physiological serum, and heparin.
6. Contrast agent: Usually, gadolinium (Gd) or iron (Fe) contain-
ing compound (chelates, complexone, nanoparticle, etc.). Also
commercial and homemade contrast media can be employed to
assess brain perfusion itself or to validate the compound.
7. Eye ointment.
The protocols described in this chapter are tailored to a high-
field preclinical MRI system with the specifications indicated in
Table 1.
Table 1
Preclinical MRI system and associated equipment/devices for developing the protocols described in
this chapter
MRI system Bruker Biospect® 7 T/16 cm bore

Gradient system Shielded, 90 mm bore
360 mT/m maximum amplitude
1
Rat-head transceiver coil H circular polarized transmit/receive birdcage volume
coil, 40 mm inner diameter
1
Mouse-head transceiver coil H circular polarized transmit/receive birdcage volume
coil, 23 mm inner diameter
1
Transmitter coil H circular polarized transmit-only volume coil, 72 mm
inner diameter
1
Rat head surface coil H circular polarized with integrated combiner and preamplifier
receive-only coil, 40 mm inner diameter
1
Mouse head surface coil H circular polarized with integrated combiner and preamplifier
receive-only coil, 23 mm inner diameter
Physiological monitoring Small Animal Monitoring and Gating System (Model 1025,
SA Instruments, Inc. NY)
Anesthesia Isoflurane (IsoFlo, Esteve labs, Barcelona, Spain) in O2 or air
(Standard calibrated vaporizer)
Halogenated gas filter Activated charcoal canister packed (Harvard Apparatus Anaesthetic,
Holliston, MA, USA)
Thermal blanket Warmed circulating water blanket (Gaymar T/PUMP TP500
Heat Therapy System)
Contrast agent Magnevist® (Bayer)
Resovist (Bayer)
Dy-DTPA (homemade)
3 Methods
3.1 Vessel For these studies, a vessel of the animal has to be cannulated for a
Cannulation quick administration of the contrast agent during acquisition.
Although an artery could be the best choice because of the faster
reaching of the brain, especially carotid arteries, in practice, due to
the higher difficulty of the procedure, a tail vein is usually the
selected one (Fig. 3, up). Besides, the cannulation of a tail vessel
is less invasive for the animals allowing the surveillance of them after
the study. Intravenous cannulation of a rat lateral tail vein can be a
challenging procedure, even more in a mouse [17]. The method
entails:
1. Place the animal in an anesthetic box induction with isoflurane
(or any other inhalational anesthesia) 3–4% in oxygen (or any
other gas carrier) at a flow of 1 L/min and then place it in an
appropriate surface with a mask supplying a lower amount of
anesthesia (1–1.5%).
Fig. 3 Cannulation of a tail vein in a rat (or mouse). Up: Schematic picture of a transversal section of the tail
showing the vessels to cannulate; down: cannulation of the dorsal venin in a rat with a catheter
2. Select the catheter according to the vessel to cannulate. Maybe

the best option is to make the catheter in the laboratory
employing a needle and polyethylene (PE) tubing of the appro-
priate length and diameter. They are good choices: PE10 tub-
ing (0.28 0.61 mm) with a 30G needle for mice and PE20
tubing (0.38 1.1 mm) with a 28G needle for rats, although
any other tailored combination can work properly. To prepare
the catheter, the needle has to be detached from the hub and
put into the tubing through the blunt end.
3. Preload the catheter with 1% heparinized saline and position
the anesthetized animal facilitating to match the angle of the
needle with the angle of the vein, when the needle is almost
parallel to the vein (Fig. 3, down). It is also useful to hold the
needle with straight or curved hemostat forceps to better
introduce it through the skin and then advanced for 3–5 mm.
To verify that the cannulation is correct and the vein is not
perforated, inject a small saline flush into the vein and you will
see as the vein clears as the liquid is passing. Put the preloaded
syringe with serum in the extreme of the cannula.
4. Secure the catheter in place with tape. Because the catheter
insertion may not be successful, the first attempt should be
made at the extreme of the tail, allowing subsequent tries on
upper parts of the vessel (see Note 1).
3.2 Animal Before starting MRI studies, make sure to observe all the healthy
Positioning and safety rules in a high magnetic field room:
in the Scanner
1. Not approximate to the magnet with any metal, either in the
animal or in the operator.
2. Check that the oxygen level sensors are operative.
3. Verify that vaporizer and oxygen bottle are full enough and the
halogenated gas filter is working properly.
After that:
1. Place the animal in the appropriate bed/holder—rat or mouse
adapted—usually in the prone position, over a heated blanket
(or any other system to maintain the body temperature at
around 37 C). The bed has to have an anesthesia inlet and a
1 L/min flow of 1–1.5% isoflurane/oxygen maintained during
the whole study. Make sure that anesthesia is flowing without
impediments. A thermometer sensor has to be inserted in the
rectum, to control the body temperature, and breathing sen-
sors allocated under the animal chest to monitor the animal
status and modulate the plane of anesthesia. Respiration rate
should be between 40 and 60 rpm (respirations per minute) to
minimize motion artifacts in images (see Note 2). The animal
head has to be immobilized during acquisition because
movements will induce processing error in the calculated para-

metric images. To that, fixed it with ears and tooth bars and/or
placing tape over the head. Put eye ointment on the eyes of the
animal to avoid damaging the cornea.
2. Select the adequate commercial or custom RF coil configura-
tion—volume receive/transmit, array, surface, etc.—and place
them as close as possible of the region of interest in the animal.
3. Introduce the animal in the scanner—either manually or using
an automatic positioning system—matching the RF coil with
the isocenter of the magnet.
3.3 MRI Acquisition 1. Identify the animal and the study in the operating software of
Protocols the system. Select the protocols to be acquired starting with a
rapid acquisition to check the correct positioning of the brain
in the magnet. Prior the first scan, an automatic or manual
adjustment of some parameters has to be achieved:
(a) Tuning and matching the coil—when it is possible—to
adjust its frequency of resonance and impedance accord-
ing to the animal.
(b) Shimming the magnet to improve its homogeneity.
(c) Determining the RF power for the 90 excitation pulse.
(d) Adjusting the basic frequency of the system.
(e) Calculating the optimal receiver gain.
2. Acquire structural images to image the brain anatomy and use
them as reference. Normally a T2W sequence is employed for
obtaining anatomical information. If a T2 study is destined to
cover the whole brain, do not forget to acquire a new one with
the same geometry—number of slices, slice thickness, and
separation—as the DSC acquisition. These T2W images will
be used to superimpose the parametric perfusion maps and
better identify the cerebral regions. (Acquisition parameters
are shown in Table 2.)
3. DSC studies are performed using ultra-fast gradient-echo
methods (EPI-T2*). These sequences that employ a single-
shot echo-planar acquisition module to get the necessary tem-
poral resolution, are highly sensitive to magnetic field inhomo-
geneities. So, a very good shimming to improve the magnetic
field homogeneities is a prerequisite for successful. There are
several automatic methods to achieve that:
(a) Optimizing the shimming of the whole sample covered by
the coil—it is a global approach—by varying the first and
second shim coil values and making adjustment according
to the signal response.
Table 2
Normal acquisition parameters used in the sequences employed to acquire a perfusion DSC MRI
experiment
Parameter T2W DSC CE-T1W

Method RARE EPI MSME
Sequence data Spin-echo acquisition Gradient-echo single-shot Spin-echo acquisition
echo-planar acquisition
TR/TE (ms) 2500–3000/ 250–400/ 250–400/
50–120 7–14 10–20
Slice thickness (mm) 1.0–1.5 1.0–1.5 1.0–1.5
Slice separation (mm) 0–0.5 0–0.5 0–0.5
a
FOV (cm) 2.0–4.0 2.0–4.0 2.0–4.0
Matrix b
256 256 64 64 256 256
128 128
Slices 1–6 1–6 1–6
Acquisition time 2–5 min 45–95 s 2–5 min
Acceleration factors RARE factor: 8 Parallel accelerationc
Partial FT
Zero filld
Others Number of repetitions: 150–200
a
FOV: 2.0 for mice, 4.0 for rats
b
See Note 3
c
When an RF array coil is employed
d
In Bruker systems works better zero fill than partial FT acceleration
(b) Employing a localized shimming in a cubic region that is

virtually divided into smaller sticks where shimming is
improved step by step. This tool is named FASTMAP in
Bruker scanners and is very efficient [18].The problem is
that not always the region of interest to be studied can be
fixed into a cubic space.
(c) Using a shimming procedure that optimizes the magnetic
field homogeneity based on a previously acquired 3D
magnetic field distribution. This protocol is called MAP-
SHIM in Bruker instruments. In contrast with FAS-
TMAP, this method allows a localized shimming in a
rectangular region of arbitrary size and orientation [19].
User has to choose the approach that better fits to the study. If
some of the localized methods are going to be used, be sure to
suitably select the cerebral ROI to be assessed. In our system,
the most suitable method is that based on the adjustment of the
first and second orders shim coils.
4. Once the homogeneity of the magnetic field is optimized, the

DSC study can be acquired. Load an optimized EPI-T2* pro-
tocol; select the optimal acquisition parameters (see Table 1)
and place the slice package covering the cerebral area to image
(see Note 4). Because an image will be acquired every
250–400 ms, the number of repetitions has to be selected
depending on the desired time of observation. Usually a total
acquisition time between 45 and 90 s is adequate enough.
Before starting acquisition, make sure that the vein cannulation
is all right by injecting a small saline flush without feeling any
restriction. Change at that time the syringe with saline for a
new one with the contrast agent preloaded. Usually, a 0.3 M
solution is employed for rats and 0.05 M for mice. The CA is
quickly administrated as a bolus 10 s after starting acquisition.
The volume injected will be calculated according to a dose of
0.2–0.3 mmol/kg b.w. The administration can be done either
manually or employing an infusion pump at a rate of approx.
100 μL/s (see Note 5).
Typical contrast media employed do not cross the BBB allow-
ing the injection to remain as a bolus at least in the first pass
across the brain. In pathological conditions that cause a disrup-
tion or alteration in the integrity of the BBB, a partial extrava-
sation of the CA can take place and cause a decrease in the T1 of
the surrounding tissue. In those cases, the perfusion signal will
be affected yielding an underestimation of the CBV.
5. In these perfusion studies based on the administration of an
exogenous contrast agent, it is recommendable to acquire a
T1W protocol after the DSC acquisition to check that the CA
effectively was into the vascular system. Usual acquisition para-
meters appear in Table 2.
The performing of a new perfusion study necessarily needs to
be postponed until the total clearance of the CA from the tissue
(at least 1–2 h).
3.4 Post-Processing 1. Data can be processed and analyzed with a dedicated tool
and Data Analysis incorporated to the software of the MRI spectrometer. Data
are also read with different analysis software like Matlab (The
MathWorks Inc., Natick, MA, 2000) or IDL (Iterative Data
Language, Research System, Boulder, CO, USA) and managed
either with commercial or homemade packages. Images can be
managed as raw-data or DICOM format (Digital Imaging and
Communication in Medicine) that is a standard for handling,
storing, printing, and transmitting information in medical
imaging (http://dicom.nema.org).
2. Images are computed by fitting the signal to the adequate
equations (Eqs. 2–4) in a pixel-by-pixel bases. Depending on
the software, acquisition parameters like number of slices,
number of repetition, contrast agent concentration, time

before agent administration, and so on have to be indicated.
In our lab perfusion maps are obtained by computing pixel
time-evolution signal where the values corresponding to the
first 10 s of each temporal series were considered to calculate
the basal level. Although software interfaces can vary a lot
between different computational applications, Fig. 4 shows
the starting screen of the homemade software that we
employed to analyze DSC studies (MyMapAnalyzer based in
Matlab R2010b).
Majority of software yield the corresponding parametric images
CBC, CBF, and MTT as color-based maps in different image
formats (jpg, tiff, bmp. . .) or data set. Metrics and statistical
analysis are obtained from these data sets.
3. Parametric images are analyzed by selecting areas of interest
(manually or automatically) and considering either each pixel as
a value or a mean of the ROI as a data. When animal models are
employed to assed a pathology and/or treatment, ROIs in
Fig. 4 Screen shot of a DSC study processing with a homemade application (MyMapAnalyzer) developed in
Matlab
different animals has to include always approximately the same

number of pixels and be placed, as much as possible, in the
same location. It is also recommendable to do the ROIs selec-
tion for more than one observer to minimize errors and check
the intra-class correlation coefficient. Parametric images can
also be analyzed with convenient image treatment software.
One of the most used is the freely available Image J (http://
rsbweb.nih.gov/ij/).
The statistical analysis of data can be done with different soft-
ware packages according to the expertise of the laboratory.
Typically employed are: Microsoft Excel, Matlab, IDL, SPSS
(http://www.ibm-spss.com), GraphPad Prism (http://www.
graphpad.com/scientific-software/prism/), and R (https://
www.r-project.org).
As it was previously indicated, absolute quantification of
hemodynamic parameters necessarily relies on the ability of
measuring the signal changes around or inside of a big vessel,
typically paraclinoid or middle cerebral arteries. In practice,
image-based determination of AIF in rodent is challenged
and very difficult to achieve due to low the spatial resolution
inherent to the protocol that entails a big imprecision in select-
ing the area of interest. This limitation is usually overpassed
with different approaches, although it can be useful enough to
carry out a semiquantitative quantification through the use of
laboratory arbitrary units (a.u.) [20], or normalization of the
obtained data with the value of a region not affected in the
acquisition protocol –for example, cerebrospinal fluid or mus-
cle—, or using a ROI in the contralateral healthy hemisphere in
the case of localized lesions like a stroke or a tumor [21].
There are also more sophisticated methods reported in litera-
ture that perform automatic AIF determination based, for
example, on the detection of pixels with maximal contrast
enhancement in different regions of interest [22], the use of
clustering methods [23], or the development of physical mod-
els for the echo-planar signal intensity from an artery [24]
between others (see Note 6).
3.5 CBF, CBV, To illustrate the protocol, some examples of parametric images
and MTT Maps obtained from perfusion DSC studies in our lab are shown in Fig. 5.
4 Notes
1. Tail vein cannulation may be a tricky procedure. Heating the

tail to dilate the veins can help to better localize them and insert
the needle. There are devices to that purpose but also a hot
water filled glove can be useful or immerse the tail in warm
Fig. 5 CBF, CBV, and MTT parametric images. Maps were obtained from perfusion MRI studies of a healthy
animal (upper images) and a rat with a C6 glioma in the right hemisphere (lower images). In both cases the
bolus tracking technique was employed 250 μL of a 0.3 M gadolinium chelate (Gd-DTPA) solution rapidly
injected 10 s after starting acquisition. MR images were computed with homemade software (MyMapAnalyzer)
water (45 C) for 45 s. Also to make a small pressure in the tail

just above of the needle insertion point can be useful.
2. To perform accurate and reliable perfusion DSC MRI studies it
is very important to maintain a constant body temperature and
respiration rate. This is achieved by increasing/decreasing the
level of isoflurane in oxygen and the temperature of the heated
blanket or the system employed to warm the animal.
3. Matrix size is an important parameter to think about in DSC
acquisitions. The smaller, the faster acquisition, the higher
signal-to-noise ratio (SNR) but also the lower resolution.
Depending on the volume and localization of the region to
be assessed, the user can vary these parameters to achieve a
major temporal or spatial resolution. If the ROI is big enough,
64 64 matrix can be used to get a good signal. Nevertheless,
if a higher spatial resolution is required, matrix size has to be
increased to 128 128 at the expense of SNR.
4. Be sure to place the animal in the correct position taking into
account that it is not recommended to angle the slices. EPI
acquisition suffers of n/2 ghosting, due to the reverse read
gradients applied, and geometrical distortions. A double
sampling in the K-space filling improves distortions and

reduces ghost, but increases TE. It is recommendable to use
the shortest TE to avoid the fast T2*, improve the SNR, and
minimize distortions. To select a high spectral bandwidth
(300,000 Hz in 7T, 400,000 MHz in 9.7T) also reduces
geometrical deformations.
5. The bolus has to be injected as quick as possible, the stronger
the concentration of the contrast agent, the higher the drop of
the signal. Acquisition time must include enough reference
images before administrating CA (to obtain a basal level in
the time curve) and cover the first pass of the bolus through
the brain vasculature (but not too long to avoid recirculation).
6. In more or less extension, all quantification methods entail
calculation errors. Users must choose that approach that better
suits the experiments. For example, when an animal model and
a control group are studied, the best option can be to maintain
animals in a similar physiological status—respiration rate, heart
rate, and body temperature—and determine hemodynamic
alterations as a deviation of the basal/normal situation. For
that, CBV, CBF, and MTT measured in cerebral region of
subjects belonging to the animal model group have to be
compared to the same regions in healthy individuals. Results
can be analyzed as a percentage of change.
Acknowledgements
This work was supported by grant SAF2014-53739-R.
References
1. Calamante F, Thomas DL, Pell GS, Wiersma J, 5. Stewart G (1894) Researches on the circulation
Turner R (1999) Measuring cerebral blood time in organs and on the influences which
flow using magnetic resonance imaging techni- affect it: I–III. J Physiol 15:1–27
ques. J Cereb Blood Flow Metab 19 6. Kim SG (2012) Perfusion MR imaging: evolu-
(7):701–735. https://doi.org/10.1097/ tion from initial development to functional
00004647-199907000-00001 studies. NeuroImage 62(2):672–675.
2. Le Bihan D (1995) Diffusion, perfusion and https://doi.org/10.1016/j.neuroimage.
functional magnetic resonance imaging. J Mal 2012.01.015
Vasc 20(3):203–214 7. MacDonald ME, Frayne R (2015) Cerebrovas-
3. Le Bihan D (1992) Theoretical principles of cular MRI: a review of state-of-the-art
perfusion imaging. Application to magnetic approaches, methods and techniques. NMR
resonance imaging. Investig Radiol 27(Suppl Biomed 28(7):767–791. https://doi.org/10.
2):S6–11 1002/nbm.3322
4. Jahng GH, Li KL, Ostergaard L, Calamante F 8. Buxton RB (2012) Dynamic models of BOLD
(2014) Perfusion magnetic resonance imaging: contrast. NeuroImage 62(2):953–961.
a comprehensive update on principles and tech- https://doi.org/10.1016/j.neuroimage.
niques. Korean J Radiol 15(5):554–577. 2012.01.012
https://doi.org/10.3348/kjr.2014.15.5.554 9. Ostergaard L (2005) Principles of cerebral per-
fusion imaging by bolus tracking. J Magn
Reson Imaging 22(6):710–717. https://doi. 17. Shimizu S (2004) Routes of administration. In:
org/10.1002/jmri.20460 Hans J Hedrich, GBaPP (ed) The laboratory
10. Koshimoto Y, Yamada H, Kimura H, mouse. Academic Press, London, p 14
Maeda M, Tsuchida C, Kawamura Y, Ishii Y 18. Gruetter R (1993) Automatic, localized in vivo
(1999) Quantitative analysis of cerebral micro- adjustment of all first- and second-order shim
vascular hemodynamics with T2-weighted coils. Magn Reson Med 29(6):804–811
dynamic MR imaging. J Magn Reson Imaging 19. Kanayama S, Kuhara S, Satoh K (1996) In vivo
9(3):462–467 rapid magnetic field measurement and shim-
11. Barbier EL, Lamalle L, Decorps M (2001) ming using single scan differential phase
Methodology of brain perfusion imaging. J mapping. Magn Reson Med 36(4):637–642
Magn Reson Imaging 13(4):496–520 20. Perez-Carro R, Cauli O, Lopez-Larrubia P
12. Conturo TE, Akbudak E, Kotys MS, Chen (2014) Multiparametric magnetic resonance
ML, Chun SJ, Hsu RM, Sweeney CC, Mark- in the assessment of the gender differences in
ham J (2005) Arterial input functions for a high-grade glioma rat model. EJNMMI Res 4
dynamic susceptibility contrast MRI: require- (1):44. https://doi.org/10.1186/s13550-
ments and signal options. J Magn Reson Imag- 014-0044-4
ing 22(6):697–703. https://doi.org/10. 21. Garcia-Palmero I, Lopez-Larrubia P, Cerdan S,
1002/jmri.20457 Villalobo A (2013) Nuclear magnetic reso-
13. Borges AR, Lopez-Larrubia P, Marques JB, nance imaging of tumour growth and neovas-
Cerdan SG (2012) MR imaging features of culature performance in vivo reveals Grb7 as a
high-grade gliomas in murine models: how novel antiangiogenic target. NMR Biomed 26
they compare with human disease, reflect (9):1059–1069. https://doi.org/10.1002/
tumor biology, and play a role in preclinical nbm.2918
trials. AJNR Am J Neuroradiol 33(1):24–36. 22. Yankam Njiwa J, Ratering D, Baltes C, Rudin
https://doi.org/10.3174/ajnr.A2959 M (2010) Increasing temporal resolution of
14. López-Larrubia P, Cañadillas-Cárdenas E, DSC perfusion MRI using the analytic image
Metelo AM, Arias N, Martı́nez-Maestro M, concept. MAGMA 23(4):251–261. https://
Salguero LA, Cerdán S (2011) Magnetic reso- doi.org/10.1007/s10334-010-0224-9
nance imaging of gliomas. In: Chen CC 23. Yin J, Sun H, Yang J, Guo Q (2015) Auto-
(ed) Advances in the biology, imaging and mated detection of the arterial input function
therapies for glioblastoma. InTech, Rijeka, pp using normalized cut clustering to determine
225–250 cerebral perfusion by dynamic susceptibility
15. Liu W, Wang B, Wolfowitz R, Yeh PH, Nathan contrast-magnetic resonance imaging. J Magn
DE, Graner J, Tang H, Pan H, Harper J, Reson Imaging 41(4):1071–1078. https://
Pham D, Oakes TR, French LM, Riedy G doi.org/10.1002/jmri.24642
(2013) Perfusion deficits in patients with mild 24. Duhamel G, Schlaug G, Alsop DC (2006)
traumatic brain injury characterized by Measurement of arterial input functions for
dynamic susceptibility contrast MRI. NMR dynamic susceptibility contrast magnetic reso-
Biomed 26(6):651–663. https://doi.org/10. nance imaging using echoplanar images: com-
1002/nbm.2910 parison of physical simulations with in vivo
16. Duong TQ (2013) Magnetic resonance imag- results. Magn Reson Med 55(3):514–523.
ing of perfusion-diffusion mismatch in rodent https://doi.org/10.1002/mrm.20802
and non-human primate stroke models. Neurol
Res 35(5):465–469. https://doi.org/10.
1179/1743132813Y.0000000211
Chapter 4
Preclinical Arterial Spin Labeling Measurement of Cerebral

Blood Flow
Eric R. Muir
Abstract
Magnetic resonance imaging has been utilized as a quantitative and noninvasive method to image blood
flow. Arterial spin labeling (ASL) is an MRI technique that images blood flow using arterial blood water as
an endogenous tracer. Herein we describe the use of ASL to measure cerebral blood flow completely
noninvasively in rodents, including methods, analysis, and important considerations when utilizing this
technique.
Key words Arterial spin labeling, Blood flow, Perfusion, Cerebral blood flow, Magnetic resonance
imaging, Rodent
1 Introduction
Arterial spin labeling (ASL) is a magnetic resonance imaging (MRI)

technique to noninvasively image quantitative, absolute blood flow
with high temporal and spatial resolution [1, 2]. Blood flow is an
important physiological parameter as the tissue metabolism is
dependent upon the vascular supply to deliver sufficient oxygen
and nutrients and to remove waste products. Vascular dysfunction
and abnormal blood flow are associated with numerous diseases.
For example, perturbations of basal cerebral blood flow (CBF) and
CBF regulation have been implicated in many neurological diseases
such as stroke, brain tumor, and neurodegenerative disease
[3–6]. Blood flow MRI has been widely used to study normal
physiology and pathophysiology clinically [4, 5] and preclinically
[6–10].
Two major MRI techniques have been developed to quantita-
tively image blood flow, ASL which will be described herein [1, 2],
and dynamic susceptibility contrast (DSC) [11]. The DSC tech-
nique requires the intravenous injection of a gadolinium contrast
agent, which generally precludes repeated or time-series measure-
ments due to the long half-life of the contrast agent in the blood.
59
60 Eric R. Muir
The ASL technique, which is completely noninvasive, uses arterial

blood water as an endogenous tracer by magnetically labeling the
water of inflowing blood (e.g., in the carotid arteries for imaging
CBF). Multiple repeated blood flow measurements can be made
using ASL because the signal of the magnetically labeled water
decays relatively rapidly (with the blood or tissue T1 which are
around 2 s at 7 T), allowing dynamic measurements or averaging
to improve the signal-to-noise ratio.
There are numerous ASL methods, most of which can be
grouped into three broad categories, including continuous, pulsed,
and velocity-selective ASL [2, 12, 13]. Continuous ASL (CASL)
uses a long radiofrequency (RF) pulse combined with magnetic
gradients applied in the direction of flow to adiabatically invert
moving spins as they flow past the labeling plane [1]. Pseudocon-
tinuous ASL (pCASL), an alternate form of continuous ASL, uses a
long train of short RF and gradient pulses to invert spins as they
flow past the labeling plane, similar to CASL [14]. In pulsed ASL
(PASL), a short RF pulse is used to invert blood within a large slab
[15, 16]. Compared to pulsed and velocity-selective ASL, continu-
ous and pseudocontinuous ASL generally provide better blood flow
signal due to higher labeling efficiency [4, 13] but deposit more
power into the subject and can be difficult to implement.
CASL can be performed using the same coil or separate coils for
labeling and imaging. Using the same coil results in strong magne-
tization transfer effects which need to be accounted for by appro-
priate control images and calculations [17, 18], while the separate
labeling coil approach has minimal magnetization transfer if the
imaging and labeling coils are decoupled while the other is trans-
mitting. The two-coil approach also easily allows multi-slice imag-
ing which is otherwise difficult due to magnetization transfer in the
single-coil approach [4, 17].
This chapter describes the animal preparation and MRI proce-
dures to noninvasively image cerebral blood flow in rodents using
the CASL technique [1, 6–9, 19–23]. Although imaging blood
flow of the brain is the most common application of ASL, this
technique can be applied to image blood flow of other organs,
such as the kidneys [24], heart [25], and retina [26, 27].
2 Materials
2.1 Animals, 1. Mice 20–30 g, or rats 200–300 g (see Note 1).

Anesthesia, and 2. Isoflurane or other anesthetic (see Note 2).
Physiological Support
3. Isoflurane vaporizer.
4. Flowmeter.
5. Compressed air (or air pump) and/or oxygen.
Arterial Spin Labeling 61
6. Anesthesia induction chamber.

7. Nose cone.
8. Eye ointment.
9. MRI-compatible temperature probe.
10. MRI-compatible pulse oximeter (MouseOx, Starr Life Sciences
Corp., Oakmont, PA, USA) and/or respiration pressure pad
with force transducer (Small animal monitoring and gating
system, SA Instruments Inc., Stony Brook, NY, USA).
11. Circulating warm water bath with water pads or warm air
blower.
12. Feedback regulated temperature controller (Digi-Sense Tem-
perature Controller, Cole-Parmer, Vernon Hills, IL, USA).
2.2 MRI 1. Preclinical 7 T MRI scanner (Bruker, Billerica, MA, USA) (see
Note 3).
2. Magnetic gradients, 150 G/cm with 6 cm inner diameter
gradient insert for mice, 40 G/cm with 12 cm inner diameter
for rats (see Note 3).
3. Animal holder equipped with ear and tooth bars.
4. Radiofrequency (RF) transmitter/receiver coil for brain imag-
ing (see Note 4).
5. RF transmitter coil for arterial spin labeling at the chest or neck
position for ASL (see Note 5).
6. Switch box to actively detune RF coils for ASL (see Note 6).
3 Methods
3.1 Animal 1. Place the animal in an anesthesia induction chamber with 4–5%
Preparation isoflurane delivered to the box in air or oxygen at 1.5–2 L/min
until the animal loses its righting reflex. Different anesthetics
can be used if preferred, but be aware this will affect blood flow
(see Note 2).
2. Remove the animal from the induction chamber, place a nose
cone over its nose, switch isoflurane delivery to the nose cone,
and reduce the isoflurane to 2%. Check paw reflexes by pinch-
ing the paws and looking for reflex withdrawal to ensure ade-
quate anesthesia.
3. Put eye ointment on both eyes of the animal to prevent the
cornea from drying out while under anesthesia.
4. Place the animal into a head holder with ear bars and tooth bar
to immobilize the skull and minimize animal motion through-
out scanning procedures.
62 Eric R. Muir
5. Move the animal to the animal delivery system of the MRI

scanner, which should have a temperature-controlled water
pad (or warm air blower) to maintain the animal’s body tem-
perature at 37 0.5 C (see Note 7). Reduce the isoflurane to
1–1.5%.
6. Place the physiological monitoring equipment on the animal. A
rectal temperature probe should be inserted using lubrication.
Place the oximetry clip on the animal to monitor arterial blood
oxygenation, heart rate, and respiration rate. The pressure
pad can be placed under the animal to monitor respiration
(see Note 8).
7. Monitor respiration rate, heart rate, oxygen saturation, and
temperature to confirm that physiological parameters remain
stable during imaging (see Note 9).
3.2 MRI 1. The animal should be positioned so that the labeling coil is at
the neck for rats or at the upper chest for mice, to allow labeling
of blood flowing to the brain (see Note 10).
2. Position the imaging RF coil as close to the center of the brain
as possible (see Note 11).
3. Move the animal into the MRI scanner.
4. If the RF coils allow for tuning and matching, then adjust the
tune and match of the imaging and labeling RF coils by adjust-
ing the 1H resonance frequency (300 MHz at 7 T) and
impedance.
5. Use a positioning scan to center the brain in the MRI scanner
in all three directions. Use a large field of view (FOV) if needed
if the brain is initially far away from the center.
6. Run automatic scanner adjustments to adjust frequency, shim,
and calibrate RF pulses.
7. Perform a pilot scan using a 2D gradient echo fast low angle
shot (FLASH) sequence (20 s). Then run a midsagittal rapid
acquisition with relaxation enhancement (RARE) or fast spin
echo scan (30 s).
8. For the following steps, plan three to nine coronal slices to
cover the region of interest in the brain using the pilot scans
from step 7 as reference. To consistently position the imaging
slices in the anterior-posterior direction between animals, use
the midsagittal scan as a reference to center one of the imaging
slices on the anterior commissure (e.g., the third-most anterior
slice out of seven slices total as in Fig. 1).
9. Perform localized volume shimming if needed to improve
magnetic field homogeneity and image quality (see Note 12).
Fig. 1 Example midsagittal RARE image of a rat at 7 T. The black arrow indicates
the anterior commissure. The centers of seven slices spaced 1.5 mm apart are
shown, with the third-most anterior slice positioned on the anterior commissure.
MRI parameters are: data matrix ¼ 128 128, field of view ¼ 25.6 25.6 mm,
spectral width ¼ 50 kHz, effective echo time ¼ 60 ms, repetition time ¼ 2.0 s,
and echo train length ¼ 8
10. For ASL, single-shot, gradient-echo, echo-planar-imaging

(EPI) acquisition is used. Paired images are acquired alter-
nately, one with arterial spin labeling (labeled image) and the
other without (non-labeled). Typical MR parameters are: data
matrix ¼ 64 64 for mice or 96 96 for rats,
FOV ¼ 12.8 12.8 mm for mice or 25.6 25.6 mm for
rats, three to nine coronal slices, slice thickness ¼ 1.0 mm for
mice or 1.5 mm for rats, spectral width ¼ 250–350 kHz, echo
time (TE) ¼ 8–10 ms, repetition time (TR) ¼ 3.0 s, partial
Fourier ¼ 77%, and 90 flip angle. Continuous arterial spin
labeling (CASL) is employed with separate imaging and label-
ing coils, using a square radio frequency pulse for the labeling
coil with a labeling duration of 2.2 s in the presence of a 1 G/
cm gradient along the flow direction such that the condition of
adiabatic inversion is satisfied (see Note 13). A post labeling
delay (PLD) of 350 ms is used for mice or 500 ms for rats (see
Note 14). The sign of the frequency offset of the label pulse is
switched for non-labeled images. The number of averages is
typically 50 or more, depending on the required signal-to-
noise ratio and spatial resolution. Example CBF images are
shown in Fig. 2.
11. A proton density weighted image is often acquired for use in
the blood flow calculation (see Note 15) using the same imag-
ing sequence, parameters, and geometry as used for ASL in
64 Eric R. Muir
Fig. 2 Example CBF maps (in mL/g/min) from continuous ASL in a (a) rat and (b) mouse. Some brain structures
should be able to be distinguished in the CBF maps, such as the corpus callosum with low blood flow
step 10, but without the ASL preparation and with a long TR
of 10 s.
12. Optionally, an anatomical reference image can be acquired
using a T2-weighted RARE sequence with parameters such as
data matrix ¼ 128 128, FOV ¼ 12.8 12.8 mm for mice or
25.6 25.6 mm for rats, slice thickness ¼ 1.0 mm for mice or
1.5 mm for rats, spectral width ¼ 50 kHz, echo train length ¼ 8,
effective TE ¼ 50 ms, TR ¼ 2.0 s, and 8 averages (4.3 min).
3.3 Animal Recovery 1. At the conclusion of imaging, remove the animal from the
scanner and holder. Discontinue anesthesia after removing
physiological monitoring probes. Place the animal on a heating
pad on low until it is fully awake and ambulatory (see Note 16).
2. Once the animal is awake and ambulatory it can then be
returned to its home cage for recovery.
3. The animal should be monitored throughout the day for signs
of distress.
3.4 Image Analysis 1. MRI image processing and analysis can be done using custom
written codes or various freely available MRI analysis tools.
Blood flow calculation from ASL data can be done using soft-
ware available online such as ASLtbx (University of Pennsylva-
nia) [28] or using custom written codes implementing the
equation in step 2. If needed, image co-registration can be
performed using software such as Statistical Parametric
Mapping (SPM, Wellcome Trust Centre for Neuroimaging,
University College London). If animals are properly secured
in ear and tooth bars and adequate anesthesia is maintained,
motion should not occur. Image display can be done with
software such as Mango (Research Imaging Institute, Univer-
sity of Texas Health Science Center at San Antonio). Several
other free software programs are also available to process and
display MRI images.
2. For two-coil CASL acquisitions (see Note 17), CBF images

with intensity in units of mL/g/min can be calculated pixel-by-
pixel using [2, 23, 29]
CBF ¼ 60λ ðS NL S L Þ expðPLD=T 1A Þ=½2α T 1 S 0 ð1 expðLD=T 1 ÞÞ ð1Þ
SNL and SL are signal intensities of the non-labeled and labeled

images, respectively. S0 is the proton density weighted image (see
Note 15). λ, the brain–blood partition coefficient of water, is
~0.9 mL/g for averaged gray and white matter [30]. T1 is the
apparent longitudinal relaxation time of the brain, which is about
1.7 s at 7 T [31], and T1A is the longitudinal relaxation time of
arterial blood which is about 2.2 s at 7 T [32] (see Notes 18 and
19). α is the labeling efficiency, which has been measured to be
0.75–0.9 in animals using CASL (see Note 20) [6, 7, 19, 23].
4 Notes
1. Animals at different sizes than the listed range of adult weights

can also be imaged with ASL. For example, we have imaged
mice from 12 g [10] up to 70 g. However, the positioning of
the RF coils and labeling plane as well as the imaging field of
view and resolution may need to be adjusted dependent on the
animal’s size.
2. Different anesthetic agents have different effects on blood flow
values. Comparisons of blood flow made under different anes-
thetics or dosages need to be made with caution. For example,
isoflurane is a vasodilator [8] while a ketamine/xylazine cock-
tail is vasoconstrictive [9].
3. Preclinical MRI systems generally have higher magnetic field
strengths and more powerful gradients than clinical MRI sys-
tems, which provide better signal to noise and higher resolu-
tion. However, ASL studies of rodents have been performed
using clinical MRI systems [33].
4. Surface coils can provide higher signal-to-noise ratio compared
to volume coils. Surface coils with inner diameter of 1.4 cm can
be used for mice and 2.0 cm for rats. The image intensity using
surface coils for transmit and receive, however will be inhomo-
geneous as both the transmit and receive fields will drop off
away from the coil. A separate volume transmit coil can be used
to minimize inhomogeneity of the transmit field while main-
taining SNR with a surface receiver coil, although this requires
additional hardware and the coils must be decoupled from each
other [34].
5. We described the use of a continuous ASL sequence with
separate labeling and imaging coils. Many variations of ASL
66 Eric R. Muir
exist and different scanners may have different sequences avail-

able or may have hardware limitations preventing some meth-
ods. All methods give reasonably similar results with some
advantages and disadvantages. For more information see
[2, 12]. If software or hardware are not available to perform
any type of ASL, blood flow can alternatively be measured with
dynamic susceptibility contrast MRI using an intravenous
injection of a gadolinium MRI contrast agent [11].
6. For two-coil continuous ASL, the imaging coil and the labeling
coil should be detuned while the other coil is transmitting to
avoid artifacts and inaccurate CBF values due to the coils
coupling.
7. The body temperature of small rodents can drop quickly under
anesthesia, so thermal support needs to be provided to main-
tain body temperature. However, animals can also be over-
heated if the heater is set to high or runs for too long, so
careful monitoring of temperature is needed. A feedback tem-
perature regulator can be used to automatically switch the
heater off if the animal’s temperature goes above 37 C.
8. Stable maintenance of animal physiology under anesthesia is
important since blood flow is affected by physiological para-
meters, such as blood gases and temperature. Maintain the
animal’s body temperature within 37 C 0.5 C, as rodents
do not maintain their body temperature under anesthesia.
Additionally, the animal’s physiology should be monitored
continuously to maintain the respiration rate, heart rate, and
arterial oxygen saturation while under anesthesia.
9. Under isoflurane, the respiration rate can be maintained at
about 100–120 breaths/min in mice and 60–70 breaths/min
in rats. If the respiration rate or heart rate are becoming too fast
or too slow, the depth of anesthesia can be adjusted easily by
adjusting the isoflurane level slightly. However, isoflurane is a
vasodilator, so this should be done carefully to avoid confound-
ing the experiment. The arterial oxygen saturation under iso-
flurane anesthesia and air is usually around 93–97%. If it drops
below 90%, some supplementary oxygen can be given. How-
ever, changing the oxygen levels can also affect blood flow, so
this should be avoided unless necessary.
10. Variability in the supplying arteries and positioning of labeling
coils and planes across animals can lead to errors with ASL. To
verify the correct placement of the labeling coil, images of the
labeling region can be obtained by temporarily connecting or
selecting the labeling coil as the transceiver coil to acquire
images. A fast T1-weighted spoiled gradient echo sequence
with short TR can provide bright blood images. Images should
show the carotid arteries at the neck or the heart at the chest.
11. For surface coils, avoid pressing the coil too hard on the
animal’s head as it would increase loading. Having the coil or
its circuitry too near to the head can also cause artifacts at the
cortical surface.
12. High-order localized shimming on the brain can improve
image quality and reduce distortion in EPI images. However,
for ASL, we have found that this can cause poor shim quality
outside of the brain leading to poor labeling efficiency at the
neck, so first order global shimming may be preferable.
13. Calibration of labeling RF power and gradients can be done by
acquiring multiple short ASL scans with arrayed RF labeling
power or gradient amplitude to find the optimal power or
gradient that provide the maximum difference between
non-label and label images [35]. However, too high a labeling
power can directly saturate the brain, which will be seen as very
bright areas in the CBF maps in the posterior brain. With the
long RF pulse used for labeling in CASL, high RF power also
poses a risk of heating and burns. Since these scans would
lengthen imaging times and the optimal values are usually
consistent, this is usually done initially on a small number of
subjects to determine optimal settings to use for future studies.
14. Accurate calculation of blood flow assumes that all the labeled
blood has been delivered to the tissue, so a delay is usually
given after the end of the labeling period to allow time for the
labeled blood to flow from the labeling region to the tissue.
The post labeling delay should thus be longer than this arterial
transit time, but with too long a PLD the labeled blood water
signal may be reduced too much as it decays with T1. The
arterial transit time to the rat gray matter is around
200–400 ms, dependent upon brain region [21, 22]. Some
diseases, such as stroke, can significantly prolong the transmit
time which can affect CBF quantification. The arterial transit
time can be measured in addition to blood flow by acquiring a
series of ASL scans with multiple PLDs or by other methods, at
the cost of longer imaging times [21, 22, 36].
15. The non-labeled images of the ASL scan are sometimes used
instead of a separately acquired proton density image for blood
flow calculation. This will introduce some inaccuracy in the
calculated blood flow as the TR used for ASL scans is usually
short enough to have some T1 weighting. However, using the
non-labeled images avoids the acquisition of an extra proton
density scan and avoids errors if the intensities of the ASL and
proton density data are scaled differently by the MRI scanner
and not accounted for. The non-labeled images can be scaled
by 1/(1exp(TR/T1)), where T1 is the tissue T1, to com-
pensate if TR is not several times longer than T1 [2]. If
68 Eric R. Muir
background suppression [20, 37] was used for ASL data, a

separate proton density image without background suppres-
sion should be acquired.
16. Animals should recover from isoflurane anesthesia quickly,
within a few minutes. However, some injectable anesthetics
may have long recovery times, in which case the animal’s
physiology should continue to be monitored until it begins
to awaken. In particular, small rodents can quickly become
under or over heated if their temperature is not monitored.
17. Many models have been used to calculate blood flow via ASL
approaches. We give a model for CBF calculation [29] includ-
ing common parameters relevant for two-coil CASL of rodents
[23]. Different or more complicated models have been used,
although the results are generally similar regardless of the exact
model [38]. Different types of ASL may also require different
models [2, 18, 29].
18. It is common to simply use a single average tissue T1 value,
such as of gray matter, for calculations of CBF. Individual T1
maps can be acquired and used for blood flow calculation, and
may be particularly helpful if tissue T1 is expected to vary
substantially [23]. However, acquiring T1 maps will increase
the imaging time, and noisy or low quality T1 maps can
degrade the quality of the blood flow map. Literature values
of arterial blood T1 are used as direct measurement of blood T1
in vivo is difficult.
19. In ASL, the labeled blood water signal will decay with the T1 of
arterial blood while in the arterial compartment and with the
T1 of tissue while in the tissue compartment. The calculation
should ideally account for the time the labeled blood water
remains in the blood as in [29], but the arterial transit time is
usually not measured (see Note 13). Equation 1 assumes that
the arterial transit time is the same as the post labeling delay
[23], although there will be some variation across the brain and
between subjects in the actual arterial transit time. However, as
the arterial transit time in rodents is short compared to humans
[21, 22], uncertainty of the actual transit time likely results in
only small errors in rodents. Some studies even use only a single
T1 value for both blood and tissue to simplify the calculations,
assuming they have similar T1 [2].
20. Labeling efficiency can be measured from the ASL signal dif-
ference in a large artery carrying the labeled blood, such as the
carotid arteries supplying the brain. The labeling efficiency can
be calculated as (SNLSL)/(2SNL), where SNL is the signal in
the artery from non-labeled images and SL is the signal in the
artery from labeled images [6, 7, 19, 23, 39]. If magnitude
data is used, then the label signal should be made negative for
labeling conditions where the labeling efficiency is greater than

0.5.
References
1. Williams DS, Detre JA, Leigh JS, Koretsky AP 10. Muir ER, De La Garza B, Duong TQ (2013)
(1992) Magnetic resonance imaging of perfu- Blood flow and anatomical MRI in a mouse
sion using spin inversion of arterial water. Proc model of retinitis pigmentosa. Magn Reson
Natl Acad Sci U S A 89:212–216 Med 69(1):221–228. https://doi.org/10.
2. Alsop DC, Detre JA, Golay X, Gunther M, 1002/mrm.24232
Hendrikse J, Hernandez-Garcia L, Lu H, Mac- 11. Calamante F, Thomas DL, Pell GS, Wiersma J,
Intosh BJ, Parkes LM, Smits M, van Osch MJ, Turner R (1999) Measuring cerebral blood
Wang DJ, Wong EC, Zaharchuk G (2015) flow using magnetic resonance imaging techni-
Recommended implementation of arterial ques. J Cereb Blood Flow Metab 19:701–735
spin-labeled perfusion MRI for clinical applica- 12. Barbier EL, Lamalle L, Decorps M (2001)
tions: a consensus of the ISMRM perfusion Methodology of brain perfusion imaging. J
study group and the European consortium for Magn Reson Imaging 13:496–520
ASL in dementia. Magn Reson Med 73 13. Wong EC (2014) An introduction to ASL
(1):102–116. https://doi.org/10.1002/ labeling techniques. J Magn Reson Imaging
mrm.25197 40(1):1–10. https://doi.org/10.1002/jmri.
3. Brown GG, Clark C, Liu TT (2007) Measure- 24565
ment of cerebral perfusion with arterial spin 14. Dai W, Garcia D, de Bazelaire C, Alsop DC
labeling: part 2. Applications. J Int Neuropsy- (2008) Continuous flow-driven inversion for
chol Soc 13:526–538 arterial spin labeling using pulsed radio fre-
4. Wolf RL, Detre JA (2007) Clinical neuroimag- quency and gradient fields. Magn Reson Med
ing using arterial spin-labeled perfusion MRI. 60(6):1488–1497. https://doi.org/10.1002/
Neurotherapeutics 4:346–359 mrm.21790
5. Telischak NA, Detre JA, Zaharchuk G (2015) 15. Wong EC, Buxton RB, Frank LR (1998)
Arterial spin labeling MRI: clinical applications Quantitative imaging of perfusion using a sin-
in the brain. J Magn Reson Imaging 41 gle subtraction (QUIPSS and QUIPSSII).
(5):1165–1180. https://doi.org/10.1002/ Magn Reson Med 39:702–708
jmri.24751 16. Kim S-G (1995) Quantification of relative cere-
6. Shen Q, Fisher M, Sotak CH, Duong TQ bral blood flow change by flow-sensitive alter-
(2004) Effects of reperfusion on ADC and nating inversion recovery (FAIR) technique:
CBF pixel-by-pixel dynamics in stroke: charac- application to functional mapping. Magn
terizing tissue fates using quantitative diffusion Reson Med 34:293–301
and perfusion imaging. J Cereb Blood Flow 17. Alsop DC, Detre JA (1998) Multisection cere-
Metab 24:280–290 bral blood flow MR imaging with continuous
7. Silva AC, Zhang W, Williams DS, Koretsky AP arterial spin labeling. Radiology 208
(1995) Multi-slice MRI of rat brain perfusion (2):410–416. https://doi.org/10.1148/radi
during amphetamine stimulation using arterial ology.208.2.9680569
spin labeling. Magn Reson Med 33:209–214 18. Alsop D, Detre J (1996) Reduced transit-time
8. Sicard K, Shen Q, Brevard ME, Sullivan R, sensitivity in noninvasive magnetic resonance
Ferris CF, King JA, Duong TQ (2003) imaging of human cerebral blood flow. J
Regional cerebral blood flow and BOLD Cereb Blood Flow Metab 16:1236–1249
responses in conscious and anesthetized rats 19. Muir ER, Shen Q, Duong TQ (2008) Cerebral
under basal and hypercapnic conditions: impli- blood flow MRI in mice using the cardiac spin-
cations for functional MRI studies. J Cereb labeling technique. Magn Reson Med
Blood Flow Metab 23(4):472–481 60:744–748
9. Lei H, Grinberg O, Nwaigwe CI, Hou HG, 20. Shen Q, Duong TQ (2011) Background sup-
Williams H, Swartz HM, Dunn JF (2001) The pression in arterial spin labeling MRI with a
effects of ketamine-xylazine anesthesia on cere- separate neck labeling coil. NMR Biomed
bral blood flow and oxygenation observed 24:1111–1118
using nuclear magnetic resonance perfusion
imaging and electron paramagnetic resonance 21. Thomas DL, Lythgoe MF, van der Weerd L,
oximetry. Brain Res 913:174–179 Ordidge RJ, Gadian DG (2006) Regional vari-
ation of cerebral blood flow and arterial transit
70 Eric R. Muir
time in the normal and hypoperfused rat brain coefficient for water? J Cereb Blood Flow
measured using continuous arterial spin label- Metab 5:65–69
ing MRI. J Cereb Blood Flow Metab 26 31. Muir ER, Cardenas D, Huang S, Roby J, Li G,
(2):274–282. https://doi.org/10.1038/sj. Duong TQ (2014) MRI under hyperbaric air
jcbfm.9600185 and oxygen: effects on local magnetic field and
22. Meng Y, Wang P, Kim SG (2012) Simulta- relaxation times. Magn Reson Med 72
neous measurement of cerebral blood flow (4):1176–1181. https://doi.org/10.1002/
and transit time with turbo dynamic arterial mrm.25027
spin labeling (Turbo-DASL): application to 32. Dobre MC, Ugurbil K, Marjanska M (2007)
functional studies. Magn Reson Med 68 Determination of blood longitudinal relaxation
(3):762–771. https://doi.org/10.1002/ time (T1) at high magnetic field strengths.
mrm.23294 Magn Reson Imaging 25(5):733–735.
23. Debacker CS, Daoust A, Kohler S, Voiron J, https://doi.org/10.1016/j.mri.2006.10.020
Warnking JM, Barbier EL (2016) Impact of 33. Wegener S, Wu WC, Perthen JE, Wong EC
tissue T1 on perfusion measurement with arte- (2007) Quantification of rodent cerebral
rial spin labeling. Magn Reson Med. https:// blood flow (CBF) in normal and high flow
doi.org/10.1002/mrm.26255 states using pulsed arterial spin labeling mag-
24. Rajendran R, Lew SK, Yong CX, Tan J, Wang netic resonance imaging. J Magn Reson Imag-
DJ, Chuang KH (2013) Quantitative mouse ing 26(4):855–862. https://doi.org/10.
renal perfusion using arterial spin labeling. 1002/jmri.21045
NMR Biomed 26(10):1225–1232. https:// 34. Doty FD, Entzminger G, Kulkarni J,
doi.org/10.1002/nbm.2939 Pamarthy K, Staab JP (2007) Radio frequency
25. Campbell-Washburn AE, Price AN, Wells JA, coil technology for small-animal MRI. NMR
Thomas DL, Ordidge RJ, Lythgoe MF (2013) Biomed 20(3):304–325. https://doi.org/10.
Cardiac arterial spin labeling using segmented 1002/nbm.1149
ECG-gated Look-Locker FAIR: variability and 35. Maccotta L, Detre JA, Alsop DC (1997) The
repeatability in preclinical studies. Magn Reson efficiency of adiabatic inversion for perfusion
Med 69(1):238–247. https://doi.org/10. imaging by arterial spin labeling. NMR Biomed
1002/mrm.24243 10(4–5):216–221
26. Muir ER, Duong TQ (2011) MRI of retinal 36. Wells JA, Lythgoe MF, Gadian DG, Ordidge
and choroidal blood flow with laminar resolu- RJ, Thomas DL (2010) In vivo Hadamard
tion. NMR Biomed 24:216–223 encoded continuous arterial spin labeling
27. Li G, De La Garza B, Shih YY, Muir ER, (H-CASL). Magn Reson Med 63
Duong TQ (2012) Layer-specific blood-flow (4):1111–1118. https://doi.org/10.1002/
MRI of retinitis pigmentosa in RCS rats. Exp mrm.22266
Eye Res 101:90–96. https://doi.org/10. 37. Maleki N, Dai W, Alsop DC (2012) Optimiza-
1016/j.exer.2012.06.006 tion of background suppression for arterial spin
28. Wang Z, Aguirre GK, Rao H, Wang J, labeling perfusion imaging. MAGMA 25
Fernandez-Seara MA, Childress AR, Detre JA (2):127–133. https://doi.org/10.1007/
(2008) Empirical optimization of ASL data s10334-011-0286-3
analysis using an ASL data processing toolbox: 38. Steger TR, White RA, Jackson EF (2005)
ASLtbx. Magn Reson Imaging 26(2):261–269 Input parameter sensitivity analysis and com-
29. Buxton RB, Frank LR, Wong EC, Siewert B, parison of quantification model for continuous
Warach S, Edelman RR (1998) A general arterial spin labeling. Magn Reson Med
kinetic model for quantitative perfusion imag- 53:895–903
ing with arterial spin labeling. Magn Reson 39. Talagala SL, Ye FQ, Ledden PJ, Chesnick S
Med 40(3):383–396 (2004) Whole-brain 3D perfusion MRI at 3.0
30. Herscovitch P, Raichle ME (1985) What is the T using CASL with a separate labeling coil.
correct value for the brain-blood partition Magn Reson Med 52:131–140
Chapter 5
Dynamic Contrast-Enhanced MRI

Jennifer Moroz and Stefan A. Reinsberg
Abstract
Dynamic contrast-enhanced MRI in pre-clinical imaging allows the in-vivo monitoring of vascular,
physiological properties in normal and diseased tissue. There is considerable variation in the methods
employed owing to the different questions that can be asked and answered about the physiologic alterations
as well as morphologic changes in tissue. Here we review the typical decisions in the design and execution of
a dynamic contrast-enhanced MRI study in mice although the findings can easily be transferred to other
species. Emphasis is placed on highlighting the many pitfalls that wait for the unaware pre-clinical MRI
practitioner and that go often unmentioned in the abundant literature dealing with dynamic contrast-
enhanced MRI in animal models.
Key words Pharmacodynamics, Gadolinium chelate, Blood flow, Vessel permeability, Ktrans, IAUC60
1 Introduction
Dynamic Contrast-Enhanced MRI (DCE-MRI) has made tremen-

dous contributions to the in-vivo study of microvascular structure
and function since its development in the 1980s by Weinberg.
Initial hopes for DCE-MRI to provide an accurate, clinical
imaging biomarker especially for the development of antivascular
agents have by now been tempered. So far there are only few
successful examples of using DCE-MRI for imaging biomarkers in
the clinic as detailed in a consensus statement by O’Connor
et al. [1]. Contrary to conventional, i.e. static contrast-enhanced
MRI, DCE-MRI is not used routinely in the clinic for diagnosis or
even treatment follow-up. However, it has by now established its
form position as a research tool in pre-clinical studies of cancers and
diseases that impact microvascular function.
When reviewing the methods of DCE-MRI in different
research settings, it becomes quickly apparent that the methods of
acquisition and data analysis can be tuned to an array of different
questions that are being addressed in the deployment of
DCE-MRI. Consequently, this review can not provide the
71
72 Jennifer Moroz and Stefan A. Reinsberg
authoritative sequence of steps for the performance of pre-clinical

DCE-MRI but rather it will lay out the design choices to be
considered for a pre-clinical DCE-MRI study and the trade-offs
that inevitable have to be made.
1.1 Contrast Agents Tissue differentiation in MR images is predominantly related to

and Mechanism of differences in the proton density, T1 and T2 relaxation times. How-
Signal Enhancement ever, Weinmann et al. discovered that some tissues (malignant
vs. benign tumors, or between normal tissues) have similar time
constants, making them difficult to differentiate [2]. Contrast
agents are used to increase the contrast between magnetically
equivalent, but histologically different, tissues. Contrast agents
are administered into a peripheral vein (tail or jugular in rodents)
as a bolus injection through a catheter. Once in the blood, the bolus
will distribute throughout the entire vasculature and diffuse into
surrounding tissues.
MR contrast agents often contain a rare-earth element, such as
Gadolinium (Gd3+) or Manganese (Mn2+), which has unpaired
electrons in the outer shell [3–5]. Chelation of these ions creates
a strong paramagnetic complex that reduces the relaxation time
constants of water protons in the near vicinity. Contrast agents
may be classified as T1 or T2 agents, depending on the relative
changes in the T1 and T2 relaxation time constants. A T1-agent
will have a greater impact on the T1 relaxation time, which leads to a
signal increase in T1-weighted images.
The Gd ion—often chelated with diethylenetriamine pentaace-
tic acid (DTPA)—has the strongest effect on reducing the T1
relaxation times of tissue, and is therefore widely used in
DCE-MRI studies. The chelate is well tolerated in animals with
clinically relevant doses of 0. 1–0. 5 mmol/kg [2]. Gd-DTPA has a
low molecular weight (570 Da), allowing it to readily pass through
the vasculature into the interstitial space of surrounding tissues.
The highly permeable nature of tumor vessel means that the tumor
will enhance at a greater rate than normal [6].
The change in tissue relaxation rates (R1 ¼ 1/T1) is linearly
related to the contrast-agent concentration over a clinically relevant
range as described by [4]:
1=T 1, obs ¼ 1=T 1, 0 þ r 1 ∗½Gd ð1Þ

where r1 is the relaxivity of the contrast agent in units of mM1 s1.
The relaxivity typically depends on the magnetic field strength.
However, the relaxivity of small molecular-weight Gd chelates,
with low proton binding (e.g., Magnevist), has a relatively low
dependence on field strength and tends to decrease at higher field
strengths. Shen et al. provide T1 relaxivities for a number of
Gd-based contrast agents at 1.5 T, 3 T, and 7 T. For Gd-DTPA,
the relaxivity is (4.3 0.4) mM1 s1 at 1.5 T, (3.8 0.2) mM1
s1 at 3 T, and (3.1 0.4) mM1 s1 at 7 T [7].
Preclinical MRI 73
Tofts et al. listed the choices of currently available, low-molecular weight

contrast agents [8]: “Any low molecular weight contrast agent can in
principle be used for DCE methodology. The initial work by Tofts et al.
(1991) was carried out with Gd-DTPA, size 570D, and then with Magnevist
(938D) [9]. Clearly larger molecules will have lower permeability and hence
Ktrans values, and the AIF may alter a little with viscosity. In view of the
concerns about [nephrogenic systemic fibrosis], there will be value in gain-
ing experience using the newer cyclic compounds. Potentially suitable can-
didate compounds are as follows. Dotarem (754D), Eovist (725D),
Gadovist (605D), Magnevist (938D), Multihance (1058D), Omniscan
(574D), Optimark (662D), Primovist (685D), Prohance (559D) and
Teslascan (757D) (see http://www.rxlist.com).”
Changes in the relaxation rate of the tissue in response of

contrast agent distribution are monitored with repeated, serial
imaging using T1-weighted sequences.
1.2 Tissue Models In order to extract physiologically relevant parameters from

DCEMRI data, tracer-kinetic models need to be fit to the acquired
signal time curves. In the 1990s, the first generation of DCEMRI
models was developed by Larsson et al. [10], Brix et al. [11], and
Tofts et al. [9]. The Tofts model became the most widely applied
model yielding parameters for the transfer constant from plasma to
the extra-cellular extra-vascular space, Ktrans and rate constant, kep
¼ Ktrans/ve. Briefly, the model describes the concentration time
curve as the convolution of an impulse response function and the
arterial input function, ca(t):
CðtÞ ¼ K trans e tkep ∗c a ðtÞ: ð2Þ
Since the intravascular tracer concentration does sometimes
appreciably contribute to the signal change in the tissue of interest,
a modified, ad-hoc modification, commonly known as extended
Tofts model, has been used:
CðtÞ ¼ vp C a ðtÞ þ K trans e tkep ∗c a ðtÞ: ð3Þ

While much has been written about the applicability and lim-
itations of the Tofts models (see, for instance, review by Sourbron
et al. [12]), it is arguably still the most commonly used model.
Note 4.2 below shows the two-compartment exchange model
(2CXM) which is a preferable alternative to the Tofts models
above should the data quality (higher temporal resolution) support
the fitting of four rather than just two (simple Tofts) or three
(extended Tofts) parameters.
1.3 Arterial Input The arterial input function (AIF) describes the concentration of a
Function contrast agent in the blood. Ideally, the AIF should be sampled in a
vessel feeding the tissue of interest and concurrently with the DCE
experiment (see [13] for an elegant example). Measuring the AIF in
small animals is difficult due to their small spatial dimensions and
limited number of vessels of a sufficient size [14]. Common sites for
AIF measurements include the heart [15], the aorta/vena cava, the
iliac artery [16], and the tail veins [17]. For increased accuracy in
the model parameters, the AIF should be sampled with a time
resolution at, or greater than, that of the DCE images [18].
There are four main methods to obtain the AIF:
1. Using a population-averaged AIF from the literature [19–22],
2. Measuring an AIF in control population using your planned
procedure,
3. Measuring the AIF from a small injection prior to the actual
DCE-MRI experiment, also known as dual-bolus
method [23, 24], and
4. Employing a reference-region approach by imaging known
tissue (commonly muscle) in addition to the site of interest
during the DCE-MRI experiment [25] (see Note 4.1).
There is much debate on which method produces the most
accurate measurement of the true curve. Naturally, methods 2, 3,
and 4 are preferred in cases where the variation between individuals
is large or where the injection and imaging protocols deviate from
those published in the literature [14, 15]. For the purposes of this
chapter, measuring an AIF separately in a control experiment
(method 2) will be outlined in greater detail in Subheading 3.
Those interested in the other methods are referred to the notes.
2 Materials
1. The experimental setup is common to all in-vivo pre-clinical

MRI setups with the addition of the DCE injection during the
MRI imaging session. A high-field ( 7 T) MRI scanner is the
preferred system.
2. A receive coil specific to the tissue of interest should be
selected. Typically, for subcutaneous tumors in rodents, surface
coils are preferred. For studies in brain, phased-array coils are
also available and provide as good a signal-to-noise as surface
coils with usually better overall coverage.
3. A power injector (e.g., KD Scientific, Holliston, MA, USA)
capable of injecting liquid with standard type syringes at variable
flow rates is required. It is not advisable to perform the DCE
injection manually, since it is not possible to reproducibly inject
a bolus at the required, constant rate without the use of a power
injector.
4. MRI-compatible animal monitoring equipment for the moni-
toring of at least internal temperature and breathing frequency
(e.g., Small Animal Instruments, Inc., Stony Brook, USA). In
our institution we use a fiber-optic temperature probe (tip
diameter o.d. ¼ 1 mm) and a pneumatic pillow for respiratory
Preclinical MRI 75
monitoring. In addition, a fiber-optic pulse oximetry sensor

can provide information on the arterial blood oxygen
saturation.
5. Drugs and other materials include isoflurane, medical air
and/or oxygen, (heparinized) saline solution, lactate ringer
solution, eye lubricant (e.g., Lacri-lube, Allergan Inc.), con-
trast agent such as Magnevist (Bayer, Inc.) or Omniscan
(GE Healthcare, Inc.).
6. Clinically available contrast agents—such as Magnivist or
Omniscan—have stock concentrations of 0.5 M and 1.0 M,
respectively. A required dose for DCE-MRI scanning is in the
range of 0.1. . .0.3 mmol/kg and would require impossibly
small volumes of the stock solutions. To handle volumes with
less error, it is advisable to dilute stock solutions to ca. 30. . .60
mM. For instance, a 25 g mouse would receive 125 μL of
60 mM Omniscan to achieve a dose of 0.3 mmol/kg.
7. For restraint and placement, surgical tape may be used. Tubing
(e.g., PE20 polyethylene tubing) of known inner diameter to
hold the saline flush and contrast agent bolus is required.
8. PE tubing of larger inner diameter extends to the power injec-
tor is required for administration of contrast agent during the
imaging session. Syringes (1 or 3 mL) are needed to flush the
lines or effect injections.
9. Our institution uses a custom-made induction chamber that
has connections to inflowing anesthetic gas and a V-shaped slit
in a side-wall allowing the restraining of the animals by its tail
for cannulation.
2.1 Analysis Almost none of the vendor-software packaged with the MRI system
Software consoles allows data analysis of DCE-MRI. Consequently, a great
number of alternative solutions have emerged with no clear leader
in sight. A further problem is the large inter-software variability for
parameter maps derived from the same data but different software
solutions as recently reviewed by Beuzit et al. [26]. A large number
of closed-source, proprietary solutions also exist and require further
on-site quality assurance to ascertain the suitability for the task. At
our institution, we use a custom-built software solution relying on
an array of open-source Python packages mostly from the scipy
ecosystem [27]; in particular, pandas [28], and nibabel [29].
Tools such as the Osirix (Pixmeo SARL and Osirix Foundation)
plugin DCE_Tool by K.H. Sung, or Brandon Whitcher’s and
Volker Schmid’s DCEMRI package for the R programming envi-
ronment streamline some of the data processing steps described in
the Methods below. However, they do not remove the need to
critically check the parameter maps obtained and, unfortunately,
sometimes make it harder to inspect the intermediate results in the
calculation and optimization.
3 Methods
3.1 Preparation The contrast agents should be administered with a line that avoids
of the Contrast the premature injection of contrast agents. Since the blood pressure
Injection of the animal fluctuates from the time the intra-venous access is
established to the time the DCE-MRI experiments is performed, it
is possible for blood to be pushed into the connected line (increase
in blood pressure) or intra-venous fluid to flow into the animal.
Figure 1 shows the assembly of the line. A heparin lock of 25 μL
volume (segment A in the figure) will act as a buffer between the
animal and the actual contrast-agent volume.
1. Dilute the stock contrast agent with saline. Determine the
volume of contrast agent to be injected based on a dose of
0.1 mmol/kg but not exceeding 0.3 mmol/kg body
weight [5].
2. Determine the length of PE20 (segment B in the figure) tubing
needed for this volume. The length of tubing is calculated from
l ¼ VCA/(π ∗ r2), where r ¼ 0. 19 mm for PE20.
3. Fill the line with the contrast agent and set aside.
4. Connect a PE20 line of a length sufficient to hold a volume of
25 μL pre-bolus heparinized saline with the contrast-agent
PE20 line. Blunt stents may be used to create dead-volume
free connections between these tubes.
5. A further tube connecting the PE20 assembly with the power
injector (segment C in the figure) is filled with (heparin-free)
saline for the post-injection flush.
3.2 Catheterization 1. Prior to anesthesia and after sufficient warming of the animal
of the Mouse Tail using a heat lamp, a catheter is placed in the mouse tail vain. A
and Anesthesia proper restraint such that the experimenter has ready access to a
motionless tail is required such as our custom-made perspex
box with a V-shaped slit in one of the walls (see Fig. 2). The
catheter should be glued or taped in place.
Fig. 1 Setup for the injection line with heparin lock, bolus line, and saline flush
Preclinical MRI 77
Fig. 2 Mouse tail catheterized (step 2) and anesthesia induced. At this stage care must be taken to maintain
body temperature. Mouse can now be transferred to the RF coil and animal holder (left in the photo). The
catheter can be cut and connected to the bolus line
2. As soon as the catheter is placed, anesthesia can be induced (the

ensuing reduction in blood pressure makes catheterization
after onset of anesthesia more difficult). Heparinized saline
should be used in the catheter line to reduce the risk of clotting.
It bears noting that considerable skill is required in successfully
placing a mouse intra-venous access that will remain function-
ing until tens of minutes later when the actual injection will
take place.
3. Isoflurane is commonly used for DCE-MRI experiments due to
its rapid induction of and recovery from anesthesia. The level of
anesthesia may also be changed rapidly. In mice, a target
breathing rate of 100 breaths per minute (bpm) leads to
motion-free imaging. Note that surgical stimulation (or a pain-
ful injection) may partially and temporarily reverse the anes-
thetic depression.
4. The usual supportive measures for survival experiments include
the protection of eyes with eye lubricant, the compensation of
dehydration by subcutaneous saline injection (approx
0.5. . .1 mL depending on length of imaging session) into
loose skin behind its neck.
5. Once the mouse has been anesthetized, move it towards the
coil setup and position animal such that the region of interest is
centered over the coil. Secure animal in place using
medical tape.
6. Remove the syringe from the catheter line, and connect the line
with contrast agent to it. This connection can be done with
stents, a butterfly needle or a regular needle tip.
7. Connect the saline flush to the end of the contrast-agent line.
Naturally, it is important to avoid the inclusion of air bubbles in
any of the liquid that will be injected into the animal.
3.3 Animal To perform reproducible DCE-MRI experiments, it is important to

Monitoring Equipment maintain a constant plane of anesthesia as well as body temperature.
To counter the effect of the anesthesia-induced, depressed breath-
ing, it is advisable to switch the breathing gas to oxygen to maintain
a sufficient tissue-oxygen saturation. When placing the animal with
its coil holder in the magnet the following support lines and tubes
need to be put in place:
1. Supply of isoflurane and inspiration air/oxygen and safe extrac-
tion of expired gas (also containing anesthetic).
2. A respiration pillow across the animals chest, secured in place to
collect a respiration trace.
3. A vaseline-capped temperature probe inserted rectally into the
animal.
4. A supply hose of warmed room air whose temperature is con-
trolled through a feedback loop with the internal body temper-
ature of the animal.
5. If used, a pulse oximeter is connected to either a thigh skin fold
or leg.
3.4 Data Acquisition DCE-MRI experiments acquire images before, during, and after
injection of a contrast agent. Most studies estimate the concentra-
tion of contrast agent in the tissue through the change in the T1
relaxation time constant, and using Eq. 1.
3.4.1 Pre-injection: Typically, a T1 map is acquired pre-injection. The T1 map can be

T1 Map estimated using a variety of methods, though an inversion recovery
or variable flip angle approach is most popular.
In an inversion recovery experiment, a 180∘ excitation RF pulse
is applied, followed by a 90∘ RF pulse after a known delay time
(referred to as the inversion time, or TI). The magnitude of the
signal is then fit to the equation:
M Z ¼ M 0 ð1 2 e T I =T 1 Þ ð4Þ
where MZ is the magnitude of the magnetization in the image, M0 is
the magnitude of the magnetization at excitation, TI is the inver-
sion time, and T1 is the relaxation time constant to be determined
via curve fitting.
Preclinical MRI 79
Alternatively, a variable flip angle FLASH experiment may be

performed using several flip angles. A minimum of two angles are
required, though measurements with more angles will improve the
T1 estimate. In a FLASH sequence the signal intensity, S, depends
on flip angle, θ, the baseline relaxation rate, T1,0:
1 e T R =T 1
S ¼ S 0 sin θ : ð5Þ
1 e T R =T 1 cos θ
The flip angle, θ, and repetition time, TR, are prescribed by the
experimenter. Alternatively, a spatially resolved flip-angle map may
be acquired, should the excitation coil have a non-uniform excita-
tion profile (i.e., B1 inhomogeneity).
Recently, faster methods have been applied to measuring base-
line T1-map such as the Look-locker technique [30].
3.4.2 DCE Image A typical DCE-MRI consists of pre-injection scans, the injection,
Acquisition and a long time course evaluation post-injection. It may be benefi-
cial to acquire a second T1 map after the full scan for a comparison.
A large number of MR sequences have been used to acquire
DCE-MRI data. A very common technique is a 3D FLASH
sequence (also known as spoiled gradient echo sequence) since it
allows for shorter repetition times, TR, by using small flip angles.
Parameters usually are TR ¼ 4. . .6 ms, TE ¼ 1. . .3 ms, flip angle, θ
¼ 10∘. . .15∘. A final determination on the ideal repetition time/flip
angle determination has to be made through the optimization of
signal and contrast to noise ratio for an expected baseline and
change of T1. The required spatial coverage, field of view, and
spatial resolution determine the final temporal resolution. A viable
alternative for certain geometries is the use of a multi-slice 2D
FLASH DCEMRI sequence. For those the repetition time, TR, is
lengthened to about 100 ms while the flip angle is increased to θ ¼
25∘. . .40∘ to continue acquiring signal in a signal-to-noise per unit
time optimized fashion.
While parallel imaging is very useful in many human DCE-MRI
studies to achieve acceleration factors of 2–3, this has not been
routinely done in pre-clinical DCE-MRI scans. When parallel imag-
ing has been applied in pre-clinical imaging it is often used over
surface coil due to need to extend homogeneous coverage of
patient anatomy rather than the acquisition speed up. An area of
active research is the use of compressed sensing, which can achieve
remarkable speed-up [31].
1. Acquire survey scan to plan positioning of DCE-MRI slices.
2. Acquire T1 baseline preferably at exactly the same location and
resolution as the subsequent DCE-MRI experiment. For a
variable-flip angle FLASH sequence, this is straightforward
due to the only difference arising from the flip angle and lack
of repeat acquisition.
3. Should there be a need for a flip-angle map this can be acquired

now or as part of the variable flip angle FLASH T1-mapping
sequence.
4. Acquire DCEMRI scan as described above. A fixed number
(5–10) of baseline DCE image frames before bolus injection
should be acquired. The bolus injection using the power injec-
tor should be triggered. To capture the majority of contrast
uptake and washout, the subsequent duration of the remaining
scan should be at least 15 min preferably longer.
5. As detailed in Note 4.3, injection success may be checked early
with a reconstruction of the partially acquired DCE data set. It
might be possible to stop an unsuccessful experiment at this
stage, save time and move to the next animal to be scanned.
6. Optionally a second (‘bookend’) T1 map may be acquired.
3.5 Arterial Input The AIF should be sampled for each individual and for each experi-
Function ment. However this may not be possible if a high temporal resolu-
tion is desired for the DCE images. One method is to estimate the
AIF from a population averaged curve using your planned proce-
dure, but in a separate experiment.
1. Identify a large vessel in the near vicinity of the tissue of
interest. Common sites are the heart, aorta/vena cava, iliac
artery, or the tail veins.
2. Set up the animal as detailed above, but position the animal
such that the site of the AIF measurement is centered above
the coil.
3. Run the desired scans for the AIF measurement. The total scan
time should be the same as the DCE experiment.
4. Repeat for several animals.
5. The AIF may be determined from the signal magnitude [32] or
phase [33– 36] in the images. Image phase has the advantages
that it has a directly linear relationship over a wide range of
concentrations, has greater SNR, it is less sensitive to flow, and
it is less sensitive to changes in the scan parameters.
6. The change in T1 can be used to determine the change in signal
for a magnitude based AIF.
7. To get the concentration from the image phase, a calibration
factor is required for your scanner. This may be attained from a
separate experiment using a large range of contrast agent con-
centrations and fitting a linear regression to the data. The phase
may need to be unwrapped first due to the 2π dynamic range. If
using the signal phase, a reference phantom should be placed
within the imaging plane to track phase drift [37].
8. Average the individual measurements for all animals.
Preclinical MRI 81
3.6 Data Processing Since the data processing requires a parameter fit of an data time
course, it is important to ensure that subsequent images are
registered with each other.
1. At a minimum, a visual inspection is required to check that all
slices line up as time advances. Slices with motion present
should be discarded. More robustly, it is possible to recover
data that might otherwise be lost to motion artifacts, by regis-
tering the DCE images to a suitable baseline image.
2. When signal intensities from separate scans are computationally
combined, the receiver gain used to acquire these different
scans has to be fixed. Many scanners will automatically adjust
the receiver gain and thereby introduce an unknown overall
intensity scaling factor. While scanner-reconstructed data
sometimes also stores information to undo the effect of a
variable receiver gain, it is advisable to fix the gain if possible.
3. Baseline T1 map is calculated and registered to the DCE-MRI
time course. The calculation of the baseline T1 may be aug-
mented with the calculated flip-angle map.
4. Using Eq. 5, the baseline T1 map, and the averaged signal prior
to contrast injection, the change in T1 for every pixel in the
tissue of interest can be calculated. The lower panel in Fig. 3
shows such two example time courses.
5. Using Eq. 1, one can calculate the concentration of the con-
trast agent for every time in the DCE-MRI time course. This
data forms the basis for semi-quantitative parameters and the
subsequent fit. Note 4.4 discusses the options of fitting signal
time courses vs. concentration time courses.
6. Simpler but less computationally involved, semi-quantitative
parameters such as area-under-the curve (AUC) may be calcu-
lated. There is now widespread consensus that parameters such
as AUC have limitations when it comes to translation and inter-
group comparisons.
7. Data quality, temporal resolution, and signal-to-noise will dic-
tate the choice of model and the number of free parameters
that can be supported by the data. Rukat et al. studied the
effects of these factors and provide a methodology for deciding
on a suitable model [38].
8. A suitable arterial input function, ca(t) is chosen.
9. Using the Tofts model as an example, the observed concentra-
tion data can be fit using Eq. 2. As a result, parameters
Ktrans and ve can be determined. See example in Fig. 3. For
different models (such as extended Tofts or 2CXM) other/
more parameters will result from this step. See Note 4.5 for
details on an additional parameter originating from the uncer-
tainty of the bolus arrival in the tissue.
Ktrans ve R2
0.00 0.02 0.04 0.06 0.08 0.10 0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.0 0.2 0.4 0.6 0.8 1.0
concentration time course and fit T1 − weighted image concentration time course and fit
0.25 0.25
0.20 0.20
0.15 Ktrans = 0.08 ve = 7% 0.15 Ktrans = 0.04 ve = 25%
0.10 0.10
0.05 0.05
0.00 0.00
0 100 200 300 0 100 200 300
Fig. 3 Parameter map and typical concentration time course and Tofts model fit. Top row shows the parameter
maps for Ktrans and ve. Top right shows R2 as a measure of fit quality. While reservations exist for the use of R2
for assessing different models in non-linear regression, it is useful to judge fit quality for one model applied to
different time curves. The lower row shows two different for a quickly enhancing pixel (left) and a slower
enhancing one (right)
10. Aggregate statistics of these parameter values in regions of

interest (e.g. tumor/tumor core/tumor periphery, etc.) can
be calculated and used for further analysis.
4 Notes
4.1 Reference DCE-MRI models are based on the indicator-dilution theory, and
Region as Alternative therefore require knowledge of the concentration-time curve (the
to Direct AIF AIF) in the blood plasma. It has been suggested that the AIF should
Acquisition be measured with a temporal resolution exceeding 5 s/image for
accurate results [39]. However, this can be difficult to achieve. To
relax the time restraint, Yankeelov et al. proposed a technique that
replaces the AIF with a calibration against another tissue with known
contrast kinetics; similar to what is used in PET imaging [25].
The technique is derived in [25], providing a final model of:

K T rans , RR K T rans , T OI
C T OI ðT Þ ¼ R C RR ðT Þ þ R
ve , RR v e , T OI
ZT
K trans , T OI
C RR ðtÞ exp ∗ðT tÞdt
ve , T OI
0
Preclinical MRI 83
where CTOI and CRR are the contrast agent concentrations in the
tissue of interest and the reference region tissue, KTrans,TOI and
KTrans,RR are the volume transfer coefficients, ve, TOI and ve, RR are
the fractional volume of the extravascular-extracellular space, and
R ¼ KTrans,TOI/KTrans,RR. This technique assumes that the AIF is
similar for both the reference tissue and the tissue of interest [39].
Procedure for model fitting with the Reference Region
model [25].
1. Acquire a T1o map over the tissue of interest. A variable flip
angle spoiled gradient echo may be used with a number of
different excitation angles. Fit the data to Eq. 5 to solve for T1.
2. Perform the DCE experiment as proposed. Allow for several
pre-bolus images to be acquired.
3. Co-register all images against the first pre-contrast image. A
mutual information-based rigid registration algorithm is
suggested.
4. Estimate the T1 time courses for each voxel in both the refer-
ence region tissue and the tissue of interest from the change in
signal intensity of the time course data.
5. Input both T1 curves into the proposed model described above
into a curve-fitting routine.
4.2 Two- Through improvements in hardware and hence temporal resolu-

Compartment tion, DCE-MRI data these days is often not sufficiently well
Exchange Model described by the Tofts model. The limitations of the Tofts model
are now well understood [12] and can be avoided through the
acquisition of high-temporal resolution DCE-MRI and vascular
input function and the application of the newer second-generation
models, such as the one initially applied by Brix et al. [40]. Of those
models, the two-compartment exchange model (2CXM) has
quickly gained widespread adoption and provides a fourth parame-
ter which characterizes permeability of the vascular compartment
boundary: the permeability surface area product, PS. Sourbron
et al. discuss the relationship of the parameters from these different
models in their review [41].
The two-compartment exchange model treats the plasma and
the interstitium as single compartments each. Mass conservation of
the interstitium is ensured by:
dc e
ve ðtÞ ¼ PSc p ðtÞ PSc e ðtÞ: ð6Þ
dt
Correspondingly, mass conservation in the vascular compartment
has to balance the influx of tracer from the input function, ca(t)
with the exchange to and from the interstitium and efflux of tracer
from the vascular compartment itself:
dc p
vp ðtÞ ¼ F p c a ðtÞ F p c p ðtÞ þ PSc e ðtÞ PSc p ðtÞ: ð7Þ
dt
The response function in a two-compartment model is a
double-exponential with four free parameters (two amplitudes
and two decay constants).
RðtÞ ¼ A 1 e B1 t þ A 2 e B2 t ð8Þ

These constants, Ai and Bi, map to the physiologically more
meaningful parameters using the following equations:
τ 1 Fp
A 1=2 ¼ F p and B 1=2 ¼ ð9Þ
τþ τ ðv p þ ve Þτ
where
ve
e¼ and τ
vp þ ve
rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi!
E Ee þ e 1 4Eeð1 EÞð1 eÞ
¼ 1 : ð10Þ
2E E Ee þ e
4.3 Predicting Since a considerable cost in time and effort is incurred with
Injection Success DCE-MRI experiments it is desirable to have an early indication
While Scanning of a successful bonus injected. At our institutions we have devel-
oped an early alert should there not have been a successful injec-
tion—this early alert can be run as soon as the contrast is injected.
Using signal from the entire field of view and reconstructing a
“global” time course of signal change. Figure 4 shows an example
of a successful injection and unsuccessful injection. The latter
allows the immediate discontinuation of the experiment thereby
saving precious scan time. This diagnostic early reconstruction
requires a tolerance to partially written data files.
4.4 Fitting The methods described in this chapter advocate the calculation of
Concentration or the concentration time curves from the DCE-MRI data and the
Signal-Time Curves subsequent fitting of this curved C(t) by a model of choice. While
this is a more obvious approach, an alternative method involves the
calculation of the expected signal time curve from a concentration
time curve C(t) in terms of the still undetermined parameters of the
model. This will form a model of the expected signal time curve, S
(t), and can be fit by the data acquired.
In principle, these two methods are equivalent. However, the
experimenter needs to take care to adjust the weighting of data
points as dictated by the transformation from C(t) to S(t). Further-
more since this transformation is not linear, the cost function used
in the minimization will result in slightly different optimization
results that are usually inconsequential in practice.
Preclinical MRI 85
2650
2800
2600
2600
2550
signal intensity / a.u.
signal intensity / a.u.

2400
2500
2200
2450
2000 2400
1800 2350
1600 2300
1400 2250
0 100 200 300 0 50 100 150 200 250
image frame number image frame number
Fig. 4 Early image showing a successful injection (left) and failed injection (right). Note the vastly different
scale in image intensity between the two. At the time of injection, motion-induced change in intensity is
sometimes visible but clear enhancement is lacking
4.5 Bolus In fitting the pharmacokinetic models such as by the Tofts equa-
Arrival Time tion (Eq. 2), there is an implicit assumption of the convolution
starting from time t ¼ 0 extending into the feature. In reality this
time zero is the bolus arrival time and needs to be determined
separately. A simple method is to set the bolus arrival time to the
time when the injection was initiated. A more sophisticated
approach is to leave this bolus arrival time flexible. It can then be
fit as part of the model optimization or it can be determined directly
from the data using control chart decision criteria [42]. In either
case, the correct choice of bolus arrival time has a strong influence
on pharmacokinetic parameters.
Acknowledgements
This work was made possible through support from the Discovery
Program of NSERC and CIHR.
References
1. O’Connor JPB, Aboagye EO, Adams JE, et al 3. Keara Boss M, Muradyan N, Thrall DE (2011)
(2016) Imaging biomarker roadmap for cancer DCE-MRI: a review and applications in veteri-
studies. Nat Rev Clin Oncol 14:169–186. doi: nary oncology. Vet Comp Oncol 11(2):87–100
10.1038/nrclinonc.2016.162 4. Zhou Z, Lu ZR (2013) Gadolinium-based
2. Weinmann HJ, Brasch RC, Press WR, et al contrast agents for magnetic resonance cancer
(1984) Characteristics of gadolinium-DTPA imaging. Wiley Interdiscip Rev Nanomed
complex: a potential NMR contrast agent. Am Nanobiotechnol 5(1):1–18. doi: 10.1002/
J Roentgenol 142:619–624 wnan.1198
5. Yan GP, Robinson L, Hogg P (2007) Magnetic function can be achieved by phase imaging.
resonance imaging contrast agents: overview Magn Reson Med 71(2):544–550. doi: 10.
and perspectives. Radiol 13:e5–e19. doi: 10. 1002/mrm.24682
1016/j.radi.2006.07.005 17. Pathak AP, Artemov D, Bhujwalla ZM (2004)
6. Carr DH, Brown J, Bydder GM, et al (1984) Novel system for determining contrast agent
Intravenous chelated gadolinium as a contrast concentration in mouse blood in vivo. Magn
agent in NMR imaging of cerebral tumours. Reson Med 51:612–615. doi: 10.1002/mrm.
Lancet 323(8375):484–486. doi: 10.1016/ 10727
S0140-6736(84)92852-6 18. Henderson E, Rutt BK, Lee TY (1998) Tem-
7. Shen Y, Goerner FL, Snyder C, et al (2015) T1 poral sampling requirements for the tracer
relaxivities of gadolinium-based magnetic reso- kinetics modeling of breast disease. Magn
nance contrast agents in human whole blood at Reson Imaging 16(9):1057–1073. doi: 10.
1.5, 3, and 7 T. Invest Radiol 50(5):330–338. 1016/S0730-725X(98)00130-1
doi: 10.1097/RLI.0000000000000132 19. Lyng H, Dahle GA, Kaalhus O, et al (1998)
8. Tofts P (2010) T1-weighted DCE imaging Measurement of perfusion rate in human mel-
concepts: modelling, acquisition and analysis. anoma xenografts by contrast-enhanced mag-
MAGNETOM Flash 3:30–39 netic resonance imaging. Magn Reson Med
9. Tofts PS, Kermode AG (1991) Measurement 40:89–98
of the blood-brain barrier permeability and 20. Parker GJM, Roberts C, Macdonald A, et al
leakage space using dynamic MR imaging. (2006) Experimentally-derived functional
1. Fundamental concepts. Magn Reson Med form for a population-averaged high-temporal-
17(2):357–367 resolution arterial input function for dynamic
10. Larsson HBW, Stubgaard M, Frederiksen JL, contrast-enhanced MRI. Magn Reson Med 56
et al (1990) Quantitation of blood-brain bar- (5):993–1000. doi: 10.1002/mrm.21066
rier defect by magnetic resonance imaging and 21. Sanz-Requena R, Prats-Montalban J, Marti-
gadolinium-DTPA in patients with multiple Bonmati L, et al (2015) Automatic individual
sclerosis and brain tumors. Magn Reson Med arterial input functions calculated from PCA
16(1):117–131. doi: 10.1002/mrm. outperform manual and population-averaged
1910160111 approaches for the pharmacokinetic modeling
11. Brix G, Semmler W, Port R, et al (1991) Phar- of DCE-MR images. J Magn Reson Imaging
macokinetic parameters in CNS Gd-DTPA 42(2):477–487. doi: 10.1002/jmri.24805
enhanced MR imaging. J Comput Assist 22. Onxley JD, Yoo DS, Muradyan N, et al (2014)
Tomogr 15(4):621–628. doi: 10.1097/ Comprehensive population-averaged arterial
00004728-199107000-00018 input function for dynamic contrast-enhanced
12. Sourbron SP, Buckley DL (2011) On the scope vmagnetic resonance imaging of head and neck
and interpretation of the Tofts models for cancer. Int J Radiat Oncol 89(3):658–665. doi:
DCE-MRI. Magn Reson Imaging 66 10.1016/j.ijrobp.2014.03.006
(3):735–745. doi: 10.1002/mrm.22861 23. Kershaw LE, Cheng MHL (2011) A general
13. McIntyre DJO, Ludwig C, Pasan A, et al dual-bolus approach for quantitative
(2004) A method for interleaved acquisition DCE-MRI. Magn Reson Imaging 29
of a vascular input function for dynamic (2):160–166. doi: 10.1016/j.mri.2010.08.
contrast-enhanced MRI in experimental rat 009
tumours. NMR Biomed 17(3):132–143. doi: 24. Kim TH, Pack NA, Chen L, et al (2010) Quan-
10.1002/nbm.868 tification of myocardial perfusion using CMR
14. Pickup S, Zhou R, Glickson J (2003) MRI with a radial data acquisition: comparison with
estimation of the arterial input function in a dual-bolus method. J Cardiovasc Magn
mice1. Acad Radiol 10:963–968. doi: 10. Reson 12(1):45–52. doi: 10.1186/1532-
1016/S1076-6332(03)00291-5 429X-12-45
15. Ragan DK, Lai SY, Bankson JA (2011) Fast, 25. Yankeelov TE, Luci JJ, Lepage M, et al (2005)
reproducible measurement of the vascular Quantitative pharmacokinetic analysis of
input function in mice using constrained DCE-MRI data without an arterial input func-
reconstruction and cardiac sampling. NMR tion: a reference region model. Magn Reson
Biomed 24:373–384. doi: 10.1002/nbm. Imaging 23(4):519–529. doi: 10.1016/j.mri.
1601 2005.02.013
16. Fruytier AC, Magat J, Colliez F, et al (2014) 26. Beuzit L, Eliat PA, Brun V, et al (2015)
Dynamic contrast-enhanced MRI in mice at Dynamic contrast-enhanced MRI: study of
high field: estimation of the arterial input inter-software accuracy and reproducibility
Preclinical MRI 87
using simulated and clinical data. J Magn for quantification of dynamic contrast-
Reson Imaging 43(6):1288–300. doi: 10. enhanced (DCE) MRI and comparison with
1002/jmri.25101 DCE-CT. Magn Reson Med 66
27. van der Walt S, Colbert SC, Varoquaux G (5):1267–1274. doi: 10.1002/mrm.22905
(2011) The NumPy array: a structure for effi- 36. Garpebring A, Wirestam R, Yu J, et al (2011)
cient numerical computation. Comput Sci Eng Phase-based arterial input functions in humans
13(2):22–30. doi: 10.1109/MCSE.2011.37 applied to dynamic contrast-enhanced MRI:
28. McKinney W (2010) Data structures for statis- potential usefulness and limitations. Magn
tical computing in Python. In: van der Walt S, Reson Mater Phy 24(4):233–245. doi: 10.
Millman J (eds) Proceedings of the 9th Python 1007/s10334-011-0257-8
in science conference, pp 51–56 37. Akbudak E, Norberg RE, Conturo TE (1997)
29. Brett M, Hanke M, Cipollini B, et al (2016) Contrast-agent phase effects: an experimental
nibabel: 2.1.0. doi: 10.5281/zenodo.60808 system for analysis of susceptibility, concentra-
30. Look DC, Locker DR (1970) Time saving in tion, and bolus input function kinetics. Magn
measurement of NMR and EPR relaxation Reson Med 38:990–1002
times. Rev Sci Instrum 41(2):250–251. doi: 38. Rukat T, Walker-Samuel S, Reinsberg SA
10.1063/1.1684482 (2015) Dynamic contrast-enhanced MRI in
31. Han S, Paulsen JL, Zhu G, et al (2012) Tem- mice: an investigation of model parameter
poral/spatial resolution improvement of uncertainties. Magn Reson Med 73
in vivo DCE-MRI with compressed sensing- (5):1979–1987. doi: 10.1002/mrm.25319
optimized FLASH. Magn Reson Imaging 30 39. Cárdenas-Rodrı́guez J, Howison CM, Pagel
(6):741–752. doi: 10.1016/j.mri.2012.02. MD (2013) A linear algorithm of the reference
001 region model for DCE-MRI is robust and
32. McGrath DM, Bradley DP, Tessier JL, et al relaxes requirements for temporal resolution.
(2009) Comparison of model-based arterial Magn Reson Imaging 31(4):497–507. doi:
input functions for dynamic contrast-enhanced 10.1016/j.mri.2012.10.008
MRI in tumor bearing rats. Magn Reson Med 40. Brix G, Kiessling F, Lucht R, et al (2004)
61:1173–1184. doi: 10.1002/mrm.21959 Microcirculation and microvasculature in
33. Moroz J, Wong CL, Yung AC, et al (2014) breast tumors: pharmacokinetic analysis of
Rapid measurement of arterial input function dynamic MR image series. Magn Reson Med
in mouse tail from projection phases. Magn 52(2):420–429. doi: 10.1002/mrm.20161
Reson Med 71:238–245. doi: 10.1002/mrm. 41. Sourbron SP, Buckley DL (2013) Classic mod-
24660 els for dynamic contrast-enhanced MRI. NMR
34. Bleeker EJW, van Buchem MA, Webb AG, et al Biomed 26(8):1004–1027. doi: 10.1002/
(2010) Phase-based arterial input function nbm.2940
measurements for dynamic susceptibility con- 42. McPhee KC, Baker JEH, Saatchi K, et al
trast MRI. Magn Reson Med 64:358–368. doi: (2012) Delayed bolus arrival time with high
10.1002/mrm.22420 molecular weight contrast agent indicates sig-
35. Korporaal JG, van den Berg CAT, van Osch nificant necrosis. In: Proceedings of the inter-
MJP, et al (2011) Phase-based arterial input national society for magnetic resonance in
function measurements in the femoral arteries medicine, Melbourne, p. 1528
Chapter 6
Diffusion-Weighted Magnetic Resonance Imaging

Irene Guadilla, Daniel Calle, and Pilar López-Larrubia
Abstract
Magnetic resonance imaging (MRI) is a technique based on the contents and relaxation features of water in
tissues. In basic MRI sequences, diffusion phenomenon of water molecules is not taken into account
although it has a notable influence in the relaxation times, and therefore in the signal intensity of images. In
fact, MRI techniques that take advantage of water diffusion have experienced a huge development in last
years. Diffusion-weighted imaging (DWI) has spectacularly evolved reaching nowadays a great impact both
in clinical and preclinical imaging—especially in the neuroimaging field—and in basic research. We present
here a protocol to perform DWI studies in a high-field preclinical setup.
Key words Magnetic resonance imaging, Preclinical MRI, Diffusion-weighted imaging, Echo-planar
imaging, Apparent diffusion coefficient
1 Introduction
Diffusion is a physical phenomenon that leads to a net transport of a

molecule in a liquid—or a gas—from one site to another in a
thermal random motion [1, 2]. The molecular diffusion processes
in tissues happen at the micron scale and currently can be studied
with magnetic resonance (MR) in a precise and noninvasive way
almost in real time [3, 4]. The signal intensity in a pixel of a MR
image depends mainly on the spin density (NH) and the longitudi-
nal (T1) and transversal relaxation times (T2, T2*) of the set of
hydrogen atoms from the corresponding element of volume (voxel)
in the object imaged [5]. Another important contrast mechanism
that entails a net loss of signal in the image is that caused by the spin
dephasing as a consequence of coherent and incoherent movements
that specially shorten T2 and T2* values. In tissues, the random
diffusion due to the thermal energy carried by the water molecules
(Brownian diffusion) in areas of variable magnetic field intensity
leads to random displacements in the phase of the spins and a lower
intensity of the magnetic resonance imaging (MRI) signal [6]. The
velocity observed and the direction of water diffusion are a
89
90 Irene Guadilla et al.
reflection of the molecular barriers or impediment of any nature

that hydrogens find across their translation motion. Water mole-
cules in tissues diffuses between 5 and 15 μm in 40 ms, approxi-
mately the half than in pure water [7], and as the cell diameter in
most of the tissues is included in this spatial range, the possibility of
obtaining images of the in vivo proton translation offers new chan-
nels of visualizing the intra- and intercellular dynamic. In this sense,
the distance that a water molecule transverses in a specific direction
(L) is given by Eq. 1 [8]:
L 2 ¼ 2 Δt ADC ð1Þ
where the proton displacement is described by an apparent diffu-

sion coefficient (ADC) for an interval of time Δt. The ADC in a
specific tissue is slow if the water molecules motion is hindered by
the presence of cell membranes, wall, and/or macromolecules.
Besides, these barriers are able not only to generate changes in
the velocity diffusion, but also in the direction. For example, the
myelin sheaths wrapped around the nerve axon induce in the water
molecules a favorite direction of movement through the lower
resistance path, that is, along the fibers and not across them
[9]. The fact that water diffuses faster in some specific not hindered
directions induces anisotropy in the displacement making different
the ADC values according to the direction in which they are
measured. Even more, the ability of analyzing the diffusion coeffi-
cient depends on the time of observation. When this is too short,
maybe the displacement of the molecules is not long enough to
experience restrictions in their movement. When measures are
made employing high Δt, ADCs mean better the impediments to
the diffusion.
1.1 Methodology The development of the diffusion-weighted imaging (DWI) relies

of Diffusion-Weighted on the availability of fast and ultra-fast acquisition techniques sen-
Imaging sitive to T1, T2, and T2* relaxation times, and this is associated to
the great advances of hardware in the scanners, mainly in the
improvement of the magnetic field gradient designs [10]. On the
other side, the ability of obtaining relevant and precise information
from these imaging studies also depends on the development of
adequate mathematical models and signal fitting algorithms to
those models. The shortening of T2 values due to diffusion phe-
nomenon in pure liquids like water disturbs the magnetic resonance
signal. The observed fact that diffusion could be detectable as a
reduction of the spin-echo (SE) signal [11], stimulated the devel-
opment of numerous MR methods able to measure the diffusion
coefficient in liquids [12, 13]. The most commonly employed one,
described by Stejskal and Tanner in 1965 [11], relies on the appli-
cation—to a SE sequence—of a couple of square pulsed field gra-
dients with a value of magnitude G, a duration of δ, and a delay
between them of Δ, being one gradient opposite to the other. The
Diffusion MRI 91
variation of the echo signal intensity generated in this case can be

described by the next equation:

S S 0 exp γ 2 G 2 δ2 ðΔ δ=3Þ ADC ð2Þ
where γ is the proton gyromagnetic ratio, S0 the signal at a zero

value of the diffusion gradient strength, and the diffusion time is
given by Δ δ/3 (obviously, the signal intensity also depends on
the intrinsic values of T1 and T2, and the repetition (TR) and echo
time (TE) selected by the operator). By agreement, the b factor
comprises the terms γ 2G2δ2(Δ δ/3) [14] leading to Eq. 3 as
follows:
S S 0 exp½b ADC ð3Þ
The Stejskal-Tanner spin-echo (ST-SE) sequence is schema-

tized in the diagram of Fig. 1, where the two diffusion gradients
are applied before and after the 180 pulse.
The major advantage of using gradient pulses is that the strong
pulses exceed the local magnetic field inhomogeneities that in vivo
are usually present in the tissues. The ST-SE sequence works firstly
dephasing and then rephasing the spins. Those water molecules
that move during diffusion time will experience a phase-shift pro-
portional to their velocity and acquire a phase that does not match
the phase of stationary hydrogens. This phase-shift is a consequence
of spin diffusion motion across regions of different magnetic field.
So, diffusion leads to an incomplete rephasing of the spins attenu-
ating the echo signal. The great advantage of this sequence is the
Fig. 1 Diffusion-weighted spin-echo Stejskal-Tanner sequence. Diffusion

sensitive can be modulated by changing gradient pulse parameters—δ:
gradient pulse duration, Δ: gradient pulse separation, G: gradient strength
possibility of controlling with accuracy the time frame (Δ δ/3)

sensitive to diffusion movements. Besides, considering the obser-
vation time and the ADC measured, the distance moved can also be
determined with 1 allowing to assess the restrictions to diffusion
[15]. The direction of diffusion gradient application can be selected
in such a way that the anisotropy of the water molecule motion can
also be evaluated by determining the ADC values in different
directions [16–18].
The most direct method of measuring the diffusion coefficient
with this technique relies on carrying out several experiments with
fixed values of δ, Δ, TR, and TE but modifying the gradient
strength from one to other, which entails variations of b values.
Afterwards, ADC can be calculated by using linear regression anal-
ysis of the natural logarithm of mean signal intensity versus the
diffusion factor b according to Eq. 4.
ln ½S ðb Þ= ln ½S ð0Þ ¼ k ADC b ð4Þ
A simplified representation of the workflow steps in obtaining

an ADC map that compiles in each pixel the diffusion coefficient of
the corresponding voxel in the tissue is shown in Fig. 2.
Fig. 2 Representative DWI workflow. Several images are acquired increasing b value from one to the next. A
pixel-by-pixel adjustment of the logarithm of the MRI signal (related to the basal signal) versus b values,
provides the ADC values from the slope of the line. A voxel placed in cerebrospinal fluid (CSF)—where proton
can freely move across—diffuse faster originating a higher slope in the linear adjustment than those placed in
gray matter (GM)—where water molecules find restrictions to their translational movement
Diffusion MRI 93
Briefly, the procedure involves the acquisition of consecutive

images increasing the diffusion weighting—that is, b value—
obtaining always a basal image (b ¼ 0). The ADC map will be
computed from the slopes of every pixel fitting in all images. In
terms of the numbers of b values employed, the more, the better,
although this obviously extends the acquisition time. So, it is
necessary to find the tradeoff between the accuracy in the measure-
ment and the requirement of shortening the study to minimize
artifacts and noise due to the time that the subject takes inside the
equipment (involuntary movements, signal fluctuations, interfer-
ences). It is important to realize that differences between fast and
slow diffusion movements are better distinguished at high b values,
even at the cost of losing signal-to-noise ratio.
2 Materials
The following materials are used in diffusion imaging studies:
2.1 MRI System 1. MR high-field horizontal magnet.

3. RF coils. Two basic MRI coil setups can be used.
(a) Transceiver coil setup: The excitation and reception of the
MR signal are performed with the same volume coil of
around 35–40 mm inner diameter for rats and 25–30
inner diameter for mice.
(b) Active decoupled transmit and receive coil setup: Excita-
tion and reception of the signal are typically made by a
volume transmit coil and a surface coil brain rat and/or
mouse dedicated. In that case, coils must be actively
detuned from each other.
2.2 Other Equipment, It is necessary to use a monitoring system that continuously reports
Devices, and Materials data of body temperature (rectal sensor) and respiratory rate (using
a pneumatic pillow placed in the animal abdomen) to follow the
2.2.1 Small Animal
physiological evolution of the anesthetized animal inside the mag-
Monitoring and Gating
net and control the level of anesthesia.
System
2.2.2 Anesthesia Device The inhalation anesthesia equipment used for small animals con-
tains an anesthetic vaporizer (dependent on the anesthetic), a flow-
meter (dependent on the gas carrier), a poly(methylmethacrylate)
(PMMA) chamber, circulating tubes, and a filter.
2.2.3 Thermal System Anesthetized animals have to be kept warmed by using a water
heated blanket (or any other appropriate and MR compatible
tool) to maintain the body temperature at 37 C during scanning.
2.2.4 Animal Holders Suitable animal beds of PMMA (or any other MR compatible
material) are used to place the rat or mouse inside the magnet,
provided with cap nose to supply anesthesia and adequate options
to immobilize the animal head (usually fixing tooth and ears with
bars). The holder will be slid to localize the region of interest in the
magnet isocenter.
2.3 Software 1. Acquisition. Commercial software, usually supplied by the

same distributor of the MR magnet.
2. Post-processing. Images usually can be processed with the
software supplied by the MR magnet distributor. Even so, for
a tailored processing of the images, the computation can be
done using some commercial software as Matlab (The Math-
Works Inc., Natick, MA, 2000) and IDL (Iterative Data Lan-
guage, Research System, Boulder, CO, USA), or with open
source software as ImageJ (https://imagej.nih.gov/ij/) [19]
or R (https://www.r-project.org) [20].
Table 1.
3 Methods
The following method presented to obtain diffusion-weighted

images is optimized for a 7T Bruker Biospect® MRI system
operating under Paravision 5.1 software (in a Linux environment)
and with a Bruker AVANCE III console.
3.1 Pre-scan Animals are anesthetized in a PMMA chamber using a mix of pure
oxygen and isoflurane (3–4%) and then placed on the animal holder
to be introduced inside the magnet. The anesthesia is maintained
during the whole acquisition and controlled based on the animal
respiration (see Note 1). The body temperature is maintained at
around 37 C—using the thermal blanket—and monitored with a
rectal thermometer. The animal head has to be immobilized during
acquisition because movements will cause artifacts and originate
processing errors in the calculated parametric images. To that,
immobilize it with ears and tooth bars and/or fix tape over the
head. Put eye ointment on the eyes of the animal to avoid damaging
the cornea.
3.2 Scan A fast-low angle shot magnetic resonance imaging (FLASH MRI)
sequence is used to obtain quickly an image of the three spatial
3.2.1 Localization
orientations (axial, coronal, and sagittal) of the animal and fix the
and Anatomical Imaging
correct position inside the magnet (see Note 2).
Diffusion MRI 95
Table 1
Preclinical MRI system and associated equipment/devices for developing the protocols described in
this chapter
MRI system Bruker Biospect® 7 T—16 cm bore (B-C 70/16)

Acquisition software Paravision 5.1 (Bruker Biospin, Ettlingen, Germany) running in a console Bruker
AVANCE III operated on a Linux platform
1
Rat-head transceiver H circular polarized transmit/receive birdcage volume coil, 40 mm inner
coil diameter
1
Mouse-head H circular polarized transmit/receive birdcage volume coil, 23 mm inner
transceiver coil diameter
1
Transmitter coil H circular polarized transmit-only volume coil, 72 mm inner diameter
1
Rat head surface coil H circular polarized with integrated combiner and preamplifier receive-only coil,
40 mm inner diameter
1
Mouse head surface H circular polarized with integrated combiner and preamplifier receive-only coil,
coil 23 mm inner diameter
Physiological Small Animal Monitoring and Gating System (Model 1025, SA Instruments, Inc.
monitoring NY)
Anesthesia Isoflurane (IsoFlo, Esteve labs, Barcelona, Spain) in O2 or air (standard calibrated
vaporizer)
Halogenated gas Activated charcoal canister packed (Harvard Apparatus Anaesthetic, Holliston,
filter MA, USA)
Thermal blanket Warmed circulating water blanket (Gaymar T/PUMP TP500 Heat Therapy
System)
Processing software MyMapAnalyzer, homemade software application written in Matlab (The
MathWorks Inc., Natick, MA, 2000)
After that, a T2-weighted sequence is acquired to obtain struc-

tural images of the animal anatomy to be used as a reference for the
diffusion-weighted parametric images.
3.2.2 Diffusion Imaging There are many options to acquire DWI, but acquisitions based on
the Stejskal-Tanner sequence are probably the most commonly
used. In these studies, as many ADC maps will be generated as
different directions are selected by the operator to apply the diffu-
sion gradients.
We present in this manuscript three different methods opti-
mized in for obtaining diffusion images depending on how the
k-space is filled, that it, the collection of the MR data.
1. Standard acquisition. A spin echo (SE) sequence is acquired in

such a way that each line in the k-space (echo signal) is filled at
each phase encoding step (Fig. 1). This sequence has a very
good signal-to-noise ratio (SNR) but suffers of a long acquisi-
tion time (see Note 3).
2. Single-shot echo-planar imaging (EPI). In that case, the DWI
is acquired in such a way that each image acquisition takes the
TR value because the whole k-space is filled after the first
90 –180 pulse set in the SE sequence. This protocol has the
advantage of being much faster than a standard acquisition but
the disadvantage that it is very sensitive to the field homogene-
ities and also suffers of a lower signal-to-noise ratio.
3. Segmented EPI. The mechanism of this sequence is similar to
the previous one but completing the k-space in different shots
or segments, that is, several lines of the k-space are filled after
each 90 –180 pulse combination. This sequence is slower
than the single-shot EPI—as many times as segments
selected—but has the advantage of a better SNR (see Note 4).
In this case, it is also mandatory to get a good shimming. All
diffusion protocols with EPI acquisition, single or multi shot,
in Bruker systems are called DtiEpi in the software.
EPI DWI acquisitions, either single or multi-shot (segmented),
require a high homogeneity of the magnetic field, being necessary
to adequately shim the system before starting the study. In Bruker
equipment there are several automatic options to achieve this:
1. Adjusting the first and second shim coil values to optimize the
shimming of the whole sample covered by the coil and making
the necessary adjustment to the signal response.
2. Using the FASTMAP tool implemented in Bruker scanners
[21]. FASTMAP employs a localized shimming in a selected
cubic region which is virtually divided in smaller sticks where
shimming is improved step by step. This method is very effi-
cient, but has the problem that not always the volume of
interest can be covered by a cubic space.
3. Employing the MAPSHIM protocol in Bruker instruments. It
is a procedure that optimizes the magnetic field homogeneity
after the acquisition of a 3D magnetic field distribution. This
method allows the shimming in a rectangular region of arbi-
trary size and orientation [22].
Once the shimming is optimized, it is also mandatory to adjust
the basic frequency of the system right before acquiring the scan to
minimize the artifacts associated to the echo-planar acquisition.
Then, DWI sequence is acquired after selecting the b-values that
will control the diffusion weighting in the study (see Note 5).
Diffusion MRI 97
Fig. 3 MRI study of a mouse brain in axial orientation. T2-weighted (A) and diffusion-weighted images of the
same slice (B). The acquisition parameters of the DWI sequence are showed in Table 2.
Table 2
Acquisition parameters of a diffusion-weighted MRI study of a mouse brain in a Bruker Biospect® 7 T
MRI system
Pulse sequence Spin Echo (segmented EPI acquisition)

TE (echo time) 31.11 ms
TR (repetition time) 2500 ms
Averages 3
Segments 4
Repetitions 1
δ 3 ms
Δ 20 ms
b values 150, 400, 1000 s/mm2
Diffusion gradient strength (percentage) 31.1, 50.8, 80.3%
In-plane resolution 0.172 mm/pixel
Matrix resolution 128 128
Slice thickness 1.5 mm (axial orientation)
Number of diffusion directions 3 orthogonal directions (1, 0, 0), (0, 1, 0), (0, 0, 1)
Figure 3 shows representative T2 and DW images obtained in a

DWI study. Table 2 shows the parameters of a common diffusion
sequence (see Note 6).
3.3 Post-scan Diffusion studies are computed in a pixel-by-pixel base using the
Analysis appropriate software. The corresponding ADC parametric images
are generated in each direction, and also the mean ADC map in all
assessed directions are generated by fitting the MRI signal to the
Eq. 4. In our case, we employ the homemade application
Fig. 4 Diffusion MRI study of a mouse in axial orientation. ADC parametric images in the head-foot (HF) (A),
left–right (L–R) (B), anterior–posterior (A–P) (C) directions and mean values of ADC considering the three
directions (D). The acquisition parameters of the DWI sequence are showed in Table 2.
MyMapAnalyzer [23] developed in MatLab (the MathWorks Inc.,

Natick, MA, 2000). Figure 4 shows ADC maps of a three-
orthogonal direction DWI study obtained after computational
processing.
After generating the parametric images, the user can directly
measure the ADC of every pixel or a mean value by selecting a
region of interest (ROI) using an image processing software as
ImageJ (U. S. National Institutes of Health, Bethesda, Maryland,
USA) [19]. Finally, data can then be statistically analyzed with
appropriated software like SPSS (IBM, Armonk, North Castle,
New York, USA), Excel (Microsoft, Redmond, Washington,
USA), or R [20].
4 Notes
1. It is important to maintain a correct level of anesthesia in the

animal during the MRI acquisition. For that several important
considerations have to be taken into account:
(a) The pneumatic pillow from the monitoring system has to
be placed in a correct position (under the abdomen of the
animal) to allow measuring the respiratory rate accurately.
Diffusion MRI 99
(b) All the anesthetic carrier tubes have to be properly

connected to the anesthetic system, animal holder/nose
mask, and filter.
(c) The nose of the animal has to be well positioned into the
mask to allow the exchange of the respiratory gases avoid-
ing a dangerous accumulation of CO2.
(d) The filter system has to be switch on during the experi-
ment to absorb the non-metabolized isoflurane gas.
(e) The level of isoflurane dissolved in O2 has to be controlled
according to the animal physiological status, usually
around 1.5–2% isoflurane at a flow of 1 L/min.
(f) A respiration rate around 40–60 breaths/min is adequate
enough to ensure that the animal will neither take a risk
nor move during the study.
2. Prior the first scan, an automatic or manual adjustment of some
parameters has to be achieved:
(a) Tuning and matching the coil—when it is possible.
(b) Shimming the magnet to improve its homogeneity.
(c) Determining the RF power of the 90 excitation pulse.
(d) Adjusting the basic frequency of the system.
(e) Optimizing the receiver gain value.
3. In practice, this protocol is only recommended for phantom
studies due to the long that it takes that increases the possibility
of having artifacts due to the movement of the animal. Never-
theless, the total acquisition time can be reduced by using some
options to accelerate the phase encoding, like:
(a) Partial Fourier-transform (FT) acceleration is a factor that
accelerates the experiment through an asymmetric trun-
cation of the k-space. Its value goes from 1 (a full sym-
metric filling of the k-space) to 2 (a half filling) and
maintains the nominal resolution but reduces the SNR.
The number of lines on the negative half of the k-space
that are sampled in a partial-FT experiment is called
partial-FT overscans.
(b) Zero-fill acceleration also oscillates between 1 and 2 but in
that case it entails a symmetric truncation of the k-space.
(c) PPI acceleration is a very good option that relies on the
partially parallel imaging (PPI) but is only possible when a
phased array coil is available. The maximum value of this
factor is given by the number of channels and decreases
the acquisition time according to that value.
4. It is important to acquire the segmented EPI acquisition trig-
gered to the respiration cycle (images are acquired in the same
moment of the breath cycle) to avoid the appearance of artifacts
and ghost due to the multiple segment acquisition. For that

purpose, ensure that the respiration of the subject is as stable as
possible and use the selection of triggering per slice.
5. In the experiment, b-value has to be selected depending on the
interested diffusion weighting: a value too low induces low
diffusion and the signal decay cannot be probably determined
in an accurate way; a value too high induces a high signal decay
in such a way that signal intensity reaches the noise level. In
general, high b-values has to be selected for assessing regions
where water molecules diffuse slowly, whereas for high diffu-
sion regions it is better to choose a low b-value.
6. Select the acquisition parameters depending on the interests of
the experiment. It is important to note that DW images usually
suffer from low SNR due to the diffusion gradient dephasing
and the long TE values. Regarding on that, some important
considerations have to be taken into account:
(a) Optimal SNR is achieved with a TR value around 5·T1 to
ensure the recovery of the initial magnetization, or
increasing the average values.
(b) A minimum TE will results in a better SNR.
(c) Increasing spectral bandwidth, the geometrical distor-
tions linked to EPI imaging are minimized and also allows
a lower TE.
(d) Basal images with a b-value of zero have to be included in
the experiment as a function of the number of direction
assessed (approx. in a 1:6 ratio).
(e) Although a 100% of diffusion gradient strength can be
achieved, in practice it is not recommended to exceed
85%.
Acknowledgments
This work was supported by grant SAF2014-53739-R. IG held a

predoctoral contract from Ministerio de Economı́a, Indrustria y
Competitividad (MINECO) of Spain. DC held a postdoctoral
contract from Consejo Superior de Investigaciones Cientı́ficas
(CSIC).
References
€
1. Einstein A (1905) Uber die von der moleku- 2. Einstein A (1956) Investigations on the theory
larkinetischen Theorie der W€arme geforderte of the Brownian movement. Dover Publica-
Bewegung von in ruhenden Fl€ ussigkeiten sus- tions, NY
pendierten Teilchen. Ann Phys-Berlin 322 3. Le Bihan D (1995) Diffusion, perfusion and
(8):549–560 functional magnetic resonance imaging. J Mal
Vasc 20(3):203–214
Diffusion MRI 101
4. Luypaert R, Boujraf S, Sourbron S, Osteaux M 15. van Gelderen P, de Vleeschouwer MH,

(2001) Diffusion and perfusion MRI: basic DesPres D, Pekar J, van Zijl PC, Moonen CT
physics. Eur J Radiol 38(1):19–27 (1994) Water diffusion and acute stroke. Magn
5. Stark D, Bradley W (1999) Magnetic reso- Reson Med 31(2):154–163
nance imaging, vol 1. Mosby, St. Louis, MO 16. Hagmann P, Jonasson L, Maeder P, Thiran JP,
6. Stark D, Bradley W (1999) Magnetic reso- Wedeen VJ, Meuli R (2006) Understanding
nance imaging, vol 3, Mosby St. Louis, MO diffusion MR imaging techniques: from scalar
7. Davis S, Fisher M, Warach S (2003) Magnetic diffusion-weighted imaging to diffusion tensor
resonance imaging in stroke. Cambridge Uni- imaging and beyond. Radiographics 26(Suppl
versity Press, Cambridge, UK 1):S205–S223. https://doi.org/10.1148/rg.
26si065510
8. Einstein A (1956) Investigation of the theory
of the Brownian movement. Dover, New York 17. Schaefer PW, Grant PE, Gonzalez RG (2000)
Diffusion-weighted MR imaging of the brain.
9. Beaulieu C (2002) The basis of anisotropic Radiology 217(2):331–345. https://doi.org/
water diffusion in the nervous system—a tech- 10.1148/radiology.217.2.r00nv24331
nical review. NMR Biomed 15(7–8):435–455.
https://doi.org/10.1002/nbm.782 18. Mukherjee P, Berman JI, Chung SW, Hess CP,
Henry RG (2008) Diffusion tensor MR imag-
10. Mori S, Barker PB (1999) Diffusion magnetic ing and fiber tractography: theoretic underpin-
resonance imaging: its principle and applica- nings. Am J Neuroradiol 29(4):632–641.
tions. Anat Rec 257(3):102–109 https://doi.org/10.3174/ajnr.A1051
11. Stejskal EO, Tanner JE (1965) Spin diffusion 19. Schneider CA, Rasband WS, Eliceiri KW
measurements: spin schoes in the presence of a (2012) NIH Image to ImageJ: 25 years of
time-dependent field gradient. J Chem Phys image analysis. Nat Methods 9(7):671–675
42:288–292
20. Team RC (2013) R: a language and environ-
12. Tofts PS, Lloyd D, Clark CA, Barker GJ, Parker ment for statistical computing. R Foundation
GJ, McConville P, Baldock C, Pope JM (2000) for Statistical Computing, Vienna, Austria
Test liquids for quantitative MRI measure-
ments of self-diffusion coefficient in vivo. 21. Gruetter R (1993) Automatic, localized in vivo
Magn Reson Med 43(3):368–374 adjustment of all first- and second-order shim
coils. Magn Reson Med 29(6):804–811
13. Holz M, Heil SR, Sacco A (2000)
Temperature-dependent self-diffusion coeffi- 22. Kanayama S, Kuhara S, Satoh K (1996) In vivo
cients of water and six selected molecular rapid magnetic field measurement and shim-
liquids for calibration in accurate 1H NMR ming using single scan differential phase
PFG measurements. Phys Chem Chem Phys 2 mapping. Magn Reson Med 36(4):637–642
(20):4740–4742 23. Garcia-Palmero I, Lopez-Larrubia P, Cerdan S,
14. Le Bihan D, Douek P, Argyropoulou M, Villalobo A (2013) Nuclear magnetic reso-
Turner R, Patronas N, Fulham M (1993) Dif- nance imaging of tumour growth and neovas-
fusion and perfusion magnetic resonance imag- culature performance in vivo reveals Grb7 as a
ing in brain tumors. Top Magn Reson Imaging novel antiangiogenic target. NMR Biomed 26
5(1):25–31 (9):1059–1069. https://doi.org/10.1002/
nbm.2918
Chapter 7
Diffusion Tensor Imaging (DTI)

Silvia Lope-Piedrafita
Abstract
Diffusion Tensor Imaging is an MRI technique that allows in vivo noninvasive measurement of the
translational motion of water, providing information about its anisotropy (or lack of it) in different tissues.
DTI has been commonly used to quantitatively measure the integrity of tissues which may be compromised
by neurological disease, such as white matter tracks of the brain, which normally impart significant
anisotropy to water motion in healthy brains. However, this anisotropic effect is diminished when axonal
or neuronal damage is present. This chapter describes a standard protocol for DTI data acquisition in
preclinical studies.
Key words Diffusion tensor imaging, Fractional anisotropy, Apparent diffusion coefficient, Axial
diffusivity, Radial diffusivity, White matter, Fiber tracking
1 Introduction
Molecular diffusion refers to random, microscopic translational

movement of water and other small molecules caused by thermal
processes (i.e., Brownian motion). Diffusion driven displacements
of the molecules probe tissue structure at a microscopic scale since
during that random motion they cross or interact with cell mem-
branes, fibers, or macromolecules. The observation of this displace-
ment distribution may thus provide unique clues to the structure
and geometric organization of tissues.
Stejskal and Tanner described how to encode molecular diffu-
sion effects in the NMR signal by using bipolar magnetic field
gradient pulses, back in 1965 [1]. This diffusion technique was
later combined with NMR imaging principles and in 1985 the first
diffusion-weighted images were published [2–4]. Thus, diffusion
sensitivity is commonly encoded by the addition of two magnetic
field gradient pulses of duration δ and amplitude GD that are
separated by a diffusion time Δ, which is the total amount of time
spins are allowed to diffuse. The first diffusion gradient encodes the
spins with a characteristic phase and the second diffusion gradient,
103
104 Silvia Lope-Piedrafita
of equal time and amplitude, rewinds this phase shift, in such a way
that for static spins, where no net translational movement has taken
place, the phase of all the spins is brought back together after the
second gradient. However, if molecules diffuse during the time
delay between the two gradients, the refocusing is incomplete and
a phase dispersion remains after the second gradient. Accordingly,
signal intensity diminishes compared to that of static molecules.
Experimentally, the equation for signal attenuation, S, is com-
monly described as [5]
S ¼ S 0 e bD ð1Þ
where

2 δ
b ¼ ðγ δ G D Þ Δ ð2Þ
3
S0 correspond to the signal amplitude obtained without diffu-

sion gradients, and γ is the gyromagnetic constant intrinsic for each
nucleus. The diffusion-sensitizing factor, b, depends on parameters
that can be controlled experimentally. Therefore, the diffusion
coefficient, D, can be estimated experimentally by acquiring
diffusion-weighted images at different b-values and fitting the
data to Eq. 1 (see Fig. 1).
Because each voxel in an image may contain multiple types of
tissue with water in both the intra- and extracellular spaces, the
parameter that is measured in a diffusion experiment is usually
called an apparent diffusion coefficient (ADC), rather than the
true diffusion coefficient. The fact that water diffusion is sensitive
to the tissues cellular and subcellular architecture and integrity
makes diffusion-weighted MRI a unique technique to investigate
a variety of diseases like stroke, cancer, and many other neurological
disorders [6–8].
When water motion is equally probable in any direction, it is
said that water diffusion is isotropic. Sometimes though, if the
medium has a peculiar physical arrangement or there are obstacles
that limit molecular movement in some directions more than in
others, water diffusion is anisotropic. For the case of anisotropic
diffusion, Eq. 1 has to be written in a more general form [9],
T
S ¼ S 0 e b g^ D g^
ð3Þ
where gb is the normalized diffusion gradient direction and D is a

three rank symmetric tensor named diffusion tensor which
describes diffusion of water in three dimensions as
2 3
D xx D xy D xz
D ¼ 4 D yx D yy D yz 5 ð4Þ
D zx D zy D zz
DTI 105
Fig. 1 Estimation of the diffusion coefficient (ADC map, top right) by the exponential fitting of data acquired at
three diffusion weightings (b ¼ 200, 600, and 1000 s/mm2) and a non-diffusion one (b0 image, top left)
Because of the symmetry of this tensor (Dxy ¼ Dyx, Dxz ¼ Dzx,

and Dyz ¼ Dzy), it will be necessary to have, at least, six diffusion-
weighted images along six non-colinear directions [9]. It is also
necessary to acquire an image without diffusion weighting, b0
image, to be able to calculate all the elements of the tensor. In
practice, many more than six directions are often acquired with the
diffusion directions spread isotropically over a sphere in order to
improve signal-to-noise and accuracy of the diffusion tensor esti-
mation [10]. Care should be taken when many gradient pulses are
used, since b calculation may become more complicated [11]. Once
all the elements of D are known, the tensor is diagonalized, which is
a standard mathematical matrix operation that rotates the coordi-
nate frame of reference into the principal axes of the tensor, and as a
result, the eigenvalues (λ1, λ2, λ3) with its corresponding eigenvec-
tors (ε1, ε2, ε3) are explicitly obtained. The eigenvalues correspond
to the principle diffusivities of the diffusion tensor and the eigen-
vectors correspond to the directions of the principal diffusivities. By
convene, λ1 always refers to the longest principle diffusivity and λ3
to the shortest. The concept of diffusion ellipsoids is often used to
clearly represent diffusion tensor data (see Fig. 2). The surface of the
ellipsoid depicts the displacement front of water molecules.
Fig. 2 The diffusion tensor visualized as an ellipsoid in the case of isotropic diffusion (left) and anisotropic
diffusion (right). The diffusion ellipsoid represents the three-dimension mean squared displacement of mobile
hydrogen within the voxel. When the diffusion has no preferred direction, λ1 ¼ λ2 ¼ λ3, it is said that diffusion
is isotropic and is represented as a sphere. When diffusion is anisotropic, the diffusion ellipsoid is oriented
such that its long axis, λ1, is parallel to the direction of greatest diffusion, ε1, known as the primary principal
diffusion direction. Two additional vectors represent the rate of diffusion perpendicular to the primary principal
diffusion direction with axis lengths proportional to λ2 and λ3, respectively
From these principle diffusivities various quantitative para-

meters, which are scalar invariants, can be calculated. The mean
diffusivity, or apparent diffusion coefficient (ADC), which gives
information about the average diffusion, is calculated as
λ1 þ λ2 þ λ3
ADC ¼ ð5Þ
3
There are also parameters that give information about the water
diffusion anisotropy of the tissue. The most commonly used mea-
sure is the fractional anisotropy (FA). FA measures the fraction of
the “magnitude” of D that can be ascribed to anisotropic diffusion
and is calculated according to [5]
rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
h iffi
3 ðλ1 hλiÞ2 þ ðλ2 hλiÞ2 þ ðλ3 hλiÞ2
FA ¼ qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
ffi ð6Þ
2 λ21 þ λ22 þ λ23
FA has values between 0 (isotropic diffusion) and 1.0 (diffusion
restricted to a single direction).
In the cases where cylindrical diffusion anisotropic symmetry is
observed, for example, in white matter tracks, often changes in
diffusivity are given as changes in parallel diffusivity, λ||, also
named as longitudinal diffusivity, and as changes in perpendicular
diffusivity, λ⊥, also named as radial or axial diffusivity, according to
λjj ¼ λ1 ð7Þ
DTI 107
λ2 þ λ3
λ⊥ ¼ ð8Þ
2
Changes in λ⊥ and λ|| may potentially be used to differentiate
myelin loss versus axonal injury. Both circumstances induce a
decrease in FA; however, a loss of myelin integrity is correlated
with an increase in λ⊥, whereas axonal damage is associated to a
decrease in λ|| [12, 13].
In addition, the physical direction of the preferred motion of
water in the laboratory frame can also be elucidated directly from
the eigenvectors which can be displayed in directional-encoded-
color (DEC) maps. Moreover, the diffusion tensor is used for fiber
orientation tracking by connecting voxels based on the primary
principal diffusivity direction [14]. However, the traditional
single-tensor DTI model described here has been shown to be
inadequate in many regions of the brain that contain crossing fibers,
leading to false connections. More advanced models, which are
beyond the scope of this chapter, have been introduced to over-
come this problem including multitensor models, high angular
resolution diffusion imaging (HARDI), Q-ball imaging, or diffu-
sion spectrum imaging (DSI) [15].
The quantitative anisotropic parameters calculated from DTI
data sets are sensitive to the geometry and integrity of brain tissue.
For example, in dense white matter tracks, especially highly myelin-
ated ones, water diffusion is restricted in directions perpendicular
to the axon’s long axis, leading to high values of anisotropy.
Accordingly, DTI has been widely used to study a number of
white matter diseases and developmental pathologies [12, 13, 16,
17]. In addition, there is an active field of DTI research in the area
of cancer where the usefulness of anisotropy parameters for tumor
classification and for detecting tumor invasion is being investigated
[18, 19].
2 Materials
1. A high field strength whole body small animal MRI scanner

(typically from 4.7 T to 21 T) with a strong and fast gradient
system with maximum gradient capabilities between 400 and
1000 mT/m and maximum slew rates of 2000–7000 mT/m/s.
For the specific protocol presented in this work a 7T Bruker
BioSpec 70/30 USR MR system (Bruker BioSpin GmbH,
Karlsruhe, Germany) equipped with a BGA12 mini-imaging
gradient insert (maximum amplitude: 400 mT/m and slew
rate: 5500 T/m/s) was used. An optimized preemphasis adjust-
ment is important for reducing image artifacts.
2. Coils: a transmitter coil with good B1 homogeneity and a

receiver with good signal-to-noise ratio (SNR) over the interest
volume to be imaged. In this case, a 72 mm inner diameter
birdcage volume coil was used for excitation and a four-channel
phased array rat brain surface coil for signal reception.
3. Anesthesia vaporizers, anesthetics, appropriate carrier gases,
and an induction chamber. Breathable isoflurane is the most
commonly used anesthetic for small animal MRI due to its
rapid effect and recovery and its excellent tolerance. Isoflurane
can be delivered with either oxygen, medical air, or a mixed of
both. Two vaporizers are recommended to be able to indepen-
dently manage anesthesia to animals inside the MRI scanner or
to animals in preparation.
4. Animal bed equipped with a fixation system (tooth-bar and
ear-bars), to avoid motion artifacts, and a tube that continu-
ously provides anesthesia while inside the magnet. Many beds
also include a heating water system.
5. MR compatible small animal monitoring and gating system
(Model 1025, SA Instruments, Inc., Stony Brook, NY, USA).
6. MR compatible physiological sensors: Temperature probe, res-
piration pad, and ECG electrodes.
7. Warming system to maintain animal body temperature, either
using a circulating warm water bath or by blowing warm air
over the animal.
8. MRI software for data acquisition and processing. ParaVision
5.1 software (Bruker BioSpin GmbH, Ettlingen, Germany)
implemented on the MRI scanner or other available software
for DTI data processing like DtiStudio[20].
9. Nonwoven tape used to fasten the respiration pad and other
cables, connectors or coils that need to be fixed to the animal
bed. Scissors are not allowed in the MRI room; therefore it is
very convenient to use tape that can be easily cut with
finger tips.
3 Methods
3.1 Pre-scanning The aim is to have everything ready so that, once the animal is
Preparation anesthetized, we are ready to run the experiment, thus minimizing
the time that the animal is under anesthesia.
1. It is very important to remove all loose metals from the opera-
tor and the animal (metal ear tags) prior to entering the
scanner room.
2. Insert and center volume coil inside the magnet.
3. Turn on heating system (water bath or heating module fan).
DTI 109
4. Turn on computer and open animal monitoring software (PC-

SAM, SA Instruments, Inc., Stony Brook, NY, USA) and check
that is working properly.
5. Open MRI acquisition software (in our case ParaVision 5.1),
and create a new study.
6. Ensure that there is enough anesthetic in the vaporizers for the
duration of the experiment.
7. Prepare the animal bed. Check that it has ear bars and a tooth-
bar, and that the anesthesia is properly connected. Have the
respiration pad and the temperature probe handy and ready for
use. If performing cardiac gating (for heart, or body DTI),
have the ECG electrodes ready as well.
8. Induce animal anesthesia: place the animal inside the induction
chamber at 4% (v/v) isoflurane in O2 at 1 L/min.
9. Once the animal is deeply anesthetized, place it rapidly into the
animal bed using the bite-bar and ear-bars for optimal head
immobilization (see Note 1). Typically, 1–1.5% isoflurane in
Oxygen at a flow rate of 1 L/min is used for maintenance.
10. Insert rectal probe and monitor the body temperature of the
animal. It should be maintained at 37 C (see Note 2).
11. Place the respiration pad in the chest area. For proper detec-
tion, the respiration pad is usually placed under the abdomen in
mouse studies and over the chest in rats. Use tape to fix the pad
to the bed, and around the animal, so that the sensor can better
detect the pressure variations. Respiration frequencies should
be maintained on 60–80 breaths/min (see Note 3).
12. Place ECG electrodes if necessary. They should be located in
opposite forepaw and hindpaw.
13. If using a surface coil as receiver, place it over the head of the
animal with the aid of a tape to ensure fixation and avoid
motion.
14. Insert the animal bed into the magnet and place the region of
interest (brain in our case) in the center of the magnet (which
should also correspond to the center of the gradients and the
center of the volume coil).
15. Tune and match the coils.
16. Make sure to closely monitor and maintain body temperature
and the respiration rate during the duration of the experiment.
3.2 Pilot Scan Load a pilot scan (in ParaVision named “tripilot”) in the MRI
acquisition software. Complete automatic adjustments which
enable to boost SNR: global shimming, setting the reference fre-
quency, and calibration of the reference frequency (see Note 4).
Adjust receiver gain and run the tripilot scan. As a result, three
orthogonal images acquired through the center of the gradients are

obtained. These images are used to confirm if the animal is well
positioned in the center of the magnet and also as reference scout
images for localizing the sections in the next scans.
3.3 DTI Scan 1. Load a diffusion protocol from the MRI software library. The
most commonly used sequence for DTI studies is the bipolar
pulse field gradient diffusion sequence in its spin-echo version
and with segmented echo planar imaging (EPI) acquisition
scheme, DTI_EPI (see Note 5).
2. Select slice geometry through the region of interest. Number
of slices, and orientation (see Note 6).
3. Choose spatial resolution, field of view (FOV), matrix size, and
slice thickness. 2D or 3D options are available, although usu-
ally 2D is performed in small animal studies since 3D requires
much longer acquisition times (see Note 7). Making the voxel
isotropic is important for fiber tracking.
4. Insert number and values for the b-value. A standard DTI data
set include acquisition of one image at low b-value (b ¼ 0
usually) and several diffusion-weighted images at only 1 high
b-value (~1000 s/mm2) (see Note 8).
5. Choose diffusion time Δ and the duration of the diffusion
gradients δ. They need to be selected trying to minimize TE
while maintaining the designated high b-value.
6. Enter number of sensitive diffusion directions, and number of
b0 images. For a proper estimation of the diffusion tensor,
5–10 measurements should be made at high b-values for
every measurement made at b0 [21, 22]. On the other hand,
a minimum SNR of 20 is recommended for the b0 image
[23]. Also, diffusion directions should be well distributed in
space (see Note 9).
7. Fat suppression should be used since aids in reducing Nyquist
ghost artifacts (see Note 10).
8. Activate respiration gating to reduce motion artifacts if neces-
sary (see Note 11).
9. Run the scan.
Sample DTI parameters for rat brain are presented in Table 1.
3.4 DTI Data 1. Diffusion tensor calculation and derived parameters, described
Processing in the introduction, can be obtained in a pixel-by-pixel basis by
using the scanner manufacture software (in our case ParaVision
5.1). It outputs maps for each component of the diffusion
tensor, each eigenvalue, each direction of the three eigenvec-
tors, the tensor trace which corresponds to the ADC value, and
the fractional anisotropy (see Fig. 3). Alternatively, free software
DTI 111
Table 1
Sample parameters for DTI acquisition in a rat brain study at 7T
Sequence Segmented DTI-EPI

Number of segments 4
Field of view 35 35 mm2
Matrix size 128 128
Slice thickness 1 mm
Number of slices 9
b Value 1000 s/mm2
Diffusion gradient duration 4 ms
Diffusion gradient separation 18 ms
Number of b0 images 5
Number of diffusion directions 30
Echo time 40 ms
Repetition time 3s
Receiver bandwidth 200 kHz
Scan time 7 min
is available which can also provide you with these maps

[24]. The software DtiStudio [20], for example, offers addi-
tional features compared to ParaVision, like image
co-registration, the option to eliminate images with large signal
dropouts or geometric distortions, prior to tensor calculation,
or fiber tracking analysis. Some groups also use their own
home-written programs for DTI data processing [25, 26].
2. Different types of data analysis can be performed depending on
the application at hand, which may include region of interest
analysis, voxel-based analysis, or fiber tracking [23].
3. Perform data statistical analysis, which also depends on the
study aim.
4 Notes
1. Diffusion is very sensitive to motion. It is very important to

perfectly fix the animal head to avoid motion artifacts during
acquisition.
2. Diffusion changes with temperature; therefore maintenance of
the right body temperature is very important in DTI experi-
ments. Body temperature usually drops when animals are under
Fig. 3 Rat brain b0 image (bottom left) and DTI parametric maps: longitudinal diffusivity (λ||, top left), axial
diffusivity (λ⊥, top middle), apparent diffusion coefficient (ADC, top right), fractional anisotropy (FA, bottom
middle), and directional encoding color (DEC, bottom right). The colors in the DEC map refer to the direction of
the primary principal diffusion direction and are coded as follows: red ¼ right-left, green ¼ dorsal-ventral, and
blue ¼ rostro-caudal. The DTI data set was acquired using the sample parameters presented in Table 1. High
ADC is observed in the ventricles and nonsignificant differences in ADC are seen between white and gray
matter. However, white matter has higher values of λ|| and FA compared to gray matter
anesthesia, thus when the animal is first located in the bed its
temperature is usually a few degrees below the target tempera-
ture (37 C). Vary water temperature in such a fashion as to
reach and keep the animal temperature constant throughout
the entire scanning time. Also try to scan at the same tempera-
ture all the animals within the same study for proper compari-
son. Automated, temperature feedback warm air systems are
usually connected to the monitoring system and work well to
compensate heating power, if body temperature changes occur.
Diffusion scans tend to increase body temperature, thus we
may anticipate these changes by lowering the bath temperature
a little bit in advance.
3. It is very important to monitor respiration rate very closely,
since different strains, and even different individuals within the
same strain, respond differently to anesthesia. Vary anesthesia
DTI 113
doses as necessary. Low respiration rates may lead to death of

the animal, whereas high respiration rates indicate that the
animal is awakening and may result in movement of the animal.
4. Many scanners run these automatic adjustments by default
when the first scan of the study is launched.
5. Although many different diffusion sequences can be found in
the literature [27–30], only a few of them are available in
commercial scanners. The standard sequence would be a diffu-
sion sequence that uses a spin-echo and a standard Cartesian
acquisition scheme. In terms of image quality this is the
sequence that gives better images and also allows higher spatial
resolutions. However it needs long acquisition times and thus,
for a basic DTI study which requires at least a set of seven
images (one b0 and 6 at high b-value along non colinear direc-
tions), experimental times may easily go over 1 h, making it
unfeasible in many circumstances. On the other hand, EPI
sequences offer much more rapid acquisitions allowing, for
example, the acquisition of diffusion images along 20–30 direc-
tions in about 10 min. The main drawbacks of using EPI are
the inherent image eddy current-induced distortions, magnetic
susceptibility artifacts, and that usually EPI images are acquired
at lower resolution. Nonetheless, these artifacts may be par-
tially lessened nowadays by using multi-shot or segmented EPI
techniques as well as with many image post-processing algo-
rithms which have been implemented in order to correct these
distortions. Local shim using automatic algorithms like Map-
shim [31] or Fastmap [32] may help to reduce magnetic sus-
ceptibility artifacts. Also, using segmented-EPI, higher spatial
resolutions may be achieved at reasonable echo times. In con-
clusion, the DTI-EPI sequence is in principle the default
choice; however standard spin-echo sequences may be of
choice when reliable geometry is required in the study (for
example, when pixel-by-pixel correlations between multimodal
MRI techniques are performed).
6. A priori, MRI allows imaging of sections in any orientation.
Nonetheless, in our experience, experimentally, if oblique sec-
tions are selected, more motion like artifacts are observed in the
images, which may be due to interactions between diffusion
and slice selection gradients. Therefore, it is important to
position the animal as straight and centered as possible to
avoid latter on the selection of tilted and off center field of
view sections in DTI scans.
7. Isotropic voxel dimensions at high spatial resolution are the
optimal approach to DTI acquisition if sufficient scanning time
is available. However spatial resolution affects the SNR; there-
fore, there is always a tradeoff between resolution, SNR, and
scan time, which will greatly depend on the study goals. An

SNR greater than 20, for the non-diffusion sensitive image b0,
is recommended [23].
8. The optimal high b-value has been numerically estimated to be
~1/ADC [21]. White and gray matter in mature healthy brains
has diffusion values around 0.6–1.2 103 mm2/s and a high
b-value of 1000 s/mm2 is commonly used. Nonetheless, when
dealing with much higher ADC values, a lower b-value is
recommended. Also, it has been shown that acquisition at
multiple high b-values do not significantly improve the estima-
tion of the anisotropy parameters [21], nonetheless, acquisi-
tions at multiple b-values are required when using multitensor
analysis or for bi-exponential fittings.
9. Determination of optimal number and distribution of gradient
directions for reliable tensor estimation is not straightforward.
Some studies have shown that FA estimation improves by
increasing the number of orientations, suggesting the acquisi-
tion of as many different directions as time allows. The use of
20–30 diffusion directions already improves the quality of the
tensor estimations and avoids orientational bias [10]. The gra-
dient encoding schemes also have an important influence on
the derived anisotropy parameters and are more relevant when
a small number of diffusion directions are used. In that case, the
use of reported optimal encoding schemes is recommended
[10, 33, 34].
10. The Nyquist ghost artifact is specific to EPI. It is due to the fact
that in EPI acquisitions k-space is traversed in opposite direc-
tions on alternate echoes. Errors in the phase of signal inten-
sities from sources like fat-water chemical shift can cause a
ghost image, shifted by one half of the FOV in the phase-
encoding direction, which overlays on the original image
[23]. The use of frequency-selective fat saturation pulses
helps to minimize these artifacts.
11. DTI is very sensitive to motion. When doing body DTI, respi-
ration and ECG gating is obviously required. Moreover, in
brain DTI, even when the animal head is pretty well restricted,
motion artifacts are observed in the images, which are greatly
diminished when using respiration gated acquisitions. How-
ever, gating may significantly increase acquisition times.
Acknowledgements
The author expresses its gratitude to Dr. Gary V. Martinez (Lee

Moffitt Cancer Center and Research Institute, Tampa, FL, USA),
who assisted in the proofreading of the chapter. Time allocated in
DTI 115
the joint nuclear magnetic resonance facility of the Universitat

Autònoma de Barcelona and Centro de Investigación Biomédica
en Red-Bioingenierı́a, Biomateriales y Nanomedicina (CIBER-
BBN) (Cerdanyola del Vallès, Spain) is gratefully acknowledged.
References
1. Stejskal EO, Tanner JE (1965) Spin diffusion revealed through MRI as increased radial (but
measurements: spin echoes in the presence of a unchanged axial) diffusion of water. Neuro-
time-dependent field gradient. J Chem Phys Image 17:1429–1436
42:288–292 13. Song SK, Yoshino J, Le TQ, Lin SJ, Sun SW,
2. Le Bihan D, Breton E (1985) Imagerie de Cross AH, Armstrong RC (2005) Demyelin-
diffusion in vivo par resonance magnetique ation increases radial diffusivity in corpus callo-
nucleaire. CR Acad Sci Paris 301:1109–1112 sum of mouse brain. NeuroImage 26:132–140
3. Merboldt KD, Hanicke W, Frahm J (1985) 14. Conturo TE, Lori NF, Cull TS, Akbudak E,
Self-diffusion NMR imaging using stimulated Snyder AZ, Shimony JS, McKinstry RC,
echoes. J Magn Reson Imaging 64:479–486 Burton H, Raichle ME (1999) Tracking neu-
4. Taylor DG, Bushell MC (1985) The spatial ronal fiber pathways in the living human brain.
mapping of translational diffusion coefficients Proc Natl Acad Sci 96:10422–10427
by the NMR imaging technique. Phys Med 15. Tournier JD, Mori S, Leemans A (2011) Dif-
Biol 30:345–349 fusion tensor imaging and beyond. Magn
5. Le Bihan D, Mangin JF, Poupon C, Clark CA, Reson Med 65:1532–1556
Pappata S, Molko N, Chabriat H (2001) Dif- 16. Lope-Piedrafita S, Totenhagen JW, Hicks CM,
fusion tensor imaging: concepts and applica- Erickson RP, Trouard TP (2008) MRI detects
tions. J Magn Reson Imaging 13:534–546 therapeutic effects in weanling Niemann-Pick
6. Moseley ME, Kucharczyk J, Mintorovitch J, type C mice. J Neurosci Res 86:2802–2807
Cohen Y, Kurhanewicz J, Derugin N, 17. Breu M, Zhang J, Porambo M, Pletnikov MV,
Asgari H, Norman D (1990) Diffusion- Goeral K, Kakara M, Johnston MV, Fatemi A
weighted MR imaging of acute stroke: correla- (2016) Diffusion tensor imaging abnormalities
tion with T2-weighted and magnetic in the cerebral white matter correlate with
susceptibility-enhanced MR imaging in cats. sex-dependent neurobehavioral deficits in
Am J Neuroradiol 11:423–429 adult mice with neonatal ischemia. Dev Neu-
7. Chenevert TL, McKeever PE, Ross BD (1997) rosci 38:83–95
Monitoring early response of experimental 18. Lope-Piedrafita S, Garcia-Martin ML, Galons
brain tumors to therapy using diffusion mag- JP, Gillies RJ, Trouard TP (2008) Longitudinal
netic resonance imaging. Clin Cancer Res diffusion tensor imaging in a rat brain glioma
3:1457–1466 model. NMR Biomed 21:799–808
8. Dmytriw AA, Sawlani V, Shankar J (2017) 19. Gimenez U, Perles-Barbacaru AT, Millet A,
Diffusion-weighted imaging of the brain: Appaix F, El-Atifi M, Pernet-Gallay K, van der
beyond stroke. Can Assoc Radiol J 68 Sanden B, Berger F, Lahrech H (2016) Micro-
(2):131–146. https://doi.org/10.1016/j. scopic DTI accurately identifies early glioma
carj.2016.10.001 cell migration: correlation with multimodal
9. Basser PJ, Mattiello J, Le Bihan D (1995) MR imaging in a new glioma stem cell model.
diffusion tensor spectroscopy and imaging. NMR Biomed 29:1553–1562
Biophys J 66:259–267 20. Jiang H, van Zijl PC, Kim J, Pearlson GD,
10. Skare S, Hedehus M, Moseley ME, Li TQ Mori S (2006) DtiStudio: resource program
(2000) Condition number as a measure of for diffusion tensor computation and fiber
noise performance of diffusion tensor data bundle tracking. Comput Methods Prog
acquisition schemes with MRI. J Magn Reson Biomed 81:106–116
147:340–352 21. Jones DK, Horsfield MA, Simmons A (1999)
11. Mattiello J, Basser PJ, Le Bihan D (1997) The Optimal strategies for measuring diffusion in
b matrix in diffusion tensor echo-planar imag- anisotropic systems by magnetic resonance
ing. Magn Reson Med 37:292–300 imaging. Magn Reson Med 42:515–525
12. Song SK, Sun SW, Ramsbottom MJ, Chang C, 22. Alexander DC, Barker GJ (2005) Optimal
Russell J, Cross AH (2002) Dysmyelination imaging parameters for fiber-orientation
estimation in diffusion MRI. NeuroImage 29. Pu LL, Trouard TP, Ryan L, Huang C, Altbach
27:357–367 MI, Bilgin A (2011) Model-based compressive
23. Mukherjee P, Chung SW, Berman JI, Hess CP, diffusion tensor imaging. 8th IEEE Int Symp
Henry RG (2008) Diffusion tensor MR imag- Biomed Imaging 1:254–257
ing and fiber tractography: technical considera- 30. Shi X, Ma X, Wu W, Huang F, Yuan C, Guo H
tions. Am J Neuroradiol 29:843–852 (2015) Parallel imaging and compressed sens-
24. Liu B, Zhu T, Zhong J (2015) Comparison of ing combined framework for accelerating high-
quality control software tools for diffusion ten- resolution diffusion tensor imaging using inter-
sor imaging. Magn Reson Imaging image correlation. Magn Reson Med
33:276–285 73:1775–1785
25. Sarlls JE, Pierpaoli C (2009) In vivo diffusion 31. Gr€u tter R (1993) Automatic, localized in Vivo
tensor imaging of the human optic chiasm at adjustment of all firstand second-order shim
sub-millimeter resolution. NeuroImage coils. Magn Reson Med 29:804–811
47:1244–1251 32. Kanayama S, Kuhara S, Satoh K (1996) In vivo
26. Totenhagen JW, Lope-Piedrafita S, Borbon IA, rapid magnetic field measurement and shim-
Yoshimaru ES, Erickson RP, Trouard TP ming using single scan differential phase
(2012) In vivo assessment of neurodegenera- mapping. Magn Reson Med 36:637–642
tion in Niemann-Pick type C mice by quantita- 33. Hasan KM, Parker DL, Alexander AL (2001)
tive T2 mapping and diffusion tensor imaging. Comparison of gradient encoding schemes for
J Magn Reson Imaging 35:528–536 diffusion-tensor MRI. J Magn Reson Imaging
27. Pipe JG, Farthing VG, Forbes KP (2002) Mul- 13:769–780
tishot diffusion-weighted FSE using PROPEL- 34. Jones DK (2004) The effect of gradient sam-
LER MRI. Magn Reson Med 47:42–52 pling schemes on measures derived from diffu-
28. Bammer R, Holdsworth SJ, Veldhuis WB, sion tensor MRI: a Monte Carlo study. Magn
Skare ST (2009) New methods in diffusion- Reson Med 51:807–815
weighted and diffusion tensor imaging. Magn
Reson Imaging Clin N Am 17:175–204
Chapter 8
Mapping Functional Connectivity in the Rodent Brain Using

Electric-Stimulation fMRI
Laura Pérez-Cervera, José Marı́a Caramés, Luis Miguel Fernández-Mollá,
Andrea Moreno, Begoña Fernández, Elena Pérez-Montoyo, David Moratal,
Santiago Canals, and Jesús Pacheco-Torres
Abstract
Since its discovery in the early 90s, BOLD signal-based functional Magnetic Resonance Imaging (fMRI)
has become a fundamental technique for the study of brain activity in basic and clinical research. Functional
MRI signals provide an indirect but robust and quantitative readout of brain activity through the tight
coupling between cerebral blood flow and neuronal activation, the so-called neurovascular coupling.
Combined with experimental techniques only available in animal models, such as intracerebral micro-
stimulation, optogenetics or pharmacogenetics, provides a powerful framework to investigate the impact
of specific circuit manipulations on overall brain dynamics. The purpose of this chapter is to provide a
comprehensive protocol to measure brain activity using fMRI with intracerebral electric micro-stimulation
in murine models. Preclinical research (especially in rodents) opens the door to very sophisticated and
informative experiments, but at the same time imposes important constrains (i.e., anesthetics, translatabil-
ity), some of which will be addressed here.
Key words fMRI, BOLD, Intracerebral micro-stimulation, Preclinical MRI
1 Introduction
Neuroscience is currently one of the most challenging areas of

biomedical research. The human brain and its function remain
one of the last frontiers for modern science. Furthermore, there is
a growing belief that its scientific understanding will have a pro-
found effect on human progress, not only because of the improve-
ment in treating neurologic and psychiatric disorders, but also
because of the impact that a deep understanding of the brain
would have on fields ranging from education, interpersonal rela-
tionships, the prolongation of a full physical and intellectual life and
also for the development of a neuro-inspired Information and
Communications Technology (ICT) society. In order to achieve
this major goal, scientists need techniques that allow proper
117
118 Laura Pérez-Cervera et al.
monitoring of brain function. Ideally, these techniques should

satisfy the following criteria:
1. Applicable in vivo to study the intact brain.
2. To provide adequate temporal and spatial resolution.
3. To reflect the electrical activity of neurons as close as possible.
4. To be applicable in human and non-human animal models.
Over the past few decades, multiple non-invasive in vivo imag-
ing methodologies have been developed to study brain function,
although none of them has been able to fulfill all of the above
mentioned criteria. These attempts can be classified on the basis
of their strategies for measuring brain activity: directly assessing the
electrical activity of the brain, or indirectly through surrogate mea-
sures as calcium transients, hemodynamic and/or metabolic
changes associated to neuronal activity. Among the former we can
include electroencephalography (EEG) and magnetoencephalogra-
phy (MEG), which present excellent temporal resolution but have
its particular Achilles’ heel in the spatial resolution. The second
category includes, among others, Positron Emission Tomography
(PET) and Magnetic Resonance Imaging (MRI). Functional PET
applications relay in the use of radioactively-labeled compounds
metabolically active (15-Oxygen, 18F-Fluorodeoxyglucose. . .)
whose measurement reports changes in blood flow, oxygen
and/or glucose consumption, which are surrogates of neuronal
activity. PET presents the highest sensitivity of all the aforemen-
tioned techniques, but the necessary use of radioactive com-
pounds—with their short half-life and the relatively low spatial
resolution—has hindered his widespread for neurological imaging
purposes (see Ref. 1 for a review on neuroimaging techniques).
Functional magnetic resonance imaging (fMRI) utilizes MRI
technology to indirectly measure brain activity by detecting
changes associated with it. The best-known form of fMRI uses
the blood-oxygen-level dependent (BOLD) contrast [2]. It relies
on the brain’s ability to locally modulate blood flow and volume to
satisfy the increased energetic needs during neuronal activation.
The relationship between increased local neural activity and
changes in cerebral blood flow (CBF) is known as neurovascular
coupling (or functional hyperemia). The correlation between
BOLD signals and concomitantly recorded electrophysiological
measures is well established [3, 4], although the exact molecular
mechanism remains unclear [5] and its homogeneity throughout
brain regions debated [4]. Functional MRI allows imaging brain
activations with a spatial resolution as low as a few hundreds of
microns (limited by the low intrinsic sensitivity of MRI, the small
expected changes in signal intensity and the point spread function;
for a review see [6]) and with a temporal resolution of seconds or
Electric-Stimulation fMRI 119
even shorter (limited by the nature of the hemodynamic response

itself).
It is important to stress that BOLD signals do not measure
neuronal activity, but changes in neuronal activity (see [6] for a
deeper discussion on the fundamentals of BOLD). Nevertheless,
these changes in neuronal activity can occur spontaneously while
the subject is not involved in any specific task. In those conditions
the changes are of small amplitude and are referred to as resting
state activity. Alternatively, the change in activity can be boosted by
introducing a specific paradigm that defines an expected low activity
state (baseline) and a high activity state (activation) that are subse-
quently contrasted statistically. Common paradigms include the
presentation of sensory stimuli, the introduction of a cognitive
task, the administration of chemical substances, or the direct stim-
ulation of the brain’s parenchyma. In animal models, the availability
of a larger repertoire of neuronal stimulation strategies allows the
precise control of the stimulation parameters and neuronal popula-
tions being targeted. These strategies include intracerebral direct
electrical micro-stimulation and optogenetic or pharmacogenetic
stimulation or inhibition of specific neuronal populations. These
might overcome some of the limitations of sensory and task stimu-
lation when the focus of the study is to unveil the precise neuronal
circuits involved in a certain functional state. For instance, while
activation of primary sensory structures is in some cases readily
available by sensory stimulation, recruitment of other regions as
some subcortical nuclei is more challenging. Also, the polysynaptic
propagation of activity in intricate brain networks initiated by sen-
sory stimuli or cognitive tasks makes it harder to investigate direct
relationships or causality. Direct activation of a specific pathway or a
neuronal population greatly eases such studies, allowing to investi-
gate the consequences of precisely timed and accurately localized
manipulation on the overall response of brain-wide networks.
Highly sophisticated and rich experiments can be done in
animal models by the combination of fMRI with electrophysiologi-
cal, optogenetic and pharmacogenetic tools. However, the use of
animals (especially small animals as rats and mice) imposes certain
limitations that need to be considered carefully. Spatial resolution
issues, due to the smaller size of rodent’s brains compared to
primates, have been partially countervailed by the implementation
of systems with ultrahigh magnetic fields and stronger radiofre-
quency gradients, yet still the obtained resolution is coarse for
imaging certain subcortical structures. Therefore, partial volume
effects need to be seriously considered. Also, the need for the
absolute immobility of the subject during image acquisition has
favored the used of anesthetized animals in most of the studies.
Besides the obvious impact of anesthesia on neuronal activation,
additional factors need to be considered in fMRI studies since
anesthetics could directly impact on the hemodynamic response
and the neurovascular coupling (discussed in detail in Subheading

3.1) [7] among other neuronal activation features. Some alterna-
tives to avoid anesthetics are present, e.g., habituation training
protocols which allow to do fMRI experiments in
non-anesthetized rodents [8–10]. Although they are very appealing
and pave the way to advance in rodent fMRI, the available protocols
involve an initial period of severe stress during habituation to the
MRI environment, which could interfere with the particular scien-
tific question at hand. Finally, further complications derived from
the combination of fMRI with other recording or stimulation
techniques include the need for MRI-compatible materials in all
implants and devices required in the experiment.
The present protocol describes, stepwise, how to perform a
BOLD fMRI experiment with intracerebral electric micro-
stimulation, which has allowed us to (1) perform precise and con-
trolled activations of selected brain regions, (2) reproducibly
acquire data within and between animals in rats and mice, (3) to
investigate the frequency-dependence of activity propagation in
brain networks, and (4) to provide several insights into the synaptic
plasticity control of long range connectivity.
2 Materials
2.1 Carbon Electrode 1. Theta-shaped glass capillary (World Precision Instruments,

Preparation Florida, USA).
2. Carbon fibers.
3. Micropipette puller.
4. Bunsen burner.
5. Standard forceps, straight.
6. London forceps, angled.
7. Epoxy resin (fast drying).
8. Silver conductive epoxy resin.
2.2 Electrode 1. Isoflurane.

Implantation Surgery 2. Urethane.
3. Bupivacaine.
4. Stereotaxic frame and stereotaxic micromanipulator.
5. Ophthalmic gel.
6. Heating pad with rectal thermal probe.
7. Surgical instruments:
(a) Small surgical scissors, straight.
(b) Standard forceps, straight.
(c) Dumont forceps, straight.

(d) Scalpel handle, straight.
(e) Non-sterile scalpel blade, curved, 22.
(f) Micro curette, straight.
(g) Micro drill trephine.
8. Sterile saline solution and hydrogen peroxide.
9. High-temperature Cautery kit.
10. Cotton.
11. Shaver.
12. Permanent marker.
13. 20G and 25G needles.
14. Tissue adhesive.
15. Bone acrylic.
16. Super-Bond C&B dental acrylic.
2.3 Functional MRI The additional materials listed here are required to produce electri-
cal brain stimulation in the rat while it remains in the magnet and
acquire and process imaging data.
2.3.1 MRI Magnet We use a horizontal 7 Tesla scanner with a 30 cm diameter bore
(Biospec 70/30v, Bruker Medical, Ettlingen, Germany). The sys-
tem has a 675 mT/m actively shielded gradient coil (Bruker, BGA
12-S) of 11.4 cm inner diameter. Data is acquired and
pre-processed with a Hewlett-Packard console running Paravision
5.1 software (Bruker Medical GmbH, Ettlingen, Germany)
operating on a Linux platform.
2.3.2 MRI Phase It is not indispensable, but it is highly convenient. The expected
Array Coil signal changes are in the range of 1–3% of the total signal intensity,
thus it will critically contribute to achieve the highest possible
signal-to-noise ratio (SNR). We employ a1H rat brain receive-only
phase array coil with integrated combiner and preamplifier, no
tune/no match, in combination with the actively detuned
transmit-only resonator (BrukerBioSpin MRI GmbH, Germany).
2.3.3 Physiological We use an MRI-compatible temperature control unit (MultiSens

Monitoring and Control Signal conditioner, OpSens, Quebec, Canada). Other physiological
System parameters as heart rate (optimal values, 300 50 beats per min-
ute), oxygen saturation (>95%), and breathing rate (90 10
breaths per minute) are monitored (see Note 1) throughout the
session using an MRI-compatible sensor with foot clip (MouseOx,
Starr Life Sciences, Oakmont, USA). Physiological parameters can
be used to feed the analysis of BOLD signals (used as nuisance
factors) which might be especially important in resting state
experiments.
2.3.4 Heating System It is crucial to keep the animal’s temperature in the physiological
range, and to maintain it stable (37 0.5 C) in order to preserve
vascular reactivity in response to neuronal activation. We use a water
blanket connected to a water bath (Thermo Scientific SAHARA
Heated Bath Circulators S5P) controlled by a temperature regu-
latory system (Thermo Scientific STANDARD Series Thermostats
SC150) (see Note 2). In this system, precise control of the animal’s
temperature requires the constant attention of dedicated personnel
to manually vary the temperature in the heating bath. In order to
automate this process, a home-made and inexpensive closed loop
regulation system has been developed to adjust water bath temper-
ature to keep a constant body temperature in the animal (see details
in Note 3).
2.3.5 Other Devices 1. Pulse generator and current source (STG2004, Multichannel
and Small Equipments Systems, Reutlingen, Germany) for electric micro-stimulation.
2. Digital Oscilloscope to check electrode functionality.
3. Eye ointment (even if it is an acute procedure).
4. MRI-compatible stereotaxic device with ear- and bite-bars.
5. Agarose (0.5%) in saline. It is prepared and introduced in a
10 cc syringe. It can be storage in the fridge before its use (see
Note 4).
6. MRI sequences: gradient Echo (GE)-Echo Planar Imaging
(EPI) sequence providing adequate temporal and spatial reso-
lution (see Notes 5 and 6); and T2-weighted anatomical
images, like a Rapid Acquisition Relaxation Enhanced
(RARE) sequence (see Note 7).
2.3.6 Image Analysis There are some commercially available software tools for fMRI
Software analysis. In our case, fMRI data are analyzed offline using our
own software developed in MatLab, which included Statistical
Parametric Mapping package (SPM8, http://fil.ion.ucl.ac.uk/
spm), Analysis of Functional NeuroImages (AFNI, http://afni.
nimh.nih.gov/afni), and FSL Software (FMRIB http://fsl.fmrib.
ox.ac.uk/fsl).
3 Methods
All animal work should be carried out only upon review and
approval of the methods by your institution’s Animal Care and
Use Committee [11, 12]. For those new to MRI and small animal
surgery, prior to initiating any studies, training and advice should
be sought from experts in the field.
Due to the presence of strong magnetic fields, surgery is per-
formed in an area separated from the magnet room. Thus, it will be
necessary to fix the stimulating and/or recording electrode to the

animal’s skull so that the animal can be safely transferred to the
magnet room at the end of the surgery. Furthermore, due to the
common use of surface coils in fMRI experiments, both the elec-
trode positioning and its fixations must be done in such a way that
allow maximal proximity between the MRI coil and the brain of the
animal. Special care must be taken with bleeding during surgery,
because any trace of blood will have a deep impact in the image
quality, making it very difficult to obtain a reliable BOLD signal.
3.1 Anesthesia As previously introduced, most fMRI experiments in rodents are

performed in anesthetized animals. Different anesthetics have been
introduced for fMRI studies in rodents, each of which presents its
particular advantages and drawbacks (for a review see [7]) and all of
them having an impact on the neurovascular coupling. At this
point, it is important to emphasize that, in our experience, at least
80% of a successful fMRI experiment in rodents relies on maintain-
ing the animal’s physiology at adequate and steady-state values.
Body temperature (37 0.5 C), oxygen saturation (>95%),
CO2 (35–50 mmHg) and levels and blood pressure
(130–140 mmHg) need to be fine-tuned. While precise monitor-
ing of some of this values require invasive interventions (i.e., blood
pressure and accurate CO2 measurements require femoral artery
cannulation and tracheotomy, respectively) or direct blood sam-
pling difficult to implement in longitudinal studies, pilot experi-
ments with full monitoring of the animal’s physiology are strongly
recommended in setting up new anesthetic protocols.
The final election of an anesthetic method will depend on
multiple factors like the species utilized (i.e., rats [13] vs. mice
[14]), the duration of the experiment [15], whether it is an acute
or longitudinal experiment, and even the type of stimuli used
[16]. As a general rule, injectable anesthetics provide a stable
imaging condition for up to 2 h, whereas inhaled anesthesia allows
longer imaging sessions. An exception to this rule is urethane,
which provides a stable and long-lasting (more than 8 h) anesthe-
tized state with a single intraperitoneal injection and minimal car-
diovascular effects [17]. Importantly, urethane also preserves most
of the characteristic electrophysiological rhythms recorded in the
hippocampus and other neocortical regions [18]. In the present
protocol of electric stimulation fMRI, urethane has been the choice
based on the above advantages [18]. However, urethane is
restricted to terminal experiments due to its hepatotoxic and carci-
nogenic effects, for which it is compulsory to euthanize the animal
at the end of the experiment. For chronic rat experiments and when
working with mice, dexmedetomidine is the usual election [5]. It
allows animal recovery but provides, in our hands, shorter periods
of stable anesthesia (in the range of 1.5–2 h). An alternative admin-
istration regime for dexmedetomidine has been proposed to extend
this period [19]. We have found significant differences between

different rats and mice strains. So we do recommend a pilot study
in order to choose the most convenient anesthesia for each particu-
lar model.
3.2 Electric Stimulating electrodes dedicated to MRI experiments have been

Stimulation developed based on existing protocols [20]. Previous studies have
shown the utility of iridium [21] or platinum-iridium electrodes
[22] for this purpose. Nevertheless, these electrodes produce large
susceptibility artifacts around the electrode’s location, especially
patent in EPI acquisition, precluding the possibility to study func-
tional responses in the area close to the implant. To overcome this
problem, we have introduced glass-coated carbon fiber bipolar
electrodes in our setup, which present several advantages: most
importantly the absence of susceptibility artifacts in the acquired
brain images, but also the possibility to produce very thin bipolar
electrodes (up to 7 μm tip diameter) [18].
To prepare carbon fiber electrodes, we use bundles of fibers
inserted into a theta-shaped glass capillary previously pulled to form
7 mm long pipettes with 200 μm tip diameter and adjusted to
produce an electrical impedance of 40–65 kΩ (see Note 8). A
regular wire with a pin connector is attached to the pipette,
connected to the carbon fibers using silver conductive epoxy
resin, and isolated with clear epoxy resin [18].
Depending on the configuration used in the MRI, the glass
electrode can be bent in order to accommodate the receiver coil,
minimizing its distance to the brain and maximizing the signal-to-
noise ratio (SNR) (see details in Subheading 3.4).
3.3 Stimulation In previous work applying electric-stimulation fMRI to study the

Protocols frequency response of the perforant pathway, the major neocortical
input to the hippocampus [23], we showed the existence of an
activity threshold to elicit a detectable fMRI response. More specif-
ically, we showed that (1) a certain level of activity, in an approxi-
mately constant population of neurons, must be reached in order to
start a detectable BOLD signal, (2) the activity-threshold for
BOLD elicitation can be reached by applying trains of pulses at
relatively low frequencies (4–5 Hz for the perforant path), (3) once
the threshold is crossed, the BOLD signal (magnitude and exten-
sion of the activation) is linearly correlated with the stimulating
current, (4) at current intensities evoking a half-maximal neuronal
spiking response, the activity spreads polysynaptically, with increas-
ing stimulation frequencies up to 20 Hz. Thus, stimulation proto-
cols often consisted in 6–10 trains of electrical pulses (100 μs
biphasic pulses) repeated every 30–60 s (total duration of the trial
180–600 s) and trials repeated three to five times per condition.
The duration of the stimulation train can be adjusted to the specific
needs, but durations between 2 and 6 s at frequencies ranging from
fMRI
Stim.
ON OFF
Dummy scans (not stored)
Acquisition of fMR images
Stimulation train ON Stimulation pulses
Fig. 1 Scheme representing the image acquisition (up) and the stimulus (down) during the time course of an
fMRI experiment. Green squares represent the time for dummy scans at the beginning of the acquisition
4 to 20 Hz produce BOLD responses of excellent amplitude (larger

than 4% change) in a variety of preparations [18, 22–25]. Off
periods between stimulation trains sufficiently long as to allow a
full recovery of the hemodynamic response (25–30 s in rats and
mice) increase the SNR of the response and the statistical power of
the analysis. A good coordination between image acquisition and
timing of stimulus presentation is necessary and easily achievable
using the TTL signals generated by the imaging protocols to syn-
chronize the pulse generator (see Fig. 1). Duration of the stimula-
tion train, pulse shape and intensity, frequency, and any
other stimulus parameter can be systematically varied for specific
purposes [26]. Within each EPI acquisition, it is advisable to
acquire long-enough baselines (4–8 volumes) before the first stim-
ulation train that will facilitate posterior quantifications of BOLD
signal change.
3.4 Intracerebral Most of the BOLD based fMRI experiments are acquired using EPI
Electrode Implant images, which are very sensitive to T2* changes. Practically
for fMRI speaking, a number of factors can confabulate causing a deteriora-
tion of the quality of the functional images. When working with
surgically manipulated animals, especially in acute preparations,
extreme care has to be taken to minimize bleeding. After the
surgery and before cementing the implant to the skull (see below)
thorough cleaning of the exposed cranium is mandatory.
In order to improve the SNR, the tip of the electrode is bent
(using a burner and some forceps) to form a 90 angle, so it could
go inside the brain leaving the main body of the pipette outside
parallel to the head of the rat, minimizing the implant’s height and
allowing a closer proximity between the MRI array coil and the
head of the animal.
The method described here is based on standard procedures
used in electrophysiological experiments with rats adapted to the
MRI requirements. Similar protocols are used for mice.
1. Weigh the animal.
2. Dissolve urethane in sterile 0.9% saline. Warm it to room
temperature before injection.
3. Place the animal in an induction chamber, and induce anesthe-

sia with 4% isoflurane in 100% oxygen (1 L/min). Wait until
the animal is superficially sedated (see Note 9).
4. Inject urethane intraperitoneally (1.3 g/Kg dose for rats and
1.5 g/kg dose for mice, see Note 10). Wait until the total
absence of withdrawal reflexes. Induction time is heteroge-
neous across animals and strains. If after 1 h of the initial dose
the animal shows reflexes, additional doses (10–20% of initial
dose) can be injected. In our experience, adjustment of the
initial dose is necessary for different strains. The slow process of
induction of anesthesia ensures a steady-state anesthesia (with
stable vital constants) during more than 8 h.
5. Upon induction, place the animal in a heating pad to maintain
the animal’s body temperature at 37 C. Use a rectal tempera-
ture probe with lubricating jelly to monitor the temperature.
6. Shave the head around the incision area.
7. When the animal reaches an appropriate level of anesthesia, fix
the animal’s head in a stereotaxic frame.
8. Inject subcutaneous local anesthetic (200 μL Bupivacaine,
0.5%) in the incision points, using a 1 cc syringe with a 20G
needle tip.
9. In order to prevent eye damage, employ ophthalmic gel on
each eye. The gel needs to be reapplied during the surgery to
ensure that eyes are covered at all times.
10. Make an incision (1 cm long) at the top of the head by
pressing firmly with a scalpel in an anteroposterior direction.
Remove excess skin to expose the skull. Cauterize the skin rims,
avoiding burning the skull to prevent image artifacts, and
apply hydrogen peroxide to remove any source of bleeding
(see Note 11).
11. Calculate the goal stereotaxic coordinates. Modern tools have
been developed to facilitate electrode localization [27].
12. Once the target site is located, trephine holes are made using a
manual drill. First-time used drill-bits require deep cleaning to
remove metal traces that can detach and enter the craniotomy
producing large image artifacts.
(a) Carefully rotate the drill bit over the skull until achieving a
circular craniotomy (2 mm diameter).
(b) Delicately pinch the dura using a curved 25G needle. For
mice, it is better to avoid this step to minimize bleeding.
The dura can be broken directly when introducing care-
fully the electrode in step 13.
(c) Add saline to the craniotomy once dura is pierced to avoid
dryness.
13. Slowly lower the MRI-compatible electrode until it reaches the

desired ventral coordinate (see Notes 12 and 13). If the dura is
not broken by the electrode in the case of mice, do not force it
(the electrode can be broken or the brain damaged). Go back
to step 12(b) and pinch it with the needle.
14. Deeply clean the skull using a dry cotton swab, eliminating any
bleeding. Add a small drop of tissue adhesive to seal the tissue,
and wait for 10 min.
15. Fix the electrode with several layers of acrylic.
(a) Pay special attention to the first layer of acrylic. This is the
critical one and should cover as much surface of the skull,
including the electrode, as possible. Use less viscous
cement in this first layer than in next ones. Wait until it
the layer is completely dry (20 min) before applying
more cement.
(b) In mice the first layer is crucial. The skull is smaller com-
pared with rats, and the cement has a smaller surface to
weld in. Thus, the electrode is more exposed being sus-
ceptible of movements and/or breakages. In order to
minimize these, we recommend generously covering the
section of the electrode closer to the mouse’s head using
acrylic.
(c) Apply extra layers around the implant until fully embed-
ding it, preventing post-surgical movement and protect-
ing it during the transport of the animal to the MRI room.
Avoid contact of the cement with the skin during the
whole process. This precaution will minimize the proba-
bility of bleeding during the imaging session.
16. When the cement is completely dry, remove the electrode
holder.
17. Detach the animal from the stereotaxic frame and place it in a
transfer cage. In order to avoid temperature dropping, main-
tain the animal in the heating pad until its transport to the
imaging facility.
3.5 fMRI In order to prevent temperature drop, preheat the magnet’s heat-
ing system before the animal arrives to the facility.
1. Place the animal on the MRI bed.
2. Check that the correct level of anesthesia has been maintained.
3. Insert the rectal temperature probe using lubricating jelly and
tape it in place.
4. Fix the animal’s head in an MRI-compatible stereotaxic device.
5. Place the physiological monitoring device (MRI-compatible
sensor with foot clip) or the breathing piezoelectric sensor.
6. Cover the exposed skull and the implant with agarose, with
special emphasis in filling all the possible empty spaces between
the head of the animal and the coil (see Note 14).
7. Connect the electrode to the current source.
8. Using the oscilloscope, cross-check the impedance of the elec-
trode that should match the value obtained during its fabrica-
tion, discarding a possible breakdown in the process of
implantation or during the accommodation of the animal in
the MRI setup.
9. Fix the coil in the MRI bed over the head of the animal, as close
as possible to the skull (see Note 15).
10. Place the animal inside the RF coil aligning the approximate
center of the brain with the magnet isocenter.
11. Acquire T2-weighted anatomical images in the three orthogo-
nal planes.
12. Even when the animal positioning is accurate, there can be
small inter-animal differences when defining an exact position.
In order to do grouped analyses, it is interesting to minimize
this variability. Thus, we recommend using anatomical land-
marks to position EPI slices always in the same orientation. A
possible strategy is:
(a) Take the plane that cut the base of cerebellum and the
anterior commissure (see Fig. 2a).
(b) Take the midline plane that separates the brain in left and
right hemispheres (Fig. 2b, c).
(c) Use the above anatomical planes to define the angle and
positioning of the slices for functional imaging. In our
case, 15 slices are positioned perpendicular to the planes
with the sixth more anterior slice containing the anterior
commissure (Fig. 2d).
13. Use a shimming procedure to adjust field homogeneity in the
brain. In our case, we use the MapShim macro implemented by
Bruker.
14. Adjust the EPI images according to the landmarks mentioned
in step 12 (see Note 16) and use saturation slices around the
brain. Acquire a set of EPI images without stimulation to check
proper image acquisition (no folding, ghosts, etc.)
15. Acquire an anatomical image with the same geometry than the
EPI images but higher (at least double) in plane spatial resolu-
tion. It will help to identify anatomical landmarks and
co-registration of brain templates for grouped analysis.
16. Start data acquisition.
Fig. 2 Position of selected slices using anatomical landmarks. Sagittal image (a) is used to calculate the angle
of the plane (pink) defined by the anterior commissure and the base of the cerebellum. Horizontal (b) and
coronal (c) images are used to calculate the plane (pink) that goes through the midline. Finally, EPI planes for
functional imaging (d) are positioned perpendicular to the anatomical planes calculated previously. To assure
a similar anteroposterior location of the EPI images across animals, an anatomical landmark is also used, in
our case the sixth most anterior slices is located on the anterior commissure (d, yellow slice)
3.6 Data Analysis The development of either commercial or open-source software

tools for fMRI analysis has greatly facilitated the applicability of
fMRI and has contributed to its massive widespread. Nevertheless,
due to the complex mathematical work behind the generation of
brain activation maps, it is important to apply proper and robust
statistical methods, e.g., to avoid false positives [28]. For a deep
discussion about the analysis see [29, 30]. The workflow for the data
analysis used in our laboratory implies linear detrending, temporal

(0.015–0.2 Hz) and spatial filtering (3 3 full width at half
maximum gaussian kernel of 1.5 sigma) of voxel time series, a
general linear model or cross-correlation analysis with a simple
boxcar model shifted forward in time (typically by the employed
TR), or a boxcar convolved with the hemodynamic response func-
tion (HRF). Typical functional maps obtained in one of our
electric-stimulation fMRI experiments are shown in Fig. 3.
4 Notes
1. Alternatively, breathing rate can be monitoring alone using a

simple custom designed piezoelectric device (sensitive to pres-
sure) positioned in the chest of the animal.
2. Usually, water bath is non-MRI compatible, so it has to be
positioned outside the 5 Gauss security area. The bath is
connected to the water blanket through two 5 m long silicon
tubes.
3. The automatic temperature control system is based on the
Arduino microcontroller (Arduino MEGA 2560, Arduino S.r.
l., Italy). To maintain the physiological temperature of the
animal stable automatically, a PID (Proportional-Integral-
Derivative) has been developed. The microcontroller obtains,
through serial communication (using a RS232 Shield V2, Link-
Sprite Technologies, Inc., Longmont, CO), the temperature of
the animal from the signal conditioner and it generates a con-
trol action that is transmitted to the thermostat to control the
temperature of the fluid pumped to the bed. This control
system allows automatic temperature control, maintaining
almost stable the temperature of the body of the animal being
scanned. The technician can interact with the automatic tem-
perature control system through a keypad and a display, being
able to set the desired temperature for the animal.
4. Agarose can be prepared in deuterated water, so it will be
invisible for MRI. Nevertheless, based on our experience,
there are no benefits in terms of fMRI maps acquisition.
5. Sequence parameters for GE-EPI images: field of view (FOV),
25 25 mm; slice thickness, 1 mm; 15 slices; matrix, 96 96;
segments, 1; flip angle, 60 ; echo time (TE), 15 ms; repetition
time (TR), 2000 ms; and four dummy scans.
6. Alternatively you could use a Spin Echo (SE) sequence with
similar parameters. There is extensive literature reviewing the
impact of the employed sequence methodology in the obtained
fMRI results [6]. Briefly, SE is more specific to
A Example of evoked functional activation maps (rat)
Corr.
5 mm
0
B Example of cerebral BOLD signal (from A)

5
4
3
BOLD (%)
2
1
0
-1
-2
0 20 40 60 80 100 120 140 160 180 200 220 240 260 280 300
Time (s)
Fig. 3 Brain-wide functional connectivity of dorsal CA3. (a) fMRI BOLD map overlaid on anatomical
T2-weighted images. Color-code denotes the correlation of the BOLD signal with the stimulation protocol
convolved with a hemodynamic function (see text). The arrow shows the artifact caused by the carbon
electrode. Arrowheads show agarose. (b) Global BOLD signal time course of the significantly (P < 0.001)
activated voxels in response to ten trains of stimulation (indicated by vertical gray bars), delivered at 10 Hz.
Graph shows the mean of 3 repetitions of the same protocol
microvasculature changes but less sensitive whereas GE is more

influenced by changes in macrovasculature but overall more
sensitive.
7. Sequence parameters for RARE images: FOV, 25 25 mm;
slice thickness, 1 mm; 15 slices; matrix, 192 192; RARE
factor, 8; effective TE (TEeff), 56 ms; TR, 2000 ms.
8. Small electrode tips will cause less tissue damage during
implantation, but the higher electric impedance would require
higher voltages to inject a same amount of current and there-
fore the possibility to overheat and damage the tissue. Thus,
there is a compromise between these two parameters. In our
experience, an electrode tip of 200 μm render good stimula-
tion while minimizing tissue damage.
9. Some animal strains are more susceptible to anesthetics mixture
and can be affected by the interaction of urethane and isoflur-
ane. In this case, we recommend injecting the animal without
previous exposure to other anesthetic.
10. Usually, urethane is injected at doses in the range of 1.2–1.4 g/
Kg for rats and 1.4–1.6 for mice.
11. In our experience, when working with mice, the complete
removal of the skin over the skull significantly increases the
quality of the fMRI images. To do that, gently cut with surgical
scissors the skin over the head. Remove the excess skin, cauter-
ize borders (with extreme caution to avoid overheating the
skull) and apply hydrogen peroxide to clean the area. Carefully
remove any trace of blood.
12. For instance, to stimulate the CA3 region of the dorsal hippo-
campus in the rat the coordinates, referenced to Bregma, are:
3.5 mm anteroposterior and 3.6 mm lateral, initial position
3.8 mm ventral to the dural surface, based on [31].
13. The same procedure can be followed to insert a recording
electrode in the desired area to be sure about stimulation
electrode positioning. Nevertheless, care must be taken when
placing the recording electrode to minimize brain damage.
Ideally, this recording electrode should be placed in a region
not fundamental for the fMRI study.
14. EPI images are highly sensible to abrupt changes in magnetic
susceptibility originating artifacts in the border where the vari-
ation occurs. Due to the specific configuration used in our set
up, when the phase array coil is positioned, there is an empty
space between the surface coil and the head of the animal. We
fill this space with agarose using a syringe previously filled with
agarose 0.5%.
15. Avoid excessive pressure between the coil and electrode. Usu-
ally electrodes are fragile and break easily during the experi-
ment if there is some of pressure on them.
16. The employed FOV usually exceeds the cross section of the
subject to prevent artifacts from image folding. The slice thick-
ness is 1 mm for rats and 0.8 mm for mice, but depending on
the SNR and the expected level of activation, it could be
decreased.
Acknowledgements
This work was supported by the Spanish Ministerio de Economı́a y

Competitividad (MINECO) and FEDER funds under grants
BFU2015-64380-C2-1-R (S.C.) and BFU2015-64380-C2-2-R
(D.M.) and EU Horizon 2020 Program 668863-SyBil-AA grant
(S.C.). S.C. acknowledges financial support from the Spanish State
Research Agency, through the “Severo Ochoa” Programme for
Centres of Excellence in R&D (ref. SEV- 2013-0317).
References
1. Crosson B, Ford A, McGregor KM, 7. Masamoto K, Kanno I (2012) Anesthesia and

Meinzer M, Cheshkov S, Li X, Walker-Batson- the quantitative evaluation of neurovascular
D, Briggs RW (2010) Functional imaging and coupling. J Cereb Blood Flow Metab 32
related techniques: an introduction for rehabil- (7):1233–1247. https://doi.org/10.1038/
itation researchers. J Rehabil Res Dev 47(2): jcbfm.2012.50
vii–xxxiv 8. Khubchandani M, Mallick HN, Jagannathan
2. Ogawa S, Lee TM, Kay AR, Tank DW (1990) NR, Mohan Kumar V (2003) Stereotaxic
Brain magnetic resonance imaging with con- assembly and procedures for simultaneous
trast dependent on blood oxygenation. Proc electrophysiological and MRI study of con-
Natl Acad Sci U S A 87(24):9868–9872 scious rat. Magn Reson Med 49(5):962–967.
3. Logothetis NK (2008) What we can do and https://doi.org/10.1002/mrm.10441
what we cannot do with fMRI. Nature 453 9. King JA, Garelick TS, Brevard ME, Chen W,
(7197):869–878. https://doi.org/10.1038/ Messenger TL, Duong TQ, Ferris CF (2005)
nature06976 Procedure for minimizing stress for fMRI stud-
4. Moreno A, Jego P, de la Cruz F, Canals S ies in conscious rats. J Neurosci Methods 148
(2013) Neurophysiological, metabolic and cel- (2):154–160. https://doi.org/10.1016/j.
lular compartments that drive neurovascular jneumeth.2005.04.011
coupling and neuroimaging signals. Front 10. Ferris CF, Febo M, Luo F, Schmidt K,
Neuroenerg 5:3. https://doi.org/10.3389/ Brevard M, Harder JA, Kulkarni P,
fnene.2013.00003 Messenger T, King JA (2006) Functional mag-
5. Jego P, Pacheco-Torres J, Araque A, Canals S netic resonance imaging in conscious animals: a
(2014) Functional MRI in mice lacking new tool in behavioural neuroscience research.
IP3-dependent calcium signaling in astrocytes. J Neuroendocrinol 18(5):307–318. https://
J Cereb Blood Flow Metab 34 doi.org/10.1111/j.1365-2826.2006.01424.
(10):1599–1603. https://doi.org/10.1038/ x
jcbfm.2014.144 11. Tennant DA, Duran RV, Gottlieb E (2010)
6. Greve JM (2011) The BOLD effect. Methods Targeting metabolic transformation for cancer
Mol Biol 771:153–169. https://doi.org/10. therapy. Nat Rev Cancer 10(4):267–277.
1007/978-1-61779-219-9_8 https://doi.org/10.1038/nrc2817
12. European Convention for the Protection of further evidence for the lack of electrically
vertebrate animals used for experimental and induced polysynaptic propagation of activity
other scientific purposes (2006) Appendix in the neocortex. Magn Reson Imaging 29
A. Guidelines for accommodation and care of (10):1374–1381. https://doi.org/10.1016/j.
animals (Article 5 of the Convention) mri.2011.04.005
13. Hendrich KS, Kochanek PM, Melick JA, Schid- 22. Alvarez-Salvado E, Pallares V, Moreno A,
ing JK, Statler KD, Williams DS, Marion DW, Canals S (2014) Functional MRI of long-term
Ho C (2001) Cerebral perfusion during anes- potentiation: imaging network plasticity. Philos
thesia with fentanyl, isoflurane, or pentobarbi- Trans R Soc Lond Ser B Biol Sci 369
tal in normal rats studied by arterial spin- (1633):20130152. https://doi.org/10.
labeled MRI. Magn Reson Med 46 1098/rstb.2013.0152
(1):202–206 23. Canals S, Beyerlein M, Murayama Y, Logothe-
14. Schroeter A, Schlegel F, Seuwen A, tis NK (2008) Electric stimulation fMRI of the
Grandjean J, Rudin M (2014) Specificity of perforant pathway to the rat hippocampus.
stimulus-evoked fMRI responses in the Magn Reson Imaging 26(7):978–986.
mouse: the influence of systemic physiological https://doi.org/10.1016/j.mri.2008.02.018
changes associated with innocuous stimulation 24. Godino Mdel C, Romera VG, Sanchez-
under four different anesthetics. NeuroImage Tomero JA, Pacheco J, Canals S, Lerma J,
94:372–384. https://doi.org/10.1016/j. Vivancos J, Moro MA, Torres M, Lizasoain I,
neuroimage.2014.01.046 Sanchez-Prieto J (2013) Amelioration of ische-
15. Sonnay S, Just N, Duarte JM, Gruetter R mic brain damage by peritoneal dialysis. J Clin
(2015) Imaging of prolonged BOLD response Invest 123(10):4359–4363. https://doi.org/
in the somatosensory cortex of the rat. NMR 10.1172/JCI67284
Biomed 28(3):414–421. https://doi.org/10. 25. Hadar R, Vengeliene V, Barroeta Hlusicke E,
1002/nbm.3263 Canals S, Noori HR, Wieske F, Rummel J,
16. Paasonen J, Salo RA, Shatillo A, Forsberg MM, Harnack D, Heinz A, Spanagel R, Winter C
Narvainen J, Huttunen JK, Grohn O (2016) (2016) Paradoxical augmented relapse in
Comparison of seven different anesthesia pro- alcohol-dependent rats during deep-brain
tocols for nicotine pharmacologic magnetic stimulation in the nucleus accumbens. Transl
resonance imaging in rat. Eur Neuropsycho- Psychiatry 6(6):e840. https://doi.org/10.
pharmacol 26(3):518–531. https://doi.org/ 1038/tp.2016.100
10.1016/j.euroneuro.2015.12.034 26. Tehovnik EJ, Tolias AS, Sultan F, Slocum WM,
17. Maggi CA, Meli A (1986) Suitability of ure- Logothetis NK (2006) Direct and indirect acti-
thane anesthesia for physiopharmacological vation of cortical neurons by electrical micro-
investigations in various systems. Part 2: Car- stimulation. J Neurophysiol 96(2):512–521.
diovascular system. Experientia 42 https://doi.org/10.1152/jn.00126.2006
(3):292–297 27. Pallares V, Moya J, Samper-Belda FJ, Canals S,
18. Moreno A, Morris RG, Canals S (2016) Moratal D (2015) Neurosurgery planning in
Frequency-dependent gating of hippocampal- rodents using a magnetic resonance imaging
neocortical interactions. Cereb Cortex 26 assisted framework to target experimentally
(5):2105–2114. https://doi.org/10.1093/ defined networks. Comput Methods Prog
cercor/bhv033 Biomed 121(2):66–76. https://doi.org/10.
19. Pawela CP, Biswal BB, Hudetz AG, Schulte 1016/j.cmpb.2015.05.011
ML, Li R, Jones SR, Cho YR, Matloub HS, 28. Eklund A, Nichols TE, Knutsson H (2016)
Hyde JS (2009) A protocol for use of medeto- Cluster failure: why fMRI inferences for spatial
midine anesthesia in rats for extended studies extent have inflated false-positive rates. Proc
using task-induced BOLD contrast and Natl Acad Sci U S A 113(28):7900–7905.
resting-state functional connectivity. Neuro- https://doi.org/10.1073/pnas.1602413113
Image 46(4):1137–1147. https://doi.org/ 29. Poldrack RA, Mumford JA, Nichols TE (2011)
10.1016/j.neuroimage.2009.03.004 Handbook of functional MRI data analysis.
20. Shyu BC, Lin CY, Sun JJ, Sylantyev S, Chang C Cambridge University Press, New York
(2004) A method for direct thalamic stimula- 30. Ashby FG (2011) Statistical analysis of FMRI
tion in fMRI studies using a glass-coated car- Data. MIT Press, Cambridge, MA
bon fiber electrode. J Neurosci Methods 137 31. Paxinos G, Watson C (2007) The rat brain in
(1):123–131. https://doi.org/10.1016/j. stereotaxic coordinates. Academic Press, Else-
jneumeth.2004.02.015 vier, New York
21. Sultan F, Augath M, Murayama Y, Tolias AS,
Logothetis N (2011) esfMRI of the upper STS:
Chapter 9
Functional Diffusion Magnetic Resonance Imaging

Rita Maria Rocha Oliveira, Irene Guadilla, and Pilar López-Larrubia
Abstract
Functional diffusion magnetic resonance imaging (fDMRI) is a noninvasive technique that allows elucidat-
ing physiological and anatomical changes at a microscopic scale by detection of water molecular displace-
ments in tissue structures. These displacements likely reflect microstructural changes associated with
neuronal or glial cells activation. In this chapter, we will describe the physical and biological concepts of
fDMRI and how images of brain activation can be acquired in a preclinical setup.
Key words Magnetic resonance imaging, Functional imaging, Diffusion imaging, Diffusion biexpo-
nential parameters
1 Introduction
In the past decades, functional magnetic resonance imaging (MRI)

techniques have been used extensively in clinical settings and
research centers. In a noninvasive way, the use of such methods
provides new information’s about brain functions of physiological
and anatomical changes. Diffusion MRI is one of the most recent
techniques with functional applications used in the study and diag-
nosis of neurological abnormalities [1–3].
To explain the concept of diffusion MRI, we will briefly go over
the principles of water diffusion. Well established by A. Einstein
[4], the phenomenon of diffusion refers to the mobility of particles,
classically called Brownian motion, due to the thermal energy
carried by these particles. Regarding on this motion, water mole-
cules in a free medium diffuse with an isotropic behavior traveling
randomly all around. However, in tissues this mobility is highly
influenced by the microenvironment because, due to the cellular
components such as cells membranes, fibers, and macromolecules,
the randomly mobility is impeded, and the water molecules display
an anisotropic diffusion. The wide analysis of the physical proper-
ties of water at a microscopic level is the base of diffusion MRI
measurements [5, 6]. Taking this into account, combining MRI
135
136 Rita Maria Rocha Oliveira et al.
Fig. 1 Diffusion-weighted spin-echo Stejskal-Tanner sequence. Diffusion sensitive can be modulated by

changing pulsed field gradient parameters—δ: gradient pulse duration, Δ: gradient pulse separation, G:
gradient strength
principles with nuclear magnetic resonance physics, it became pos-

sible to evaluate the diffusion of water molecules in biological
tissues, commonly called diffusion-weighted imaging (DWI)
methodology.
Stejskal and Tanner [7] described one of the most acceptable
methods for measuring molecular diffusion with magnetic reso-
nance. They demonstrated that molecular displacement distribu-
tions occurring during an interval of time could be detected by a
spin-echo (SE) experiment with an additional pair of pulsed field
gradients applied before and after the 180 pulse (Fig. 1). The first
gradient is applied to cause instantaneous phase shift in the nucleus
along the direction of application (whatever can be chosen). The
second gradient is applied with the same magnitude but opposite
direction than the first one to reverse the changes made by it. With
this boundary condition, we can detect encoding water molecules.
If there are water molecule motions in-between the application of
the gradients, the spins of hydrogens will be displaced and an
incomplete rephasing with signal attenuation is observed. Never-
theless, in the absence of diffusion all the spins will be phasing in the
nuclear positions recorded by the first gradient pulse, and the final
coherence will remain the same.
The eco signal intensity acquired depends on the strength (G),
the duration of the gradients (δ), their temporal separation (Δ), and
the apparent diffusion constant (ADC) along the direction of the
field gradient. Therefore, just by fitting the echo signal to a mono-
exponential function (Eq. 1), it is possible to build diffusion
Functional Diffusion MRI 137
coefficients maps by measuring the echo attenuation in each three-

dimensional element (voxel) [8].
S ðb Þ
¼ exp γ 2 G 2 δ2 ðΔ δ=3Þ ADC ð1Þ
S ð0Þ
As mentioned above, in brain the water molecules interact with

tissue components and the diffusion process is not the same if the
water can move freely or the molecules are linked to cellular com-
ponents. This strongly suggests that diffusion motion cannot be
modeled by a single Gaussian distribution and the data collection
from the signal attenuation should not be treated as a monoexpo-
nential function. Several non-monoexponential analysis models
have been proposed to describe the diffusion-related signal decay
[2, 3, 5, 9–11]. From these, the one that is more accepted is the
biexponential model that perform a curve fitting based on the
following equation,
S ðb Þ
¼ SDP:e b:Dslow þ FDP:e b:Dfast ð2Þ
S ð0Þ
where S(b) reflects the signal decay in the presence of diffusion

sensitization, S(0) represents the echo signal obtained with a zero
value of diffusion gradient, Dslow and Dfast represent the slow and
fast diffusion coefficients, with their corresponding contributions
to the signal from slow- and fast-diffusion phases, SDF and FDP,
respectively, and b is the b-value defined by the gyromagnetic
constant of protons (γ), the strength, duration and the delayed
time (Δ δ/3) of the magnetic field gradient pulses, [γ 2G2δ2(Δ
δ/3)], and reflects the degree of diffusion sensitizing of the MRI
acquisition.
By selecting the gradient direction, we can acquire different
diffusion coefficient values. In addition, the DWI experiment
involves the acquisition of consecutive images at different
b-values. At high b-values the movements between fast and slow
diffusion are better distinguished, but the signal-to-noise ratio is
decreased. Moreover, the application of strong pulses can improve
diffusion sensitization to exceed local magnetic field inhomogene-
ities that in vivo are usually present in the tissues [12].
This model assumes that motions from water molecules linked
to the cellular components are slow, and those movements far away
from those represent fast water molecules. This conception is based
on the assumption that in tissue microstructures, both in extra- and
intracellular compartments, there are water molecules moving in
the vicinities of membranes and molecules moving outlying these
structures. Accordingly, we will have a slow diffusion coefficient
that reflects the movements from the linked-water and a fast diffu-
sion coefficient sensitized to molecular displacements from mole-
cules in solution [13]. There are, however, others biexponential
behaviors [14–16] that have been published but from our point of
view that one is the most correct approach to interpret molecular
water motion as a sensitive marker to study changes associated with
physiological and pathological states.
Some authors have demonstrated changes in water coefficients
diffusion during neural activation. In their investigations, they
quantified slow and fast diffusion phases and they observed micro-
structural properties changes that are probably linked to volume
variations and cell swelling. Such cell size variations were inter-
preted as a reflection from the neuronal soma, dendritic areas, or
also glial cells [17–19]. These results strongly suggest that diffusion
MRI can be used as a powerful technique to monitor neural activa-
tion. Based on this, the biexponential DWI analysis brings forward
a new approach to produce brain images directly associated with
neural activation.
2 Materials
The following materials are used in preclinical functional DWI:
2.1 MRI System 1. MR high-field horizontal magnet.

3. RF coils. It can be used two basic MRI coil setups:
(a) Transceiver coil setup. The same volume coil of around
35–40 mm inner diameter for rats and 25–30 mm inner
diameter for mice excite and receive the MR signal.
(b) Active decoupled transmit and receive coil setup. Volume
transmit coil and a surface coil brain rat/or mouse make
the excitation and reception of the signal.
2.2 Other Equipment, 1. Monitoring System: Small animal monitoring and gating sys-
Devices, and Materials tem with a rectal sensor and a pneumatic pillow placed in the
animal abdomen that continuously reports data of body tem-
perature and respiratory rate. This allows to monitor the state
of the anesthetized animal inside the magnet and controlling
the level of anesthesia.
2. Anesthesia device: It is formed by an anesthetic vaporizer, a
flowmeter, a poly(methylmethacrylate) (PMMA) chamber, a
nose cap, gases circulating tubes, and a filter. The small rodent
inhalation anesthesia equipment allows to induced and control
the anesthetized state of the animal.
3. Thermal blanket: Anesthetized animals are placed on a water
heated blanket to maintain the body temperature at approx.
37 C during scanning (any other heating setup can be used).
4. Animal holders: Different beds/holders of PMMA (or any

other MR compatible material) are used to place the animal
inside the magnet. The holder will be slid to place the region of
interest in the magnet isocenter.
2.3 Software 1. Acquisition: Commercial software is usually supplied by the

same distributor of the MR magnet.
2. Processing: The processing of images can be done using com-
mercial software as Matlab (The MathWorks Inc., Natick, MA,
USA) and IDL (Iterative Data Language, Research System,
Boulder, CO, USA) or with open source software as ImageJ
[20] (https://imagej.nih.gov/ij/) and R [21] (https://www.r-
project.org).
Table 1.
3 Methods
The following method presented to obtain functional diffusion-

weighted images is optimized for a 7T Bruker Biospect® MRI
system operating under Paravision 5.1 software (in a Linux envi-
ronment) and with a Bruker AVANCE III console, also for rats and
mice.
3.1 Pre-scan Animals are anesthetized in a PMMA chamber with isoflurane/O2

(3–4%) before placing them in an adequate holder to be introduced
inside the magnet. By monitoring the animal respiration, anesthesia
is maintained and controlled during the whole acquisition (see Note
1). The body temperature is monitored with a rectal thermometer
and maintained at around 37 C by using the thermal blanket
(or any other appropriate device). Animal movements will cause
artifacts and originate processing errors in the calculated parametric
images, so, the animal head has to be immobilized during the
acquisition with ears and tooth bars and/or fix tape over the
head. Put some eye ointment on the eyes of the animal to avoid
damaging the cornea in a prolonged study.
If it is possible, the coil has to be tuned once the animal is in the
magnet.
3.2 Scan Acquire a position image of the animal in the three spatial orienta-
tions (axial, coronal, and sagittal) by using a fast-low angle shot
3.2.1 Localization/
(FLASH) sequence (these sequences are called Tripilot or Location
Positioning Images
in Bruker systems). This allows to fix the correct position inside the
magnet (see Note 2).
Table 1
Preclinical MRI system and associated equipment/devices for developing the functional DWI
protocols described in this chapter
MRI system Bruker Biospect® 7 T—16 cm bore (B-C 70/16)

Acquisition software Paravision 5.1 (Bruker Biospin, Ettlingen, Germany) running in a console Bruker
AVANCE III operated on a Linux platform
1
Rat-head transceiver H circular polarized transmit/receive birdcage volume coil, 40 mm inner
coil diameter
1
Mouse-head H circular polarized transmit/receive birdcage volume coil, 23 mm inner
transceiver coil diameter
1
Transmitter coil H circular polarized transmit-only volume coil, 72 mm inner diameter
1
Rat head surface coil H circular polarized with integrated combiner and preamplifier receive-only coil,
40 mm inner diameter
1
Mouse head surface H circular polarized with integrated combiner and preamplifier receive-only coil,
coil 23 mm inner diameter
Physiological Small Animal Monitoring and Gating System (Model 1025, SA Instruments, Inc.
monitoring NY)
Anesthesia Isoflurane (IsoFlo, Esteve labs, Barcelona, Spain) in O2 or air (standard calibrated
vaporizer)
Halogenated gas Activated charcoal canister packed (Harvard Apparatus Anaesthetic, Holliston,
filter MA, USA)
Thermal blanket Warmed circulating water blanket (Gaymar T/PUMP TP500 Heat Therapy
System)
Processing software “Diffusion,” homemade software application written in Matlab (The MathWorks
Inc., Natick, MA, 2000)
3.2.2 Morphological Employ a T2-weighted sequence to acquire structural images as

Images references for the diffusion-weighted parametric maps. For that, a
spin-echo (SE) or fast-SE sequence can be used. Do not forget to
acquire the images with the same geometrical information as diffu-
sion studies will be acquired.
3.2.3 Functional For obtaining functional diffusion images, DWI sequences avoid-
Diffusion Imaging ing low b-values have to be used. We present here three different
methods optimized in a Bruker Biospect® MRI system for acquir-
ing fDWI depending on the collection of the MR data (k-space
filling):
1. Standard acquisition. It is based on a SE sequence in which
each line of the k-space (echo signal) is filled at each phase
encoding step (Fig. 1). This method entails a very good signal-
to-noise ratio (SNR) but suffers of a long acquisition time,
increasing the possibility of animal movements that cause arti-
facts (see Note 3). The operator has to select the number of
spatial directions to apply the diffusion gradients.
2. Single-shot echo-planar imaging (ss-EPI). The whole k-space is
filled after the first 90 –180 pulse set in the SE sequence, and
each image acquisition takes the TR value. EPI acquisition is
much faster than standard one, nevertheless suffers from the
presence of artifacts (EPI ghost) and has low SNR. These
disadvantages make mandatory to achieve a very good shim-
ming of the magnetic field to reduce the artifacts associated to
the echo-planar readout (see above). Also, the adjustment of
the basic frequency need to be made previously the acquisition
of each diffusion EPI sequence.
3. Segmented EPI. In this case, K-space is filled in different shots
or segments, that is, several lines are collected after each
90 –180 pulse combination. This sequence has a better SNR
(see Note 4) but it is slower than the ss-EPI as many times as
segments selected. In this case, it is also crucial to get a good
shimming.
Before starting EPI DWI acquisition, it is necessary to shim
adequately the system to get the higher homogeneity of the mag-
netic field. In Bruker systems there are several automatic options to
achieve that:
1. Adjusting the first and second shim coil values to optimize the
shimming of the whole sample covered based on the signal
response after each modification.
2. Using the FASTMAP module implemented in Bruker scanners
[22]. This efficient tool allows to shimming in a localized cubic
region which is virtually divided in smaller sticks. The main
disadvantage of this method is that not always the volume of
interest can be covered by a cubic space.
3. Employing the MAPSHIM protocol that optimizes the mag-
netic field homogeneity in a rectangular region of arbitrary size
and orientation [23] after the acquisition of a 3D magnetic
field distribution.
Once the shim is improved, the DWI sequence is acquired after
selecting the b-values that will control the diffusion weighting in
the study (see Note 5). Figure 2 shows representative T2 and DW
images obtained in a diffusion study. Table 2 shows the parameters
of a common diffusion sequence (see Note 6).
3.3 Post-scan The values of fast and slow diffusion coefficients (FDC and SDC,
Analysis respectively) and fast and slow diffusion phases (FDP and SDP) can
Fig. 2 MRI study of a rat brain in axial orientation. T2-weighted (a) and diffusion-weighted (b) images of the
same slice. The acquisition parameters of the DWI sequence are showed in Table 2
Table 2
Acquisition parameters of a diffusion-weighted MRI study of a mouse brain in a Bruker Biospect® 7T
MRI system
Pulse sequence Spin echo (segmented EPI acquisition)

TE (echo time) 69.04 ms
TR (repetition time) 3000 ms
Averages 4
Segments 4
Repetitions 1
δ 4 ms
Δ 20 ms
b-Values 300, 450, 600, 750, 900, 1200, 1500, 1650, 1800 s/mm2
Diffusion gradient strength (percentage) 32.9, 40.3, 46.54, 52.04, 57.6, 65.8, 73.6, 77.2, 80.6%
In-plane resolution 0.273 mm/pixel
Matrix resolution 128 128
Slice thickness 1.5 mm (axial orientation)
Number of diffusion directions 3 orthogonal directions (1, 0, 0), (0, 1, 0), (0, 0, 1)
be calculated in a pixel-by-pixel base using the appropriate software.

Images have to be processed to generate the corresponding para-
metric images in each direction of study. In our case, we employ a
homemade application developed in MatLab (the MathWorks Inc.,
Natick, MA, 2000) that fits the MRI signal to a biexponential
distribution based on Eq. 2. Figure 3 shows coefficients and phases
maps of a three-orthogonal direction DWI study obtained after
computational processing.
Fig. 3 Functional diffusion MRI study carried out in a rat model of paclitaxel induced-peripheral neuropathy.
Parametric images shown correspond to the application of the diffusion gradient in the left-right (L–R)
direction, and acquired in axial orientation. (a) Diffusion-weighted image, (b) fast diffusion coefficient (FDC)
map, (c) slow diffusion coefficient (SDC) map, (d) slow diffusion phase (SDP) map. The acquisition parameters
of the DWI sequence are showed in Table 2
From parametric maps the user can measure the coefficients

and phases of every pixel, or a mean value by selecting a region of
interest (ROI), with an appropriate image analysis software such as
ImageJ [20]. Finally, data can then be statistically analyzed with
programs like SPSS (IBM, Armonk, North Castle, New York,
USA), Excel (Microsoft, Redmond, Washington, USA), or R [21].
3.4 Applications Functional diffusion MRI is a recent approach for assessing brain
functionality also in animal models and human being (see Note 7).
In our group we have previously used this methodology for analyz-
ing hypothalamic activity of healthy subjects under an appetite
paradigm, also in the mouse and human brain [17], and with
mice genetically deficient in leptin or neuropeptide Y [24]. Besides,
we characterized the hypothalamic metabolic compartmentation
during appetite regulation in mice [25], and carried out fDWI
studies in glioma bearing rats and an animal model of chemothera-
peutic induced-peripheral neuropathy (unpublished results).
This tool can be employed for studying specific activation path-
ways in the brain (using or not dedicated paradigm) and for
characterizing the cerebral functionality in a wide spectrum of

diseases developed in small animals, like tumors, brain injury, men-
tal disorders, or ischemia, among others. These studies can be
carried out in a resting state condition, to evaluate the basal neuro-
nal activation, or under a specific stimulus like feed, fasting, pain,
stress, etc. to analyze the brain response in physio- and pathological
status. In all cases, analysis of diffusion parametric maps (FDC,
SDC, FDP, and SDP) will allow identify cerebral regions in which
either an increase or a decrease of the diffusion coefficients—or
diffusion phases—takes place.
4 Notes
1. It is important to maintain a correct anesthesia level of the

animal during the MRI acquisition. Several considerations
have to be taken into account to obtain good results:
(a) The measurement of the respiratory rate is made with a
pneumatic pillow from the monitoring system. For accu-
rately doing this, the pillow has to be placed in a correct
position, right under the abdomen of the animal.
(b) Anesthesia vaporizer has to be properly connected by
tubes to the rest devices of the anesthetic circuit (animal
holder/nose mask and filter).
(c) The animal nose has to be well positioned in the nose
mask to properly allow the gases exchange (CO2 accumu-
lation can be dangerous).
(d) To avoid the accumulation of the non-metabolized iso-
flurane in the environment, the filter system has to be
switch on during all the experiment.
(e) The anesthetic conditions during the MRI acquisition are
usually 1.5–2% isoflurane dissolved in O2 at a flow of 1 L/
min, but they have to be controlled according to the
animal physiological status.
(f) To ensure that the animal will neither take a risk nor move
during the study the respiratory rate should be around
40–60 breaths/min.
(g) To compare results obtained from different experimental
groups, animals have to be anesthetized in the same con-
ditions and maintaining a similar physiological status.
2. Some parameters have to be adjusted (automatically or manu-
ally) before acquiring the first scan:
(a) Tuning and matching the coil—when it is possible—to
adjust its frequency of resonance and impedance of the
coil according to the animal.
(b) Shimming the magnet to improve its homogeneity

(c) Determining the RF power for the 90 pulse
(d) Basic frequency adjustment of the system
(e) Calculating the optimal receiver gain
3. The use of certain options can reduce the total acquisition time
by accelerating the phase encoding process:
(a) Partial Fourier-Transform (FT) acceleration. An asym-
metric truncation of the k-space allows to accelerate the
experiment. It maintains the nominal resolution although
it reduces the SNR. Its value goes from 1 (a full symmetric
filling of the k-space) to 2 (a half filling of the k-space).
The factor partial-FT overscans indicates the number of
lines on the negative half of the k-space that are sampled in
a partial-FT experiment.
(b) Zero-fill acceleration. As Partial Fourier-Transform (FT)
acceleration, it is a factor that oscillates between 1 and
2 but with a symmetric truncation of the k-space.
(c) PPI acceleration. Only possible when a phased array coil is
available, this option relies on the partially parallel imag-
ing (PPI). The number of channels gives its maximum
value which decreases the acquisition time as many times
as that value.
4. Regulation of respiration is very important to acquire the seg-
mented EPI sequences. The appearance of artifacts and ghost
due to multiple segment acquisition is avoided by triggering
the acquisition to the respiratory cycle because images have to
be acquired in the same equivalent point of the breath cycle. So,
the respiration of the subject should be as stable as possible.
5. In the fDWI experiments, b-values have to be selected depend-
ing on the interested diffusion weighting: a value too low
induces small dephasing of spins which means that the signal
decay will be low and mainly due to perfusion movements
(water molecules of blood in capillaries that feed the tissue).
So, diffusion cannot probably be accurately determined. A
b-value too high induces a big signal decay, in such a way that
signal intensity can reach the noise level. Usually, low b-values
are used to assess high diffusion regions while high b-values
better estimate slow diffusion regions. The described protocol
is for high b-values (corresponding to a diffusion gradient
strength from 30 to 85%) that avoid the blood flow contami-
nation of the MRI signal. Because a fitting of the signal to a
biexponential model has to be done, the higher number of
b-values, the better adjustment and the most accurate results.
6. The choice of acquisition parameters depends on the interests
of the experiment. DW images usually suffer from low SNR
due to long TE values and the diffusion gradient dephasing, so

some important considerations have to be taken into account:
(a) TR value around 5·T1 ensures the recovery of the initial
magnetization which translates in an optimal SNR.
(b) A better SNR is achieved with a minimum TE.
(c) Geometrical distortions because of EPI acquisition are
minimized by increasing spectral bandwidth which also
allows a lower TE.
(d) Basal images A0 (that is, a DW image with a b-value of
zero) has to be included in the experiment as a function of
the number of spatial directions in a 1:6 ratio.
(e) Although in theory a 100% of diffusion gradients strength
can be achieved, it is not recommended to exceed 85% to
protect the hardware.
7. Some relevant issues have to be considered before planning a
fDWI experiment. Studies need to be performed in the same
experimental conditions for all the animals involved to achieved
soundness results. To identify a different functionality due to a
pathology, comparisons have to be made between healthy and
diseased subjects. On the other hand, to detect the activation
due to a specific stimulus, the same animal has to be assessed
before and after being submitted to the stimulation.
At the same time, it is very important to previously identify
collateral considerations that can influence in the cerebral func-
tion. So, it is of crucial importance to do MRI acquisitions in
the same moment of the day, because circadian cycles strongly
influence in the activity of some cerebral regions. In the same
line, it is recommendable to measure blood glucose level in
animals before the study to properly signal variations due to a
different feeding status that also condition the neuronal
activity.
Acknowledgments
This work was supported by grant SAF2014-53739-R. RMRO

held a contract form the Programme Erasmus þ and IG held a
predoctoral contract from Ministerio de Economı́a, Indrustria y
Competitividad (MINECO) of Spain.
References
1. Padhani AR, Liu G, Koh DM, Chenevert TL, imaging as a cancer biomarker: consensus and
Thoeny HC, Takahara T, Dzik-Jurasz A, Ross recommendations. Neoplasia 11(2):102–125
BD, Van Cauteren M, Collins D, Hammoud 2. Le Bihan D (2007) The ‘wet mind’: water and
DA, Rustin GJ, Taouli B, Choyke PL (2009) functional neuroimaging. Phys Med Biol 52
Diffusion-weighted magnetic resonance
(7):R57–R90. https://doi.org/10.1088/ 15. Assaf Y, Freidlin RZ, Rohde GK, Basser PJ

0031-9155/52/7/r02 (2004) New modeling and experimental
3. Le Bihan D (2003) Looking into the func- framework to characterize hindered and
tional architecture of the brain with diffusion restricted water diffusion in brain white matter.
MRI. Nat Rev Neurosci 4(6):469–480. Magn Reson Med 52(5):965–978
https://doi.org/10.1038/nrn1119 16. Yacoub E, Uludağ K, Uğurbil K, Harel N
4. Einstein A (1956) Investigations on the theory (2008) Decreases in ADC observed in tissue
of the Brownian movement. Courier Corpora- areas during activation in the cat visual cortex
tion, Dover Publications, NY at 9.4 T using high diffusion sensitization.
5. Le Bihan D (1995) Molecular diffusion, tissue Magn Reson Imaging 26(7):889–896
microdynamics and microstructure. NMR 17. Lizarbe B, Benı́tez A, Sánchez-Montañés M,
Biomed 8(7–8):375–386 Lago-Fernández LF, Garcia-Martin ML,
6. Chilla GS, Tan CH, Xu C, Poh CL (2015) López-Larrubia P, Cerdán S (2013) Imaging
Diffusion weighted magnetic resonance imag- hypothalamic activity using diffusion weighted
ing and its recent trend—a survey. Quant magnetic resonance imaging in the mouse and
Imaging Med Surg 5(3):407–422. https:// human brain. NeuroImage 64:448–457
doi.org/10.3978/j.issn.2223-4292.2015.03. 18. Darquié A, Poline J-B, Poupon C, Saint-
01 Jalmes H, Le Bihan D (2001) Transient
7. Stejskal EO, Tanner JE (1965) Spin diffusion decrease in water diffusion observed in human
measurements: spin echoes in the presence of a occipital cortex during visual stimulation. Proc
time-dependent field gradient. J Chem Phys 42 Natl Acad Sci 98(16):9391–9395
(1):288–292 19. Miller KL, Bulte DP, Devlin H, Robson MD,
8. Stejskal E (1965) Use of spin echoes in a pulsed Wise RG, Woolrich MW, Jezzard P, Behrens
magnetic-field gradient to study anisotropic, TE (2007) Evidence for a vascular contribution
restricted diffusion and flow. J Chem Phys 43 to diffusion FMRI at high b value. Proc Natl
(10):3597–3603 Acad Sci 104(52):20967–20972
9. Clark CA, Le Bihan D (2000) Water diffusion 20. Schneider CA, Rasband WS, Eliceiri KW
compartmentation and anisotropy at high b (2012) NIH Image to ImageJ: 25 years of
values in the human brain. Magn Reson Med image analysis. Nat Methods 9(7):671
44(6):852–859 21. Team RC (2013) R: a language and environ-
10. Le Bihan D, Urayama S, Aso T, Hanakawa T, ment for statistical computing. R Foundation
Fukuyama H (2006) Direct and fast detection for Statistical Computing, Vienna, Austria.
of neuronal activation in the human brain with 22. Gruetter R (1993) Automatic, localized in vivo
diffusion MRI. Proc Natl Acad Sci U S A 103 adjustment of all first- and second-order shim
(21):8263–8268. https://doi.org/10.1073/ coils. Magn Reson Med 29(6):804–811
pnas.0600644103 23. Kanayama S, Kuhara S, Satoh K (1996) In vivo
11. Yablonskiy DA, Bretthorst GL, Ackerman JJ rapid magnetic field measurement and shim-
(2003) Statistical model for diffusion attenu- ming using single scan differential phase
ated MR signal. Magn Reson Med 50 mapping. Magn Reson Med 36(4):637–642
(4):664–669. https://doi.org/10.1002/ 24. Lizarbe B, Lopez-Larrubia P, Cerdan S (2015)
mrm.10578 fDWI evaluation of hypothalamic appetite reg-
12. Mori S, Barker PB (1999) Diffusion magnetic ulation pathways in mice genetically deficient in
resonance imaging: its principle and applica- leptin or neuropeptide Y. Neurochem Res 40
tions. Anat Rec 257(3):102–109 (12):2628–2638. https://doi.org/10.1007/
13. Le Bihan D (2012) Diffusion, confusion and s11064-015-1596-z
functional MRI. NeuroImage 62 25. Lizarbe B, Benitez A, Pelaez Brioso GA,
(2):1131–1136 Sanchez-Montanes M, Lopez-Larrubia P,
14. Steier R, Aradi M, Pal J, Perlaki G, Orsi G, Ballesteros P, Cerdan S (2013) Hypothalamic
Bogner P, Galyas F, Bukovics P, Janszky J, metabolic compartmentation during appetite
Doczi T, Schwarcz A (2012) A biexponential regulation as revealed by magnetic resonance
DWI study in rat brain intracellular oedema. imaging and spectroscopy methods. Front
Eur J Radiol 81(8):1758–1765. https://doi. Neuroenerg 5:6. https://doi.org/10.3389/
org/10.1016/j.ejrad.2011.03.058 fnene.2013.00006
Part III
In Vivo Spectroscopy
Chapter 10
In Vivo 1H Magnetic Resonance Spectroscopy

M. Carmen Muñoz-Hernández and Marı́a Luisa Garcı́a-Martı́n
Abstract
In vivo Magnetic Resonance Spectroscopy (MRS) allows the non-invasive detection and quantification of a
number of metabolites from localized volumes within a living organism. MRS localization techniques can
be divided into two main groups, single voxel and multi-voxel. Single voxel techniques provide the
metabolic profile from a specific small volume, whereas multi-voxel techniques are used to obtain the
spatial distribution of metabolites throughout a large volume subdivided into small contiguous voxels. This
chapter describes standard protocols for the acquisition and processing of in vivo single voxel1H MRS data
from the rodent brain.
Key words Localized spectroscopy, Single voxel, 1H MRS, Brain metabolites
1 Introduction
The history of magnetic resonance spectroscopy (MRS) dates back

to the first studies by Bloch and Purcell in 1946 [1]. At first, nuclear
magnetic resonance (NMR) was mainly used by physicists to mea-
sure the gyromagnetic ratio of nuclei. It was after the discovery of
chemical shift and spin-spin couplings [2, 3] that NMR became a
fundamental tool in chemistry research, giving rise to the beginning
of magnetic resonance spectroscopy.
During the first two decades spectra were recorded in
continuous-wave mode, which consisted of using a fixed magnetic
field and sweeping the frequency, or, more commonly, keeping the
frequency source constant while the magnetic field was varied. The
true revolution in NMR came in 1966 when Ernst and Anderson
introduced the Fourier Transform NMR [4], which is the basis of
modern magnetic resonance spectroscopy.
The first studies on spectroscopy applied to living organisms
appeared shortly after, in the early seventies [5, 6]. Since then, the
biomedical applications of MRS have grown enormously, not only
in basic research [7], but also in the clinical environment, where
MRS is currently a routine tool for diagnosis and treatment
151
152 M. Carmen Muñoz-Hernández and Marı́a Luisa Garcı́a-Martı́n
Table 1
NMR properties of isotopes commonly used in biological studies
Spin quantum Gyromagnetic ratio γ Natural Relative

Isotope number I (107 rad s1 T1) abundance (%) sensitivity (%)
1
H 1/2 26.7522208 99.9885 100
2
H 1 4.10662919 0.0115 0.97
13
C 1/2 6.728286 1.070 1.59
15
N 1/2 2.7126189 0.364 0.10
19
F 1/2 25.16233 100 83.30
31
P 1/2 10.8394 100 6.63
Information from the following sources: De Laeter, J.R. et al. (2003) Pure Appl Chem 75 (6):683–800; James,
T.L. (1975) Nuclear magnetic Resonance in biochemistry. Academic Press. https://doi.org/10.1016/B978-0-12-
380950-6.50004-9
follow-up [8, 9]. From the biological point of view, MRS can be
classified into three different modalities: in vitro (in-solution),
in vivo, and ex vivo (high resolution magic angle spinning,
HR-MAS). The aim of all of them when applied to biomedicine is
the characterization of the metabolic profile of living organisms.
Whereas in vitro and ex vivo spectroscopy are high resolution
techniques based on the same principles as the classical spectros-
copy used in chemistry labs, in vivo spectroscopy combines imaging
and spectroscopy methodology to obtain spectra from specific
volumes within a living organism.
Most in vivo spectroscopy studies are devoted to1H spectros-
copy, but any other nucleus with magnetic moment that is present
in biomolecules, such as phosphorus or carbon, could, in principle,
be observed.1H MRS has the advantage of its higher sensitivity
(Table 1), although it also has the drawback of the large water
signal, which has to be suppressed to allow for the observation of
tissue metabolites.31P and13C MRS, on the other hand, do not
need water suppression to be performed, but they have the disad-
vantage of their much lower sensitivity and, in the case of13C, the
low natural abundance and, therefore, the need for enriched
substrates.
Finally, in vivo spectroscopy techniques can be divided into two
main modalities, single- and multi-voxel, which allow for the detec-
tion of the average metabolic profile within a small single volume or
the spatial distribution of metabolites within a larger volume, sub-
divided into smaller ones.
In this chapter we will focus on the in vivo1H MRS studies of
the rodent brain at 9.4 T using single voxel techniques. Multi-voxel
and heteronuclear MRS will be discussed in other chapters.
1
H MRS 153
1.1 Single-Voxel1H There are several acquisition schemes that allow spectra to be
Spectroscopy obtained from a defined region within the tissue under study. In
Sequences the case of1H spectroscopy, there are basically two different meth-
ods, as described below:
1.1.1 Point RESolved This technique employs one 90 excitation pulse together with a
Spectroscopy (PRESS) [10] field gradient to select a slice in one spatial direction, and two 180
refocusing pulses, also combined with field gradients, in the other
two spatial directions, so that the intersection of the three selected
slices defines a cubic voxel. This technique is also known as double
spin echo (Fig. 1A). The first 180 pulse is applied at time τ1 after
the 90 pulse, giving rise to the formation of a spin echo at time 2τ1,
and the second 180 pulse refocuses this echo at time 2τ2. There-
fore, the final spin echo is formed at time 2τ1 þ 2τ2, which is the
total echo time (TE) of the sequence.
Fig. 1 Pulse sequences for in vivo single voxel1H NMR spectroscopy. (a) Point RESolved Spectroscopy
(PRESS), (b) STimulated Echo Acquisition Mode (STEAM), (c) voxel localization by the intersection of three
orthogonal slices
1.1.2 STimulated Echo The STEAM sequence uses three 90 pulses combined with three
Acquisition Mode (STEAM) orthogonal field gradients to select a cubic voxel (Fig. 1b), similarly
[11] to the PRESS sequence. To understand how the signal is generated
from this selected volume, let us focus just on the effect of the three
rf pulses, leaving the gradients aside. If the two first pulses are
separated by a time TE/2 and the last two by a time TM, four
spin echoes will be generated as follows: SE12 (formed by the
combination of pulses 1 and 2), SE13 (combination of pulses
1 and 3), SE23 (combination of pulses 2 and 3), and SE123
(formed by the combination of the three pulses). SE12 will be
formed at time TE after the first pulse, SE13 will be formed at
time TEþ2TM, SE23 at time TE/2þ2TM, and S123 at time 2TM.
Moreover, besides the spin echoes, other type of echoes, the
so-called stimulated echoes (STE), will appear at time TEþTM. If
we now take into consideration the field gradients, needed for
spatial localization, together with the three rf pulses, we can
observe that the first signal that appears after the three pulses is
precisely the STE, and therefore this is the signal used in the
STEAM sequence to acquire the spectrum. Finally, the spin-echo
signals are eliminated using crusher gradients during TM to avoid
undesired contribution of these signals to the spectrum.
The main advantage of STEAM as compared to PRESS is that
the former allows for very short echo times; however, it has the
disadvantage that the stimulated echo amplitude corresponds to
about half of that obtained with the spin echo in the PRESS
sequence. This is due to the fact that the 90 pulse only takes half
of the total magnetization to the longitudinal axis, while the other
half is de-phased by the crusher gradient. Furthermore, STEAM is
more sensitive to motion and diffusion artifacts. Consequently,
PRESS will be the method of choice unless very short echo times
are needed. Indeed, PRESS is by far the most commonly used
sequence in clinical practice as well as in preclinical research.
1.2 Brain The number of metabolites that can be detected in vivo is limited by
Metabolites Detectable the intrinsic low sensitivity of magnetic resonance and by the
In Vivo reduced spectral resolution as compared to liquid and HR-MAS
NMR. As this chapter is focused on brain spectroscopy, a list of the
main cerebral metabolites typically measured in vivo, and their
metabolic meaning, is described below:
1.2.1 N-acetyl aspartic This metabolite shows a very prominent signal at 2.01 ppm,
acid (NAA) corresponding to its acetyl group. Is considered a neuronal marker
[12] and its decrease is associated with neuronal loss, which can be
due to multiple causes, such as tumors [13–15], stroke [16], mul-
tiple sclerosis [17], etc.
1.2.2 Total Choline (tCho) A prominent peak appears at 3.20 ppm which we refer to as total
choline (tCho) and contains contributions from the trimethyl
1
H MRS 155
amine groups of three metabolites, free choline, phosphocholine

(PC), and glycerophosphocholine (GPC). In vivo they appear as a
single peak because of the low spectral resolution we mentioned
above. These three metabolites are intermediates in the synthesis
and degradation of phosphatidylcholine, the main component of
cell membranes, and therefore they are related to membrane turn-
over metabolism. Thus, tCho is considered a biomarker of cell
proliferation (tumor marker), but it is also elevated in other situa-
tions involving degradation of membranes, such as demyelinating
processes.
1.2.3 Creatine (Cr) The methyl group of creatine appears as a singlet at 3.03 ppm. This
singlet actually contains signals from two metabolites that cannot
be resolved in vivo, creatine and phosphocreatine, both involved in
energy metabolism. Creatine is synthesized in the liver and trans-
ported to the brain, therefore the diseases affecting the liver chron-
ically can lead to a decrease in brain creatine [18]. Moreover,
creatine levels can be altered in other conditions, such as brain
tumors, where decreased creatine levels have been reported as a
feature of high grade glioma [19] or as a favorable prognostic factor
in low grade gliomas [20], among others.
1.2.4 Lactate The methyl group of lactate appears at 1.31 ppm, but it can barely
be detected in the brain under normal conditions because of its low
concentration. However, lactate levels increase and can be easily
detected in various pathological conditions, such as cancer [21, 22]
or ischemia [23–25]. Sometimes lactate can be masked by over-
lapping with a large signal from mobile lipids. A variety of editing
methods have been developed to differentiate both signals. The
simplest is based on the use of an echo time corresponding to 1/J
(144 ms), where J is the homonuclear coupling constant. The scalar
coupling gives rise to a phase evolution of the methyl doublet,
which depends on the echo time. At 144 ms this phase evolution
is 180 , which results in an inverted doublet in the1H NMR
spectrum [26].
1.2.5 myo-Inositol (myo- myo-Ins is much more abundant in glial cells as compared to neu-
Ins) rons and therefore its increase is associated with an increase in the
glial component. At short echo times a large myo-Ins signal can be
observed at 3.56 ppm. At long echo time, however, since it is a
strongly coupled system, its signal disappears. The myo-Ins signal
increases significantly in low grade glial tumors, where cell dediffer-
entiation is still low, and decreases in most undifferentiated tumors
(high grade) [27].
1.2.6 Glutamate (Glu) These two molecules are essential in brain metabolism, with Glu
and Glutamine (Gln) being the predominant neurotransmitter. They are both closely
related through the glutamate-glutamine cycle [28]. The methine
proton of both Glu and Gln can be detected at short echo times
around 3.7 ppm, and the resonances from the four protons of the
two methylene groups are closely grouped between 2.1 and
2.4 ppm. At low magnetic fields there is a large overlap between
both metabolites and they are generally measured together, being
designated as Glx. At higher magnetic fields, like those used in
preclinical studies, they can sometimes be resolved, depending on
the spectral quality, but in general it is difficult to obtain reliable
measurements of both metabolites separately. Consequently, sev-
eral editing techniques have been developed in recent years to allow
both metabolites to be measured independently even at clinical
magnetic fields [29]. Glu and Gln can be altered in a wide range
of pathological conditions, such as depression [30], cancer
[31–33], hepatic encephalopathy [34, 35], multiple sclerosis
[17], etc.
2 Materials
2.1 MR Equipment 1. Horizontal 9.4 T small animal MR imaging and spectroscopy

system (Biospec 94/20) equipped with an Avance III console
(Bruker BioSpin, Ettlingen, Germany).
2. Actively shielded gradient coil BGA-12S with integrated shims,
providing gradient amplitudes up to 400 mT/m.
3. Receive-only rat brain quadrature surface coil with a shape
adapted to the rat head for optimized B1 homogeneity and
sensitivity in the area of the brain. This coil is used in combina-
tion with a transmit-only volume coil for homogeneous
excitation.
4. AutoPac-motorized positioning system, which includes: ani-
mal bed with teeth and ear bars to fix the head of the animal,
integrated waterbeds to maintain the body temperature during
anesthesia, and anesthesia mask for continuous delivery of iso-
flurane during the procedure.
5. MRI scanner workstation running ParaVision 6.0 (Bruker
BioSpin, Ettlingen, Germany).
2.2 Other Equipment 1. Vaporizer and induction chamber to anesthetize the animals
outside the MR spectrometer.
2. Small Animal Monitoring and Gating system (Small Animal
Instruments Inc. Stony Brook, NY, USA), provided with respi-
ration and temperature sensors.
3. Animal physiology monitoring software (PC-SAM 32 v8.02,
Small Animal Instrument Inc.).
1
H MRS 157
2.3 Data Analysis 1. Manufacturer software for spectroscopic data processing Top-
Spin 3.0 (Bruker, Ettlingen, Germany).
2. Third party software for quantitative analysis (as an example in
this chapter we describe the use of LCModel [36, 37], but there
are several options available).
3 Methods
3.1 Animal Handling 1. Turn on the gas flow (oxygen or air at a flow rate of 1–1.5 mL/
min) and anesthesia (4–5% isoflurane), directed to the induc-
tion chamber, and place the rat inside.
2. Once the animal is asleep, redirect the anesthesia to the animal
holder already placed on the scanner, reduce the isoflurane to
2–2.5%, and quickly move the animal to the animal cradle,
which should have a temperature control system (water or
air) to maintain temperature between 36.5 and 37.5 C during
the experimental procedure.
3. Immobilize the animal’s head using ear bars and tooth bars to
minimize motion effects during acquisition.
4. Place the breathing sensor under the lower part of the abdo-
men, so that interferences from the heartbeat are avoided, then
place the rectal temperature probe and check the animal moni-
toring unit. Anesthesia should be adjusted throughout the
experiment to maintain a breathing rate between 50 and
80 breaths per minute.
5. At the end of the experiment, remove the animal from the
animal holder and place it on a warm location until it is awake
(2–3 min). Then return the animal to its cage.
3.2 MRS 1. Position the surface RF coil as close to the skull as possible and
also properly centered respect to the brain (see Note 1).
2. Move the animal into the scanner so that the center of the
surface coil gets to the center of the volume coil, which should
have been previously position at the isocenter of the magnet.
3. Adjust the tune and match of the volume coil (see Note 2).
4. Load a positioning protocol and acquire the image with a field
of view (FOV) of 40–50 mm. Use these images to re-position
the animal inside the scanner if necessary (see Note 3).
5. Acquire axial, sagittal, and coronal T2-weighted images using
the RARE (rapid acquisition with relaxation enhancement)
sequence with the following or approximate parameters:
FOV ¼ 32 mm, matrix size ¼ 256 256, slice thick-
ness ¼ 1 mm, number of contiguous slices ¼ 10–15, repetition
Fig. 2 Voxel positioning and shim volume selection. The voxel (4 4 4 mm3) from which the1H NMR
spectrum will be acquired is outlined in yellow, and the volume for local shim adjustment
(5.5 5.5 5.5 mm3) in green
time (TR) ¼ 2.5 s, echo time (TE) ¼ 33 ms, rare factor ¼ 4 (see
Note 4).
6. Perform a local shim adjustment. In our case this will be done
using the “2_Localized_Shim” protocol and the “Map_Shim”
subroutine, as follows:
(a) Load the protocol.
(b) Define the voxel size (4 4 4 mm3 in the example
reported herein) and the position on the T2-weighted
images (Fig. 2, see Note 5).
(c) Define the shim volume. This can be done by adding a
margin to the voxel volume (Fig. 2, see Note 6).
(d) Acquire a B0 map (see Note 7).
(e) Run the localized shim scan (see Notes 8 and 9).
7. Load the localized spectroscopy protocol, PRESS or STEAM,
and copy the voxel position from the previous experiment.
8. Define acquisition parameters, which in the current example
using PRESS are the following: sweep width ¼ 10 ppm, num-
ber of points ¼ 2048, TR ¼ 3 s, TE ¼ 35 or 144 ms, number of
averages ¼ 128.
9. Adjust the bandwidth of the RF pulses to minimize the chemi-
cal shift displacement error (see Note 10). This step is funda-
mental to avoid lipid contamination from the skull (Fig. 3).
10. Optimize water suppression for by fine-tuning the power level
of the selective pulses. In Paravision 6.0 follow the next steps:
(a) In the “Water Sup” sub-card of the “Setup” card deselect
“Calc Sup Pulse Ampl.”
(b) Start the acquisition in setup mode and manually adjust
the power level of the suppression pulses (“Vp Pulse
1
H MRS 159
Fig. 3 RF bandwidth and chemical shift displacement error. (a) The RF bandwidth has been optimized to
minimize the chemical shift displacement error and thus avoid lipid signal contamination (bandwidths of 90
and 180 are 5400 and 3600 Hz, and the corresponding chemical shift displacements are 1.04 mm and
1.57 mm); (b) The default RF bandwidths have been used, which gives rise to a chemical shift displacement of
2.24 mm in x and y directions, resulting in substantial lipid signal contamination
1 Attenuation” and “Vp Pulse 2 Attenuation”) keeping

the power difference constant (see Note 11).
(c) In the “Acq/Reco” display you can observe the effect on
the water signal. Look for the minimal pulse power that
provides a good water suppression, which can be esti-
mated not only by looking at the residual water peak,
but also by looking at the digitizer filling.
(d) Stop the scan and keep the “Calc Sup Pulse Ampl” dese-
lected, otherwise the manually adjusted values would
be lost.
11. Adjust the outer volume suppression (OVS) (see Note 12). In
Paravision 6.0 proceed as follows:
(a) Default thickness and gap-to-voxel values of the suppres-
sion slices are defined in the “OVS” sub-card of the
“Preparation” card. Start from these values to do the
adjustment.
(b) In the “Optimization” card select the OVS pulses and
start the scan in setup mode.
(c) Adjust the thickness and the gap in all spatial directions
until you get a profile where the middle region,
corresponding to the voxel, is unsuppressed, whereas the
adjacent regions are suppressed (see Note 13).
(d) Stop the scan.
Fig. 4 In vivo1H NMR spectra acquired with PRESS. (a) Echo time ¼ 35 ms; (b) Echo time ¼ 144 ms. NNA: N-
acetyl aspartic acid, Glx: glutamate þ glutamine, tCr: total Creatine (phosphocreatine þ creatine), tCho: total
Choline (choline þ phosphocholine þ glycerophosphocholine), myo-Ins: myo-Inositol
12. Do eddy-current correction (ECC) if available. In Paravision

6.0 activate “Reference Scan” and “Eddy Current Compensa-
tion” in the “Optimization” card.
13. Acquire the spectra at the chosen echo times (Fig. 4). Water
suppression should be readjusted for each echo time value.
14. Water reference for quantification. Acquire an unsuppressed
water scan with the same parameters as the suppressed one,
but turning off the water suppression and reducing the number
of averages (4 in the current example) (see Note 14).
1
H MRS 161
3.3 Spectra Basic manual analysis can be performed using the manufacturer
Processing software (TopSpin in Bruker) or using third party software, such
as MestReNova (Mestrelab Research, Galicia, Spain). Although the
3.3.1 Basic Processing
step-by-step processing of NMR spectra is out of the scope of this
chapter, some basic steps with TopSpin software are described
below:
1. Transfer the spectra from Paravision to TopSpin (right click on
the acquired scan and transfer to TopSpin).
2. In the command line write “lb ¼ 3” and “si ¼ 4096.” Then
write “efp,” which executes exponential multiplication, Fourier
transform, and phase adjustment.
3. If phase is not properly adjusted, re-adjust it manually.
These steps should provide spectra similar to those in Fig. 4.
3.3.2 Metabolite For quantitative analysis, specific software able to provide auto-
Quantification matic processing with high reproducibility is advised, otherwise,
data comparison and hence robust statistical analysis could be
hampered. Among the most widely used are LCModel [37] and
jMRUI [38]. In this chapter we have chosen LCModel as an
example. The main steps of the analysis are described below:
1. Open FID data in LCModel (see Note 15).
2. Activate “Do water-scaling,” but do not activate “Do eddy-
current correction” (see Note 16).
3. Choose the basis-set corresponding to your acquisition con-
ditions (field strength, vendor, sequence, echo time. . .) (see
Note 17).
4. In advance settings, edit the control parameters and make sure
you add the modifications needed for your specific data type/
source. With Bruker data using Paravision 6.0, a correction
factor that accounts for the difference in receiver gain between
the suppressed and unsuppressed spectra must be applied (see
Note 18).
5. Click on “Run LCModel” and select the unsuppressed water
FID when asked. The quantified spectrum will pop up as
shown in Fig. 5 (see Note 19).
4 Notes
1. For maximal sensitivity and homogeneity in signal reception

the coil must be carefully positioned so that is it close to the
skull without pushing the animal, which could affect its breath-
ing capacity, and making sure that the center of the coil is
approximately at the center of the brain in the rostrocaudal
Fig. 5 LCModel analysis of an in vivo1H NMR spectrum acquired with PRESS. The black trace corresponds to
the experimental spectrum and the red trace to the fitted spectrum. The table shows, from left to right, the
calculated concentrations in mM (Conc.), the estimated standard deviations expressed in percent of the
estimated metabolite concentrations (% SD), the concentration ratios relative to total-creatine (/Cr þ PCr), and
the abbreviation of metabolites (Metabolite). Highlighted in blue are the metabolites with concentration
estimates of acceptable reliability (SD < 15%). Ala: alanine, Asp: aspartic acid, Cr: creatine, PCr: phospho-
creatine, GABA: gamma-aminobutyric acid, Glc: glucose, Gln: glutamine, Glu: glutamate, GPC: glyceropho-
sphocholine, PCho: phosphocholine, GSH: glutathione, Ins: myo-inositol, Lac: lactate, NAA: N-acetyl aspartic
acid, NAAG: N-acetylaspartylglutamic acid, Scyllo: scyllo-inositol, Tau: taurine, Gly: glycine. Sums of meta-
bolites are also provided when strong overlap occurs and therefore the uncertainty of the estimates in greatly
reduced by computing the sum instead of the individual metabolites
direction, or alternatively, centered on the region from which

the spectra will be acquired.
2. Surface coils come pre-adjusted from the vendor and no fur-
ther adjustment should be required.
3. The positioning protocol we refer to is generally called “Loca-
lizer” in Paravision 6.0 or above and “Tripilot” in previous
versions. It is based on the fast low angle shot (FLASH) gradi-
ent echo sequence and uses three orthogonal slices that inter-
sect at the isocenter producing a signal drop, due to partial
saturation, which can be used as a reference to re-position the
animal. For optimal homogeneity and therefore good spectra
quality, it is important to locate the brain region from where we
want to obtain the spectra as close as possible to the isocenter of
the magnet.
1
H MRS 163
4. The acquisition parameters of the RARE sequence are standard

parameters that can be changed according to the particularities
of the study. The purpose of these images is to serve as anato-
mical reference for voxel localization, therefore, other type of
images, such as T1-weighted images, can also be used. The
protocol described herein has been optimized to provide
good contrast and anatomical detail of the rat brain at 9.4 T.
5. For the rat brain 4 4 4 mm3 (64 μL) is a good voxel size,
however, depending on the region we want to explore this
might not be adequate. The voxel does not need to be sym-
metric, but it is important to keep in mind the final volume. We
typically use between 27 and 64 μL for rats and between 15 and
27 μL for mice, at 9.4 T.
6. In Paravision 6.0 select “Map_Shim” and “Automatic Shim
Volume” in the “Autoshim” tab and add the desired shim
volume margin (1.5 mm in the current example).
7. To acquire a B0 map go to “On Demand Protocol” and run the
“B0 Map” procedure.
8. A local shim procedure is performed that optimizes the shim
vales based on a B0 field map. An alternative option, which is
also available on some systems, is the “Fastmap”
technique [39].
9. Once the local shim adjustment has finished, a window will pop
up with the linewidth of the water peak. At the high magnetic
fields we normally use in preclinical imaging (7–9.4 T), reason-
able FWHM values are around 10–14 Hz.
10. The chemical shift displacement error (CSDE) is due to the
difference in the Larmor frequency among different metabo-
lites and the presence of magnetic field gradients employed for
the localization of the voxel. Thus, if we apply a magnetic field
gradient (G) in x direction, for instance, the frequency of a
given signal (ωx) in that direction will be given by
ωx ¼ ω0 + γGx. Consequently, for two different metabolites,
with two different ω0, there will be a spatial displacement of the
voxel for one of the metabolites respect to the other
(Δx ¼ Δω/γGx). Since the gradient strength required to select
the desired voxel size is proportional to the bandwidth of the
RF pulse, the CSDE can be reduced by increasing the RF pulse
bandwidth. However, the bandwidth that can be used will be
limited by our specific configuration, i.e., gradient system, type
of antenna, etc. Thus, when adjusting the bandwidth, it is
crucial to make sure that the RF pulses remain 90-180-180
for PRESS and 90-90-90 for STEAM.
11. Two different water suppression schemes are of common use,
CHESS (chemical shift selective) and VAPOR (variable power
and optimized relaxation delays). In both cases, the power of
the selective pulses should be fine tuned for optimal water

suppression. In the case of VAPOR the power difference
between the selective pulses should be kept constant for opti-
mal water suppression.
12. The outer volume suppression scheme implemented by Bruker
in the PRESS protocol consists of three repetitions of six
suppression slices placed parallel to the surface of the voxel. It
can be used to improve the voxel definition, being especially
useful in cases where the gradient spoilers need to be reduced
to allow the use of very short echo times.
13. If the gap to voxel is too small, a loss of signal from the voxel
may occur, and if the thickness is not large enough, contami-
nation from the surrounding tissue may get into the voxel.
14. Most of the quantification programs, such as LCModel
[36, 37], require an unsuppressed scan to be used as a refer-
ence. Depending on the MRI scanner software, it can be
acquired together with the suppressed one or independently.
Although in Paravision 6.0 there is the option of acquiring a
water reference scan (to be used for eddy current compensa-
tion), the FID is not stored and hence not available for further
analysis. In Paravision 6.0.1 and above, the water reference is
stored as fid.refscan in same directory as the water-
suppressed data.
15. In Bruker equipments the raw data are stored in “/opt/
PV6.0/data/username/”.
16. The stored FID is already corrected if you turned on the
“Eddy Current Compensation” option in Paravision (see Sub-
heading 3).
17. Specific basis-sets are provided together with the LCModel
license for different vendors and acquisition conditions. Alter-
natively, you can build your own basis-set following the
instructions in the LCModel manual [40].
18. Bruker data are not divided by the receiver gain (RG), and
therefore the metabolite concentrations must be multiplied
by the following correction factor: (RGunsuppresed/RGsup-
pressed). In Paravision 5.1 an additional correction must be
done because the number of scans (NS) is not taken into
account either. Thus, the correction factor in this case is:
(RGunsuppresed/RGsuppressed) (NSunsuppressed/NSsuppressed).
These correction factors can be included in the control para-
meters by imputing the “ATTH2O” value multiplied by that
factor. In this way, the concentrations provided by LCModel
are already corrected.
19. Although LCModel can provide absolute metabolite concen-
trations, the accurate estimate of these concentrations requires
1
H MRS 165
the use of internal or external concentration references, and

also that the effects of relaxation (T1 and T2) are taking into
account. The activation of “Do Water-Scaling” button in the
example reported in this chapter makes the unsuppressed water
spectra to be used as internal reference [41]. The use of water
as internal reference involves that both concentration and sig-
nal attenuation (due to relaxation effects, receiver gain, etc.)
are accurately known, which is usually not the case. Another
source of error comes from relaxation effects. Since the relaxa-
tion times in vivo, especially in pathological conditions, are
usually not accurately known, their effects on metabolite con-
centrations are often ignored. However, these effects can be
significant for large echo times (T2 effects) and/or short repe-
tition times (T1 effects). Thus, the use of short echo times and
reasonably large repetition times is always advisable for quanti-
tative MRS. Also, the effects of T1 can be greatly reduced by
using water-scaling. In summary, even with the use of an
internal reference, and taking into account the relaxation
times, the estimated absolute concentrations will still be off
by an unknown amount, and that is the reason why the con-
centration units are often referred to as “institutional units.”
Nonetheless, if relaxation times and acquisition conditions do
not change significantly during a study, all measurements will
deviate from true concentrations by the same factor, and hence
comparisons between different subjects or groups will still be
valid.
For further information about how to interpret the quantita-
tive results provided by LCModel, please refer to the LCMo-
del’s manual.
Acknowledgements
The MRI system used in this work has been funded by the Spanish
Ministry of Science and Innovation (National Plan for Scientific
Research, Development and Technological Innovation 2008-
2011) and the European Regional Development Fund
(PCT-420000-2010-3).
References
1. Bloch F (1946) Nuclear induction. Phys Rev 4. Ernst RR, Anderson WA (1966) Applications
70:460–474 of Fourier transform spectroscopy to magnetic
2. Gutowsky HS, McCall DW (1951) Nuclear resonance. Rev Sci Instrum 37:93–102
magnetic resonance fin structure in liquids. 5. Moon RB, Richards JH (1973) Determination
Phys Rev 82:748–749 of intracellular pH by 31P magnetic resonance.
3. Proctor WG, FC Y (1950) The dependence of J Biol Chem 248(20):7276–7278
a nuclear magnetic resonance frequency. Phys 6. Hoult DI, Busby SJ, Gadian DG, Radda GK,
Rev 77:717 Richards RE, Seeley PJ (1974) Observation of
tissue metabolites using 31P nuclear magnetic 19. Yerli H, Agildere AM, Ozen O, Geyik E,
resonance. Nature 252(5481):285–287 Atalay B, Elhan AH (2007) Evaluation of cere-
7. Garcia-Martin ML, Garcia-Espinosa MA, Cer- bral glioma grade by using normal side creatine
dan S (1998) Biochemistry detectable by MRS. as an internal reference in multi-voxel 1H-MR
In: S. Cerdán AHaFT (ed) Methodology, spec- spectroscopy. Diagn Interv Radiol 13(1):3–9
troscopy and clinical MRI, Springer, Milan, pp 20. Hattingen E, Raab P, Franz K, Lanfermann H,
7–13 Setzer M, Gerlach R, Zanella FE, Pilatus U
8. Gillies RJ, Morse DL (2005) In vivo magnetic (2008) Prognostic value of choline and crea-
resonance spectroscopy in cancer. Annu Rev tine in WHO grade II gliomas. Neuroradiology
Biomed Eng 7:287–326 50(9):759–767. https://doi.org/10.1007/
9. Castillo M, Kwock L, Mukherji SK (1996) s00234-008-0409-3
Clinical applications of proton MR spectros- 21. Herholz K, Heindel W, Luyten PR, denHol-
copy. Am J Neuroradiol 17(1):1–15 lander JA, Pietrzyk U, Voges J, Kugel H,
10. Bottomley PA (1987) Spatial localization in Friedmann G, Heiss WD (1992) In vivo imag-
NMR spectroscopy in vivo. Ann N Y Acad Sci ing of glucose consumption and lactate con-
508:333–348 centration in human gliomas. Ann Neurol 31
(3):319–327. https://doi.org/10.1002/ana.
11. Frahm J, Merboldt KD, Hanicke W (1987) 410310315
Localized proton spectroscopy using stimu-
lated echoes. J Magn Reson 72:502–508 22. Law M, Yang S, Wang H, Babb JS, Johnson G,
Cha S, Knopp EA, Zagzag D (2003) Glioma
12. Simmons ML, Frondoza CG, Coyle JT (1991) grading: sensitivity, specificity, and predictive
Immunocytochemical localization of N-acetyl- values of perfusion MR imaging and proton
aspartate with monoclonal antibodies. Neuro- MR spectroscopic imaging compared with con-
science 45(1):37–45 ventional MR imaging. Am J Neuroradiol 24
13. Negendank WG, Sauter R, Brown TR, Evel- (10):1989–1998
hoch JL, Falini A, Gotsis ED, Heerschap A, 23. Higuchi T, Fernandez EJ, Maudsley AA,
Kamada K, Lee BC, Mengeot MM, Moser E, Shimizu H, Weiner MW, Weinstein PR
Padavic-Shaller KA, Sanders JA, Spraggins TA, (1996) Mapping of lactate and N-acetyl-L-
Stillman AE, Terwey B, Vogl TJ, Wicklow K, aspartate predicts infarction during acute focal
Zimmerman RA (1996) Proton magnetic res- ischemia: in vivo 1H magnetic resonance spec-
onance spectroscopy in patients with glial troscopy in rats. Neurosurgery 38
tumors: a multicenter study. J Neurosurg 84 (1):121–129. discussion 129–130
(3):449–458. https://doi.org/10.3171/jns.
1996.84.3.0449 24. Graham GD, Blamire AM, Howseman AM,
Rothman DL, Fayad PB, Brass LM, Petroff
14. Poptani H, Gupta RK, Roy R, Pandey R, Jain OA, Shulman RG, Prichard JW (1992) Proton
VK, Chhabra DK (1995) Characterization of magnetic resonance spectroscopy of cerebral
intracranial mass lesions with in vivo proton lactate and other metabolites in stroke patients.
MR spectroscopy. Am J Neuroradiol 16 Stroke 23(3):333–340
(8):1593–1603
25. Woo CW, Lee BS, Kim ST, Kim KS (2010)
15. Brandao LA, Castillo M (2013) Adult brain Correlation between lactate and neuronal cell
tumors: clinical applications of magnetic reso- damage in the rat brain after focal ischemia: An
nance spectroscopy. Neuroimaging Clin N Am in vivo 1H magnetic resonance spectroscopic
23(3):527–555. https://doi.org/10.1016/j. (1H-MRS) study. Acta Radiol 51(3):344–350.
nic.2013.03.002 https://doi.org/10.3109/
16. Barker PB, Gillard JH, van Zijl PC, Soher BJ, 02841850903515395
Hanley DF, Agildere AM, Oppenheimer SM, 26. Kelley DA, Wald LL, Star-Lack JM (1999)
Bryan RN (1994) Acute stroke: evaluation Lactate detection at 3T: compensating J cou-
with serial proton MR spectroscopic imaging. pling effects with BASING. J Magn Reson
Radiology 192(3):723–732. https://doi.org/ Imaging 9(5):732–737
10.1148/radiology.192.3.8058940
27. Castillo M, Smith JK, Kwock L (2000) Corre-
17. Sajja BR, Wolinsky JS, Narayana PA (2009) lation of myo-inositol levels and grading of
Proton magnetic resonance spectroscopy in cerebral astrocytomas. AJNR Am J Neurora-
multiple sclerosis. Neuroimaging Clin N Am diol 21(9):1645–1649
19(1):45–58. https://doi.org/10.1016/j.nic.
2008.08.002 28. Sonnewald U, Schousboe A (2016) Introduc-
tion to the glutamate-glutamine cycle. Adv
18. Ross B, Michaelis T (1994) Clinical applica- Neurobiol 13:1–7. https://doi.org/10.1007/
tions of magnetic resonance spectroscopy. 978-3-319-45096-4_1
Magn Reson Q 10(4):191–247
1
H MRS 167
29. Ramadan S, Lin A, Stanwell P (2013) Gluta- 35. Cordoba J, Sanpedro F, Alonso J, Rovira A
mate and glutamine: a review of in vivo MRS in (2002) 1H magnetic resonance in the study
the human brain. NMR Biomed 26 of hepatic encephalopathy in humans. Metab
(12):1630–1646. https://doi.org/10.1002/ Brain Dis 17(4):415–429
nbm.3045 36. Provencher SW (1993) Estimation of metabo-
30. Yildiz-Yesiloglu A, Ankerst DP (2006) Review lite concentrations from localized in vivo pro-
of 1H magnetic resonance spectroscopy find- ton NMR spectra. Magn Reson Med 30
ings in major depressive disorder: a meta- (6):672–679
analysis. Psychiatry Res 147(1):1–25. https:// 37. Provencher SW (2001) Automatic quantitation
doi.org/10.1016/j.pscychresns.2005.12.004 of localized in vivo 1H spectra with LCModel.
31. Hazany S, Hesselink JR, Healy JF, Imbesi SG NMR Biomed 14(4):260–264
(2007) Utilization of glutamate/creatine ratios 38. Naressi A, Couturier C, Castang I, de Beer R,
for proton spectroscopic diagnosis of menin- Graveron-Demilly D (2001) Java-based graph-
giomas. Neuroradiology 49(2):121–127. ical user interface for MRUI, a software pack-
https://doi.org/10.1007/s00234-006-0167- age for quantitation of in vivo/medical
z magnetic resonance spectroscopy signals.
32. Li Y, Park I, Nelson SJ (2015) Imaging tumor Comput Biol Med 31(4):269–286
metabolism using in vivo magnetic resonance 39. Gruetter R (1993) Automatic, localized in vivo
spectroscopy. Cancer J 21(2):123–128. adjustment of all first- and second-order shim
https://doi.org/10.1097/PPO. coils. Magn Reson Med 29(6):804–811
0000000000000097 40. Garcia-Martin ML, Adrados M, Ortega MP,
33. Rijpkema M, Schuuring J, van der Meulen Y, Fernandez Gonzalez I, Lopez-Larrubia P,
van der Graaf M, Bernsen H, Boerman R, van Viano J, Garcia-Segura JM (2011) Quantita-
der Kogel A, Heerschap A (2003) Characteri- tive (1) H MR spectroscopic imaging of the
zation of oligodendrogliomas using short echo prostate gland using LCModel and a dedicated
time 1H MR spectroscopic imaging. NMR basis-set: correlation with histologic findings.
Biomed 16(1):12–18. https://doi.org/10. Magn Reson Med 65(2):329–339. https://
1002/nbm.807 doi.org/10.1002/mrm.22631
34. Chavarria L, Alonso J, Garcia-Martinez R, 41. Barker PB, Soher BJ, Blackband SJ, Chatham
Simon-Talero M, Ventura-Cots M, JC, Mathews VP, Bryan RN (1993) Quantita-
Ramirez C, Torrens M, Vargas V, Rovira A, tion of proton NMR spectra of the human
Cordoba J (2013) Brain magnetic resonance brain using tissue water as an internal concen-
spectroscopy in episodic hepatic encephalopa- tration reference. NMR Biomed 6(1):89–94
thy. J Cereb Blood Flow Metab 33
(2):272–277. https://doi.org/10.1038/
jcbfm.2012.173
Chapter 11
In Vivo Heteronuclear Magnetic Resonance Spectroscopy

Blanca Lizarbe, Antoine Cherix, and Rolf Gruetter
Abstract
Magnetic Resonance Spectroscopy is a technique that has the capability of measuring metabolites in vivo
and, in appropriate conditions, to infer its metabolic rates. The success of MRS depends a lot on its
sensitivity, which limits the usage of X-nuclei MRS. However, technological developments and refinements
in methods have made in vivo heteronuclear MRS possible in humans and in small animals. This chapter
provides detailed descriptions of the main procedures needed to perform successful in vivo heteronuclear
MRS experiments, with a particular focus on experimental setup in 13C MRS experiments in rodents.
13
Key words Magnetic Resonance Spectroscopy, X-nuclei, Surface coil, Infusion, Tracer, C labeled
glucose, Mouse
1 Introduction
Heteronuclear Magnetic Resonance Spectroscopy (MRS) is based

in the detection of nuclear magnetic resonances (NMR) from
nuclei other than protons. In preclinical research, applications rely
on the ability of MRS to noninvasively measure metabolites inside
living organs, assessing the content of nuclei like 13C, 15N, 17O,
19
F, and 31P. One of the most common applications in hetero-
nuclear MRS is in vivo 13C spectroscopy, which can be used to
determine rates of neurotransmission and neuroenergetics
[1]. Notably, 17O and 31P measurements can determine cerebral
metabolic rates of oxygen and ATP inside the mitochondria [2] and
its usage is quite extended in preclinical MRS. Nonetheless, in vivo
MRS of X-nuclei is challenging due to the low sensitivity detection
of the non-proton nuclei; they have a substantially lower gyromag-
netic ratio and typically lower natural abundance than the 1H
nucleus. Consequently, technical demands for its detection are
very high, and, currently, in vivo heteronuclear MRS in humans is
limited to a number of research groups in the world. In the past two
decades, however, sensitivity and spectral resolution of MRS data
has been enhanced by implementation of scanners with higher
169
170 Blanca Lizarbe et al.
magnetic field strength, as well as improved radio-frequency coil

designs, better radio-frequency pulse sequences, and more precise
methods for localized shimming. Therefore, X-nuclei MRS techni-
ques are becoming more and more accessible, and, especially in
preclinical investigations, further studies have been published in the
last years. Non-proton detection techniques, however, are—even in
animal MRS—still hampered by lower sensitivity detection. Thus,
implementation of a precise methodology is needed to guarantee
accurate and reproducible results. On these grounds, the aim of this
book chapter is to detail the procedures that are routinely imple-
mented in preclinical in vivo heteronuclear MRS experiments. For
the sake of completeness, protocols will be described first in a more
general way—in terms of the nuclei investigated or substrate used—
and, second, and more specifically, examples for 13C detection after
labeled glucose infusion in rodents will be given. Hence, the first
part of this chapter is dedicated to give a general description of
basics of heteronuclear NMR signal and hardware requirements;
the second part provides a concise description of the materials used
and the third part elaborates the description of the methodologies.
Finally, an additional section called Notes includes extra information
about specific methods that might, however, be of great interest for
the reader.
1.1 Basics Nuclear magnetic resonance spectroscopy is based, among other

of Heteronuclear NMR things, in detecting the resonances of the nuclei with non-zero spin
Signal (I 6¼ 0) after the application of a constant magnetic field and a
radiofrequency pulse. The most commonly studied nuclei are 1H,
31
P, and 13C, although nuclei from isotopes of many other elements
(e.g., 2H, 6Li, 10B, 11B, 14N, 15N, 17O, 19F, 23Na, 29Si, 35Cl, 113Cd,
129
Xe, 195Pt) have been used in high-field NMR spectroscopy
[3]. Overall, the frequency at which a specific nucleus resonates
depends on the external magnetic field strength applied, the gyro-
magnetic ratio of the nucleus, and its chemical environment. Par-
ticularly, the dependence on the chemical milieu has its origin in
two different phenomena: the shielding of the magnetic field by the
surrounding electrons, on one hand, and the interactions between
nuclei in the same molecule, on the other. Notably, interaction
between nuclear spins is not limited to nuclei of the same type—
the so-called homonuclear coupling—but also between different
types—the heteronuclear coupling. This interaction between differ-
ent nuclei results in a splitting of resonances into several smaller
lines at different frequencies [3], and all atoms that have I 6¼ 0 are
subjected to it accordingly to the chemical structure of the mole-
cule. The splitting of the resonances complicates the understanding
of the spectra, not only due to the incorporation of the additional
lines but also for the decrease in signal intensity of the original (and
not coupled) signal. In the study of 1H NMR signals, heteronuclear
coupling is normally neglected because, even though most protons
In Vivo Heteronuclear MRS 171
are bond to carbon molecules (at least in organic compounds), the

natural abundance of 13C is only around 1.1%, which almost elim-
inates the potential coupling effects. In non-proton detection MRS
experiments, however, heteronuclear coupling becomes critical
because the X-nuclei investigated are often bonded to the
NMR-visible H atoms, making proton decoupling mandatory—
and sometimes technically very challenging [4]—to obtain spectra
with enough signal intensities.
There are two main techniques detecting non-proton nuclei by
MRS: direct detection and indirect detection [4]. The principal
characteristics that define—and distinguish—both methods are
increased spectral dispersion of the resonances from the metabolites
for the first one, and higher sensitivity for the second one. Conse-
quently, depending on the type of study that is to be performed,
one of the methods can be chosen. For MRS in small volumes, for
example, indirect methods with higher sensitivity might be more
appropriate. If a higher separation between resonances is required,
then a direct detection approach should be chosen.
Direct detection methods measure the NMR signal of the
X-nuclei. Excitation through RF pulses is performed at the
frequency of the X-nuclei investigated, where the resonances
are also received, and decoupling is performed in the 1H chan-
nel. This has two immediate consequences: larger chemical shift
dispersion (~30 ppm for 31P NMR and >200 ppm for 13C
NMR) and substantial loss of sensitivity. Indeed, decreased sen-
sitivity is a consequence of either the lower natural abundance
13
(1.1% for C) and/or low gyromagnetic ratio
ð =γ1H ¼ 0:251; γ31P =γ1H ¼ 0:405 Þ. Nevertheless, this low sensi-
γ 13C
tivity can be overcome by the implementation of optimized

methodologies, such as the exogenous administration of labeled
precursors—mandatory for isotopes with low natural abun-
dance—, polarization transfer techniques [5], or methods taking
advantage of the Nuclear Overhauser Effect (NOE) [6]. Sensitiv-
ity enhancement by polarization transfer is achieved by transfer-
ring the nuclear spin polarization of protons to the X-nucleus
through chemical bonds, while NOE transfers the polarization
via cross-relaxation. Figure 1 shows a scheme of the original
distortionless enhancement by polarization transfer (DEPT)
sequence [5] (top), and a more developed version (bottom)
implemented by Henry P.G. et al. [7]. In this scheme, localiza-
tion is performed on the 1H magnetization and the localized
magnetization is subsequently transferred to carbon. This avoids
potential chemical shift displacement errors induced by the large
chemical shift displacement of metabolites in the 13C spectra.
Another way to improve sensitivity in direct detection methods
is to use hyperpolarization [8]. Hyperpolarized 13C MRS has
the potential to increase the sensitivity by several orders of magni-
tude by using a nonequilibrium polarization [9]. Particularly,
Fig. 1 1H localized broadband 13C NMR spectroscopy of the rat brain in vivo at 9.4 T. (a) the original DEPT
sequence; (b) modified sequence. The 13C part of DEPT is replaced by a segmented adiabatic 0 BIR-4 pulse.
Dephasing gradients (1 ms duration, 46 mT/m) are added between the RF pulses in DEPT to eliminate
unwanted, offset-dependent coherences. The flip angle of last 1H pulse is set to a nominal 45 to detect
signals simultaneously from the CH, CH2, and CH3 groups
hyperpolarized 13C MRS, due to the many technological challenges

related to the generation and preservation of nonequilibrium polar-
ization, is commonly limited to the detection of fast and dynamic
metabolic pathways.
In indirect detection methods, like 1H-[13C], the measured
signal arises from the protons that are bond to 13C. In this case,
the signal is transmitted and received through the 1H channel and
decoupling is performed at the X-nucleus frequency. The main
advantage of this approach is that the larger gyromagnetic ratio of
protons and its increased natural abundance leads to a significantly
higher sensitivity. In fact, the sensitivity detection has been recently
shown to be high enough to perform in vivo MRS experiments in
the mouse brain [10], or even in small areas of it [11]. Generally,
1
H-[13C] methods can be divided into single-shot multiple-quan-
tum coherences-based procedures (MQC) [12] and J-difference-
based methods [13]. The MQC-based methods selectively detect
the 1H NMR signals attached to 13C nuclei by destroying, though
dephasing, all other 1H coherences. However, the removal of all
these other 1H NMR signals prevents the calculation of 13C frac-
tional enrichments (FE) and the monitoring of metabolite levels
without 13C label incorporation. J-difference methods, on the
other hand, use the different evolution (refocusing) of the magne-
tization of the protons that are coupled to 13C atoms, compared to
Fig. 2 (a) Sequence diagram of SPECIAL-BISEP (TE ¼ 2.8 ms). The editing is achieved by applying an AFP in
the 13C channel on alternate scans; (b) in vitro validation of the editing scheme performed at 14.1 T on a
solution of 67% enriched [2-13C]sodium acetate (no 13C decoupling during the acquisition is applied). When
the 13C AFP is turned off, the inversion BISEP pulse acts as a 0 BIR-4 pulse (top), while when AFP is turned on,
13
C-coupled 1H resonances are inverted (middle). In the difference spectrum (bottom), the uncoupled
resonances (1H-[2-12C]) are minimized and only the 13C-coupled 1H resonances (1H-[2-13C]) are detected
the refocusing magnetization of non-coupled protons, to

detect both the total (12C þ 13C) content and the 13C fraction.
Figure 2 depicts a scheme of the SPECIAL-BISEP sequence [14]
that has been recently implemented in preclinical scanners and that
allows the detection of full signal intensity 1H-[13C] in the mouse
and rat brain.
1.2 Hardware The performance of successful heteronuclear MRS experiments

Requirements in vivo can depend on the technical requirements of the methodol-
ogy. Notably, because poor sensitivity is one of the main technical
issues, non-proton MRS is favored by the use of high field scanners,
where the signal to noise ratio (SNR) is increased with the square
root of the magnetic field B0. Naturally, at high magnetic fields, the
frequency separation between metabolites is also improved, result-
ing in a largely increased number of detectable metabolites at high
spatial specificity. However, high fields typically are associated with
increased technical challenges, such as inhomogeneous transmit
(B1) fields, an increased impact of microscopic and macroscopic
susceptibility differences on static B0 inhomogeneity, shortened T2
and lengthened T1 relaxation times and a lower effective B1 field
strength. Nonetheless, several studies demonstrate that the
improvements achieved with the high field can largely overcome
the technical challenges, extending the number of quantifiable
metabolites and increasing spatial resolution and specificity
[15–17]. Other limiting factors of heteronuclear MRS experiments
are the need of designing specific RF coils, implementation of addi-

tional RF channels, application of decoupling schemes and adiabatic
pulses. Each of these requirements will be briefly described in the
following paragraphs.
1.2.1 Design of Specific RF coils are frequency-dependent, and a coil that is to be used
RF Coils during heteronuclear NMR studies should consist of several
loops, each one corresponding to the different nuclei to be inves-
tigated. Moreover, the RF coil should be designed in a way that the
respective RF fields of the different loops are electrically isolated
[18]. A more detailed description of the characteristics of the RF
coils used in direct and indirect non-proton MRS is given in the
materials section.
1.2.2 Additional RF Heteronuclear MRS experiments are only possible in spectrometers

Channels equipped with at least an additional RF broadband channel. Fur-
thermore, acquisition has to be done in the two channels concomi-
tantly, either during the application of decoupling schemes—
proton or X-nuclei decoupling for direct or indirect detection,
respectively—or during the application of inversion pulses or dur-
ing polarization transfer sequences [4].
2 Materials
In the following paragraphs, a list including the main materials used

during a typical in vivo MRS experiment is provided.
1. Coil. Standard design for RF coil for 13C direct and indirect
detection is generally as follows: two proton transmit/receive
surface coils in quadrature lying on top of a single carbon coil
[19]. However, in order to increase proton sensitivity for 1H-
[13C] spectroscopy, an alternative has been proposed by put-
ting the proton loops on the bottom, in order to be closer to
the sample [20].
2. External reference. Typically, a glass sphere containing 13C
labeled formic acid (FA) can be at the center of the coil. This
allows measuring the 13C FA resonance in the magnet to set the
offset for the RF pulses in the carbon channel. Furthermore,
the reference allows to measure the power performance of the
coil and to adjust the power for efficient 90 and 180 pulses in
both proton and carbon channels. Briefly, to adjust the power
values on the different channel the process is as follows: first, an
experiment with a phantom is performed, and power value of a
180 hard pulse in the FA sphere is measured. Secondly, the
same pulse is performed in vivo in the FA. If values in the FA
change, values in the sample should change concomitantly.
13
3. C labeled substrates. Glucose is the most common substrate
that allows investigating both oxidative and glycolytic metabo-
lism; however, other substrates have been used to target specific
metabolic pathways. Notably, acetate can be used to measure
specifically astrocytic oxidative metabolism [21], whereas lac-
tate can be used for investigating neuronal oxidative metabo-
lism [22]. Besides the differences in cost, the different 13C-
labeled glucose substrates can be used for different purpose. It
is thus primordial to choose the appropriate substrate accord-
ing to the technique (mainly direct and indirect 13C MRS) and
the need for an increased signal relative to cost. After glycolysis,
carbon positions 1 and 6 of glucose will label the methyl group
of pyruvate, leading primarily to the labeling of glutamate C4
following the pyruvate dehydrogenase pathway and glutamate
C2 for the pyruvate carboxylase pathway [23]. [1, 6-13C]
glucose is thus twice as efficient as [1-13C] glucose labeling,
and thus detecting, TCA cycle aminoacids. The use of [1,
6-13C] glucose is particularly important for 13C direct detec-
tion, in order to avoid splitting of peaks due to coupling
between adjacent 13C carbons. Because of this coupling,
[U-13C] glucose should best used in 1H-[13C] MRS, where
decoupling in the carbon channel allows to eliminate reso-
nances from heteronuclear coupling. [2-13C] glucose can be
used to investigate the anaplerotic metabolism of glial cells as it
will label specifically glutamine positions C2 and C3 due to the
presence of pyruvate carboxylase [24].
4. Infusion pump. Infusion can be done with a pump suitable for
infusion from any syringe and operating in both infusion and
withdrawal mode. The syringe containing the labeled glucose
solution is connected to a polyethylene line that has to be long
enough to cover the distance between the catheter and the
pump, i.e., the distance between the animal in the magnet
and the operator in the monitoring room. Adding a transparent
tube connector at the end of the line can help for controlling
the adjunction of the 99% enriched 13C glucose bolus solution.
5. Physiology monitoring. Temperature of the animal can be con-
trolled with a water circulating system covering the animal and
controllable in the preparation room. Respiration and temper-
ature can be monitored with a module located near the animal
in the magnet bore and a control/gating module connected to
a computer located near the operator console module. Glucose
can be measured with a single drop of blood using any com-
mercial system. Small volume sampling is critical for small
animals with small blood volumes such as mice or for repeated
measurements.
6. Anesthesia. Animal should be anesthetized in an induction

chamber with 3–4% isoflurane delivered by an isoflurane vapor-
izer. Chamber should be connected with a gas outlet system
linked to a scavenge system for isoflurane elimination. Vapor-
izer should be connected to both air and oxygen for a better
control of animal oxygenation. The use of α-chloralose should
be only limited to non-painful procedures, which require a
relatively high level of brain awakeness, as this drug does not
produce complete anesthesia and has poor analgesic properties.
Furthermore, the only route for α-chloralose anesthesia is
through intravenous (IV) administration, which can render
the experiment more challenging than with isoflurane.
3 Methods
From a methodological perspective, there are two main types of

heteronuclear in vivo MRS studies: experiments that require
administration of exogenous tracers to increase the abundance of
the isotope investigated (like 13C MRS) or studies that do not need
it (like in 31P MRS). Naturally, studies involving infusion of labeled
molecules are methodologically more complex, because they
require the preparation, administration and maintenance of the
solution infused. More specifically, the experimental setup for a
typical 13C in vivo MRS study can be divided into four main
procedures, including: (1) preparation of the labeled solution,
(2) animal setup, (3) MRS preparation and acquisition, and
(4) administration of the labeled solution. A detailed description
of each step is given in the following paragraphs. Note that for
experiments not using infusion of tracers the rest of the procedures
are equivalent.
3.1 Substrate In vivo heteronuclear MRS experiments involve the utilization of

Solution and Infusion two labeled solutions with different labeling concentration. Initi-
Line Preparation ally, a solution at a very high enrichment, ideally 99% enriched
substrate (in labeled glucose 13C MRS this means 99% content of
13
C and 1% content of 12C in the glucose solution) is given to the
animal during a short period of time. This increases rapidly the total
content of labeled glucose in the animal. Secondly, a solution at a
slightly lower enrichment—normally around 70%—is administered
at a steady rate during the rest of the experiment. This should keep
the targeted isotopic enrichment (IE) also at steady state. Remark-
ably, high IE’s (between 50 and 100%) are convenient to allow
reliable quantifications incorporated labeled carbons into glucose
and its metabolites. To prepare two solutions at a different frac-
tional enrichment, it is convenient to obtain first the 99% enriched
solution and then dilute it to get the 70%. With this aim, the 99%
labeled substrate should be dissolved into phosphate-buffered saline
(PBS) at a proportion 20% w/v. Secondly, a part of this solution can

be used—we can fix 50 mL as an example—to obtain the 70%
enriched one. With this aim, an additional solution of non-labeled
glucose (also at 20% w/v) has to be prepared and mixed with the
50 mL of 99% enriched glucose. The specific amounts can be
calculated as described in Eqs. (1)–(3).
20 0:99 20 0:7
50 ¼X ð1Þ
100 1 100 1
9:9∗ 100
)X ¼ ¼ 70:71 ½mL ð2Þ
20∗ 0:7
Y ¼ 70:71 50 ¼ 20:71 ½mL ð3Þ
where X represents the mL of 70% enriched glucose and Y the mL

of non-labeled glucose. These calculations take advantage of the
fact that the amount of 13C glucose in the 99% enriched solution
has to be the same as in the 70% enriched preparation [Eqs. 1 and
(2]. In our example, this means that by adding 20.71 mL of the
non-labeled glucose solution to the 50 mL 99% enriched we will
obtain 70.71 mL of 70% enriched solution at 200 g/L. Note that
depending on the complexity of the study two solutions might be
infused at the same time. For example, during [3-13C] lactate
infusion studies a non-labeled solution of glucose is given concom-
itantly to avoid production of 13C-enriched glucose from [3-13C]
lactate in peripheral tissues [22].
The line containing the substrate to infuse entails two separate
parts; a catheter and a long polyethylene tube connecting it with
the syringe containing the tracer solution. The catheter consists in a
cannula that is inserted into the animals’ vein plus a short plastic
tubing extension, filled with PBS. The length and thickness of the
line depends on the conditions of the laboratory and type of exper-
iment performed. The line and the syringe should be filled first with
the 70% enriched solution, while only a small amount of the 99%
enriched substrate is used. This amount, calculated depending on
the blood glucose levels and body weight (see Subheading 3.4), is
inserted in the tip of the line and will be the first part entering the
animal when infusion starts.
3.2 Animal Setup Before putting the animal inside the scanner, the following steps
should be followed:
1. Checking general status of the animal. Before starting an exper-
iment, the health of the animal needs to be checked and his
status reported. For 13C glucose infusion experiments, the
animal’s body weight and blood glucose levels should be
measured (see Note 1). Importantly, these values will be
needed to calculate the amount and rate of the substrate infu-

sion (see Subheading 3.1).
2. Anesthesia. Animals can be anesthetized before insertion of the
cannula(s). To sedate the animals, volatile anesthetics like iso-
flurane are generally used, but other sedatives might be used,
depending on the type of study that is to be performed [25]. In
the case of isoflurane, it should be administrated at 3–4% for
induction (during approximately 3 min) and at 2% during
surgery, usually within a mixture of air (70%) and oxygen
(30%). If an anesthetic other than isoflurane is used—such as
α-chloralose or thiopental—a second catheter in a femoral vein
has to be placed for its infusion. In these cases, animals should
be intubated with an endotracheal catheter and ventilated with
a pressure-driven ventilator to ensure proper respiration.
3. Insertion of the cannula. It requires considerable expertise. To
perform the femoral vein catheterization, the inguinal region of
the anesthetized mouse has to be shaved. The femoral vessels
should be exposed after an inguinal cut parallel to abdominal
muscles. The subcutaneous branch of femoral vessels can be
eventually cauterized for fewer disturbances. With help of a
hemostat, the abdominal muscles must be hold above the
femoral vein, pulled and put under tension. This should expose
completely the femoral vein and compress it causing vasodila-
tation. Fibrous tissue covering the vein should be removed
using blunt forceps. Catheter will be inserted 2 mm in the
vein as distally as possible, while pulling the vein toward the
abdomen with help of forceps in order to stretch it sufficiently.
After letting go the forceps and losing tension on the hemostat,
the tube of the catheter (not the needle!) has to be pushed 1 cm
in the vein and the needle completely retracted. If the proce-
dure succeeded, blood from the vein should go up in the
catheter. A PBS-containing syringe is then connected to the
catheter from which injection of a few microliters in the vein
and withdrawal from the blood should indicate that the infu-
sion is working. The catheter has to be fixed on the vein and leg
using glue and the wound closed with silk suture. For more
safety, the catheter itself should be fixed on the leg with medical
tape. In experiments with rats, two catheters can be inserted:
one into a femoral vein to infuse the labeled solution and a
second into a femoral artery to extract blood and monitor
concentration of gases, glucose, lactate, and arterial blood
pressure. In studies using mice, however, because their total
blood volume is small, blood extraction during infusion pro-
tocols is not routinely performed. Alternatively to femoral vein
cannulation, which requires euthanasia of the animals at the
end of the experiment, animals can be cannulated at the level of
the lateral caudal vein (right or left). This is appropriate for
Fig. 3 Animal immobilization in holders and supporting instrumentation. Left: The mouse head is fixed through
ear bars (in white) which are secured by small screwdrivers (in red). A small piece of acrylic glass covers the
animal’s mouth to avoid loss of anesthesia and prevent potential movements. An extra bar connects the
anesthetic tubing with the animal’s mouth. On the edge of this bar a small hole (in red) is used to place the
animal’s teeth. Blue water tubing is placed on top of the animal to maintain body temperature. The
thermometer cable can be seen coming out from the mouse body. The blue tubing on the right lateral of
the holder belongs to the respiration sensor, which is placed underneath of the animal’s belly. Right: The rat
head is fixed also through ear bars. A surface coil (double 1H in quadrature and single 31P) is place centered in
the animals’ head
longitudinal studies in which several in vivo heteronuclear MRS

measurements are intended. However, tail vein cannulations
are technically more challenging and several precautions need
to be considered (see Note 2 for extended explanation). In
both cases, it is important that temperature is maintained at
37 C during surgery.
4. Transfer to the holder. Pictures in Fig. 3 illustrate the elements
needed to keep animals in optimal conditions during the in vivo
MRS studies. Ideally, animals should be immobilized with
stereotaxic fixation, to minimize potential motion during the
MR session. Figure 3 shows examples of devices used to fix the
rat (right) and mice (left) heads.
5. Monitoring of the animal’s physiology. Body temperature should
be monitored during all MR procedures, since anesthesia is
known to decrease body temperature [25]. This is normally
done using a rectal temperature probe that should be inserted
in the animals’ rectum using with some vaseline. Moreover, to

maintain body temperature in a physiological range—around
37 C—, circulating water tubes can be placed covering the
animal, as illustrated in Fig. 3 (left). It is also convenient to
monitor breathing rhythm. In this sense, concentration of
isoflurane can be adjusted in order to keep 70–100 bpm.
Finally, and to avoid corneal desiccation, a protective gel can
be applied on the animal’s eyes.
6. Positioning of the surface coil. We can carefully position the
surface coil into the animals’ head, as shown in Fig. 3 (right)
for the rat’s head (see Note 3 for more details) and transfer the
holder into the magnet.
7. Connecting Infusion line. Finally, and before placing the animal
in the center of the scanner, we have to arrange and connect a
catheter that links the cannula in the animals’ vein with the
syringe containing the labeled solution.
3.3 MRS: Preparation Once the animal is in the magnet, the MRS preparation before
and Acquisition acquisition consists in: (1) positioning the holder in such a way
that the area to be investigated is in the center of the magnet;
(2) defining the voxel of interest (VOI), which normally involves
acquisition of high resolution anatomical images, and (3) adjusting
the first and second order gradient shims in the voxel [26], a
process that leads to an optimized resolution of the 1H spectra.
Note that for defining the VOI with a good anatomical reference it
might be convenient to perform T2-weighted acquisitions. Once
the VOI localization and shimming processes are performed, the
type of the study and the particular conditions of the experiment
will determine the type of sequences employed and the calibrations
performed. A description of the potential techniques to be used and
their respective adjustments is provided in the next subsections.
3.3.1 Localization To perform single voxel MRS, approaches that limit the excitation
Techniques in specific areas are used. Usually, localization is performed on the
1
H magnetization for both direct and indirect detection methods,
but localization on the 13C is frequently possible [18]. Generally,
the specific conditions of the study will determine which localiza-
tion method is more suitable. Single voxel MRS can be accom-
plished by intersecting three different RF pulses, like in Point
Resolved Spectroscopy (PRESS) [27] and STimulated Echo Acqui-
sition (STEAM) [28] methodologies. Both techniques are based on
the spin echo (SE) sequence, with PRESS using a 90 –180 –180
scheme and STEAM employing three 90 pulses. In PRESS, the
90 pulse shifts the magnetization into the transverse plane, the
180 refocuses it only from one column and the last pulse refocuses
magnetization in the third dimension. Since the second pulse in
STEAM flips only half of the magnetization along z, only half of the
possible signal intensity is acquired. Nevertheless, the larger band-

width of the 90 in STEAM for the same RF power implies less
errors due to chemical shift displacement has and can be used with
shorter echo times (TE) [29]. When short TE is used, homonuclear
J-coupling evolution among protons is minimized and additional
metabolites, such as glutamine, glutamate, γ-aminobutyrate, aspar-
tate, and myo-inositol can be detected with higher signal ampli-
tudes. Image Selected In vivo Spectroscopy (ISIS) [30] defines the
VOI using an add-subtract scheme. Briefly, the localization is per-
formed through multiple scans: in the first round of scans magneti-
zation is flipped into the transverse plane, and in the second
magnetization is inverted only in the VOI, and then flipped to
the opposite side of the transverse plane. The resulting signal is
subtracted, remaining only signal coming from the VOI. Since
localization is achieved through multiple scans, ISIS is prone to
be affected by movement, and is often employed together with
other localization techniques such as Outer Volume Suppression
(OVS). OVS approach defines the voxel by first exciting magneti-
zation outside the VOI and then dephasing it, while the magneti-
zation inside the VOI is not affected. Notably, SPin ECho full
Intensity Acquired Localized spectroscopy (SPECIAL) [31] is a
hybrid sequence that uses 1D ISIS localization with the first scan
(adiabatic 180 pulse) combined with a SE sequence (90 –180 ), as
shown in Fig. 2a. It has some of the advantages of STEAM (used
with short TE) but it keeps the full signal. However, the
add-subtract scheme of the 1D ISIS can cause artifact movements
and it is often used with OVS. Note that for all localization
approaches the power needed to apply the 90 or 180 pulses
needs to be calibrated for each experimental condition.
3.3.2 Suppression Suppression of the water signal can be achieved using CHEmical
of the Water Signal Shift Selective pulses (CHESS) [32]. Values have to be calibrated
during each experimental session to obtain a minimum contribu-
tion of the water signal to the spectra. This can be achieved by using
a sequence with variable-power RF pulses with optimized relaxa-
tion delays, like (VAPOR) [33], and selecting the power that mini-
mizes the longitudinal magnetization of water.
3.3.3 Calibrations in 13C In general, due to the low sensitivity of the 13C signal, calibrations
Channel cannot be done in vivo and should be performed previously in vitro
in a phantom containing a known concentration of 13C labeled
molecules. Interestingly, using external bodies during calibration
and during the in vivo studies, such as formic acid spheres, is very
convenient (see Subheading 2).
3.3.4 Decoupling Heteronuclear coupling between the X-nucleus and the 1H is rela-
Schemes tively large, and decoupling schemes [18] need to be implemented
to avoid splitting of the signal in several lines. 1H (or 13C) decou-
pling is usually done by applying a RF field (B2) at the Larmor
frequency of 1H (or 13C) [13], which can result in increased RF
power deposition and cause excessive heating in the tissue. In this
sense, it is useful to optimize decoupling power to the minimum.
The value should correspond to the minimum power that elimi-
nates most if not all effects of heteronuclear coupling signal.
3.3.5 13
C Editing Block For 1H-[13C] methods, Adiabatic Full Passage (AFP) pulses in the
13
C channel should be also calibrated previously in a phantom
containing a known concentration of 13C labeled molecules. Opti-
mized power should invert the total 1H-[13C] signal. Figure 2b
shows an example of a non-inverted, not-decoupled spectra of
labeled acetate (top), and inverted not-decoupled spectra of labeled
acetate (middle) after the application of the AFP pulse in the 13C
channel. The use of an external reference allows detecting potential
variations on the inversion performance of AFP pulses (see Sub-
heading 2).
3.4 Tracer Infusion Infusing a tracer inevitably increases the content of the substance in
particular. Because of the non-toxicity of the tracers used and the
intrinsic low sensitivity of heteronuclear MRS, the label is normally
administered in high quantities, with enrichments ranging between
50 and 100% [34]. Typically, the targeted final IE at steady state is
around 70%, and the infusion process is divided in two parts:
(1) administration of a bolus of labeled substrate (99% enriched)
during a short period of time at an exponential rate and (2) infusion
at a constant rate of 70% enriched solution that keeps the IE at
steady state during the rest of the experiment. Several factors such
as the volume of the glucose bolus infused, rates and time constant
of its administration need to be calculated beforehand. The next
sections explain in detail these calculations.
3.4.1 Glucose Plasma The amount of 13C glucose in blood at the steady state of IE
Levels (Glcfinal) (Glcfinal) should be equal to the sum of 13C glucose in blood at
initial basal levels (Glcbasal), enriched with natural abundance of
1.1%, and the administration of the 99% enriched bolus. This
relationship is expressed in Eqs. (4) and (5).
1:1 99
Glcfinal ∗ IR ¼ Glcbasal ∗ þ ðGlcfinal Glcbasal Þ∗ ð4Þ
100 100
With IE ¼ 70%
1:1 99
Glcfinal ¼ Glcbasal ∗ ð5Þ
70 99
1:1 99 1 1
if ffi ) Glcfinal ¼ Glcbasal ð6Þ
70 99 ð1 70Þ ð1 IR Þ
With IE ¼ 70%
Glcfinal ¼ 3:33∗ Glcbasal ð7Þ
This means that considering Eq. 7 final glucose values can be

calculated depending on the targeted IE. If physiological values are
to be maintained initial values of blood glucose should be low.
3.4.2 Glucose Bolus Taking into account that the percentage of extracellular fluid in
Volume rodents is around 20% of total body weight [35], and using Eq. 7 to
calculate the difference between basal and final glucose concentra-
tion, we can estimate that the total glucose bolus administrated in
g/kg of animal is 0.0047 * Glcbasal, as expressed in Eq. 8.
Total glc bolus½g=kg ¼ 0:2∗ ðGlcfinal Glcbasal Þ
¼ 0:0047∗ Glcbasal ð8Þ
As an example, and to proceed with calculations, we will con-

sider a basal glucose level of 90 mg/dL—typical value mice after
overnight fasting in mice. We obtain a total bolus of 0.423 g/kg.
To transform expression 8 into volume values, we will consider an
estimated glucose disposal rate (eGDR) of 33.2 mg/kg/min [36],
duration of the bolus of 5 min, and a glucose solution of 200 g/L.
The total volume can be calculated as shown in Eq. 9 which
corresponds to 2.94 mL/kg of animal in our example.
∗
eGDR ∗ 5 g 1000 mL
Total glc bolus½mL=kg ¼ Total glc bolus½g=kg þ ð9Þ
1000 kg 200 g
3.4.3 Rate The glucose bolus is habitually administered during 5 min follow-
of Administration ing an exponential decay. To calculate the time constant of the
of the Bolus decay (k), we can use the system of equations expressed in
Eqs. 10 and 11.
V ðt ¼ 0Þ ¼ V 0 ¼ 0:45 Total glc bolus ð10Þ
V ðt ¼ 5Þ ¼ V 0 ek ¼ inf :rate

5
ð11Þ
In Eq. 10 we express the assumption that the 45% of the total

bolus is infused in the first minute, while Eq. 11 states the fact that
in the last minute of the bolus the amount administered equals the
amount calculated with the constant infusion rate of glucose
administration (inf. rate) (see Subheading 3.4.4 for calculations),
both in 1 min.
3.4.4 Infusion Rate After the administration of the glucose bolus, 70% enriched [1,
6-13C] glucose is infused at a rate equivalent to the eGDR,
33.2 mg/kg/min. If we are using a glucose solution of 200 g/L,
this infusion rate can be also expressed as 9.96 mL/kg/h. Ideally,
to keep the IE at a steady state, this value might need to be adjusted
based depending on concomitantly measured plasma glucose con-
centrations; if plasma glucose levels decrease infusion rate should be
increased by the same percentage, and if plasma glucose levels
increase the speed of administration of the glucose solution should
be decreased. In rats, these concomitant measurements can be
easily done by analyzing the sample extracts of the femoral artery,
which are normally performed every 15–30 min. In mice, however,
blood extractions are not performed routinely and alternative ways
of warranting IE stability are needed. In this respect, there are
basically two ways of checking this constancy of IE. The first one
consists in taking out the holder from center of the magnet and
removing a small drop of blood from the tip of the tail, with the
counterpart of having to reposition the animal in the exact same
place afterwards. Finally, values can also be checked using the brain
glucose and/or lactate FE calculations from MRS measurements.
4 Notes
1. Glucose measurements. In experiments with labeled glucose

infusions it is critical to measure the blood glucose levels in
several time points. Indeed, glucose values are used to calculate
the amount of 99% enriched substrate given. Interestingly, it is
important to know that the use of some anesthetics, like iso-
flurane, tend to increase glucose concentration in blood, which
is especially true after using high dosages like during induction.
Thus, before starting infusion it is useful—particularly in
experiments with mice, where blood extractions cannot be
performed once the animal is inside the scanner—to measure
blood glucose at the following time points: before induction
(1), after induction (2) and (3) in the holder before entering
the magnet. Ideally, values in (3) should be similar than values
in (1), indicating that a sort of stability is reached. If it is the
case, and the time between putting the mouse inside the scan-
ner and starting infusion is not too long, the value in (3) can be
used to calculate the 99% bolus volume. Otherwise it is recom-
mended to perform a fourth measurement right before starting
infusion, removing the mouse form the scanner, with the
counterpart of having to position two times the mouse in the
magnet. In experiment with bigger animals this can be solved
by having a line to extract blood from the femoral artery.
2. Tail vein vs. femoral vein. Tail vein cannulations are more fragile
than cannulations in femoral vein, but they are mandatory to
perform longitudinal studies with measurements at different
time points. The protocol for cauterization is similar than for
femoral vein, but its smaller size and localization makes it easier
for the needle to accidentally come out. This means that animal
manipulation has to be done with special care. Also, the smaller
size of the vein makes tail vein cannulation more prone to be
closed and blocked (partially or completely) the entrance of
infusion. To avoid this, two precautions should be taken: keep
the tail at warm temperatures while animal preparation and
check its circulation by infusing of small amounts of PBS during
animal preparation. In fact, once the animal is in the scanner it is
always convenient to start infusion only with PBS to keep the
line opened. Finally, it is important to take special care of the tail
at the end of each MRS session. Sometimes small injuries caused
by cannulations can end up with damages in the tail that prevent
its usage for the following time points.
3. Careful positioning of the surface coil. Being poor sensitivity one
of the main issues in heteronuclear MRS, the position of the
coil should be optimized to increase SNR in the area of interest.
With this aim, it is convenient to perform pilot studies to
determine which position enhances mostly the signal in this
area. Besides, it is recommended to implement periodically
additional tests in phantoms that make sure performance of
the coil is not affected.
References
1. Rothman DL, De Feyter HM, de Graaf RA, 5. Doddrell DM, Pegg DT, Bendall MR (1982)
Mason GF, Behar KL (2011) 13C MRS studies Distortionless enhancement of NMR signals by
of neuroenergetics and neurotransmitter polarization transfer. J Magn Reson 48
cycling in humans. NMR Biomed 24 (2):323–327. https://doi.org/10.1016/
(8):943–957. https://doi.org/10.1002/nbm. 0022-2364(82)90286-4
1772 6. Lagemaat MW, van de Bank BL, Sati P, Li SZ,
2. Zhu XH, Du F, Zhang N, Zhang Y, Lei H, Maas MC, Scheenen TWJ (2016) Repeatability
Zhang X, Qiao H, Ugurbil K, Chen W (2009) of P-31 MRSI in the human brain at 7T with
Advanced in vivo heteronuclear MRS and without the nuclear Overhauser effect.
approaches for studying brain bioenergetics NMR Biomed 29(3):256–263. https://doi.
driven by mitochondria. Methods Mol Biol org/10.1002/nbm.3455
489:317–357. https://doi.org/10.1007/ 7. Henry PG, Tkac I, Gruetter R (2003)
978-1-59745-543-5_15 1H-localized broadband 13C NMR spectros-
3. de Graaf RA (2007) Basic principles. In: In vivo copy of the rat brain in vivo at 9.4 T. Magn
NMR spectroscopy. Wiley, pp 1–42. doi: Reson Med 50(4):684–692. https://doi.org/
https://doi.org/10.1002/9780470512968. 10.1002/mrm.10601
ch1 8. Schroeder MA, Atherton HJ, Ball DR, Cole
4. de Graaf RA, Rothman DL, Behar KL (2011) MA, Heather LC, Griffin JL, Clarke K, Radda
State of the art direct 13C and indirect GK, Tyler DJ (2009) Real-time assessment of
1H-[13C] NMR spectroscopy in vivo. A prac- Krebs cycle metabolism using hyperpolarized
tical guide. NMR Biomed 24(8):958–972. 13C magnetic resonance spectroscopy. FASEB
https://doi.org/10.1002/nbm.1761 J 23(8):2529–2538. https://doi.org/10.
1096/fj.09-129171
9. Ardenkjaer-Larsen JH, Fridlund B, Gram A, 19. Adriany G, Gruetter R (1997) A half-volume

Hansson G, Hansson L, Lerche MH, coil for efficient proton decoupling in humans
Servin R, Thaning M, Golman K (2003) at 4 tesla. J Magn Reson 125(1):178–184.
Increase in signal-to-noise ratio of >10,000 https://doi.org/10.1006/jmre.1997.1113
times in liquid-state NMR. Proc Natl Acad Sci 20. Lizarbe B, Cherix A, Lei H, Gruetter R (2015)
U S A 100(18):10158–10163. https://doi. In vivo 13C spectroscopy of the mouse hypo-
org/10.1073/pnas.1733835100 thalamus. In: ESMRMB, Edinburgh
10. Xin LJ, Lanz B, Lei HX, Gruetter R (2015) 21. Waniewski RA, Martin DL (1998) Preferential
Assessment of metabolic fluxes in the mouse utilization of acetate by astrocytes is attribut-
brain in vivo using H-1-[C-13] NMR spectros- able to transport. J Neurosci 18
copy at 14.1 Tesla. J Cerebr Blood F Met 35 (14):5225–5233
(5):759–765. https://doi.org/10.1038/ 22. Duarte JM, Girault FM, Gruetter R (2015)
jcbfm.2014.251 Brain energy metabolism measured by (13)C
11. Lizarbe B, Cherix A, Xin L, Lei H, Gruetter R magnetic resonance spectroscopy in vivo upon
(2016) In vivo detection of hypothalamic glu- infusion of [3-(13)C]lactate. J Neurosci Res 93
cose metabolism in HFD and regular fed mice. (7):1009–1018. https://doi.org/10.1002/
Proceedings of the 24th Annual Meeting jnr.23531
ISMRM, Singapore, p.110. 23. Henry PG, Adriany G, Deelchand D,
12. Vuister GW, Ruizcabello J, Vanzijl PCM Gruetter R, Marjanska M, Oz G, Seaquist ER,
(1992) Gradient-enhanced multiple-quantum Shestov A, Ugurbil K (2006) In vivo 13C
filter (ge-MQF)—a simple way to obtain NMR spectroscopy and metabolic modeling
single-scan phase-sensitive HMQC spectra. J in the brain: a practical perspective. Magn
Magn Reson 100(1):215–220. https://doi. Reson Imaging 24(4):527–539. https://doi.
org/10.1016/0022-2364(92)90381-G org/10.1016/j.mri.2006.01.003
13. Pfeuffer J, Tkac I, Choi IY, Merkle H, 24. Sibson NR, Mason GF, Shen J, Cline GW,
Ugurbil K, Garwood M, Gruetter R (1999) Herskovits AZ, Wall JE, Behar KL, Rothman
Localized in vivo 1H NMR detection of neu- DL, Shulman RG (2001) In vivo (13)C NMR
rotransmitter labeling in rat brain during infu- measurement of neurotransmitter glutamate
sion of [1-13C] D-glucose. Magn Reson Med cycling, anaplerosis and TCA cycle flux in rat
41(6):1077–1083 brain during. J Neurochem 76(4):975–989
14. Xin LJ, Mlynarik V, Lanz B, Frenke H, Gruet- 25. Gargiulo S, Greco A, Gramanzini M,
ter R (2010) (1)H-[(13)C] NMR spectroscopy Esposito S, Affuso A, Brunetti A, Vesce G
of the rat brain during infusion of [2-(13)C] (2012) Mice anesthesia, analgesia, and care,
acetate at 14.1 T. Magn Reson Med 64 Part I: anesthetic considerations in preclinical
(2):334–340. https://doi.org/10.1002/ research. ILAR J 53(1):E55–E69. https://doi.
mrm.22359 org/10.1093/ilar.53.1.55
15. Duarte JM, Lei H, Mlynarik V, Gruetter R 26. Tkac I, Henry PG, Andersen P, Keene CD,
(2012) The neurochemical profile quantified Low WC, Gruetter R (2004) Highly resolved
by in vivo 1H NMR spectroscopy. NeuroImage in vivo 1H NMR spectroscopy of the mouse
61(2):342–362. https://doi.org/10.1016/j. brain at 9.4 T. Magn Reson Med 52
neuroimage.2011.12.038 (3):478–484. https://doi.org/10.1002/
16. Duarte JM, Gruetter R (2013) Glutamatergic mrm.20184
and GABAergic energy metabolism measured 27. Bottomley PA (1987) Spatial localization in
in the rat brain by (13) C NMR spectroscopy at NMR spectroscopy in vivo. Ann N Y Acad Sci
14.1 T. J Neurochem 126(5):579–590. 508:333–348
https://doi.org/10.1111/jnc.12333 28. Frahm J, Bruhn H, Gyngell ML, Merboldt
17. Cudalbu C, Lanz B, Duarte JM, Morgenthaler KD, Hanicke W, Sauter R (1989) Localized
FD, Pilloud Y, Mlynarik V, Gruetter R (2012) high-resolution proton NMR spectroscopy
Cerebral glutamine metabolism under hyper- using stimulated echoes: initial applications to
ammonemia determined in vivo by localized human brain in vivo. Magn Reson Med 9
(1)H and (15)N NMR spectroscopy. J Cerebr (1):79–93
Blood F Met 32(4):696–708. https://doi. 29. Tkac I, Starcuk Z, Choi IY, Gruetter R (1999)
org/10.1038/jcbfm.2011.173 In vivo 1H NMR spectroscopy of rat brain at
18. Gruetter R, Adriany G, Choi IY, Henry PG, 1 ms echo time. Magn Reson Med 41
Lei H, Oz G (2003) Localized in vivo 13C (4):649–656
NMR spectroscopy of the brain. NMR Biomed 30. Ordidge RJ, Connelly A, Lohman JAB (1986)
16(6-7):313–338. https://doi.org/10.1002/ Image-selected invivo spectroscopy (Isis)—a
nbm.841
new technique for spatially selective 34. Gruetter R (2002) In vivo 13C NMR studies
NMR-spectroscopy. J Magn Reson 66 of compartmentalized cerebral carbohydrate
(2):283–294. https://doi.org/10.1016/ metabolism. Neurochem Int 41(2-3):143–154
0022-2364(86)90031-4 35. Chapman ME, Hu L, Plato CF, Kohan DE
31. Mlynarik V, Gambarota G, Frenkel H, Gruetter (2010) Bioimpedance spectroscopy for the
R (2006) Localized short-echo-time proton estimation of body fluid volumes in mice. Am
MR spectroscopy with full signal-intensity J Physiol Ren Physiol 299(1):F280–F283.
acquisition. Magn Reson Med 56 https://doi.org/10.1152/ajprenal.00113.
(5):965–970. https://doi.org/10.1002/ 2010
mrm.21043 36. Jucker BM, Schaeffer TR, Haimbach RE,
32. Haase A, Frahm J, Hanicke W, Matthaei D McIntosh TS, Chun D, Mayer M, Ohlstein
(1985) 1H NMR chemical shift selective DH, Davis HM, Smith SA, Cobitz AR, Sarkar
(CHESS) imaging. Phys Med Biol 30 SK (2002) Normalization of skeletal muscle
(4):341–344 glycogen synthesis and glycolysis in
33. Tkac I, Gruetter R (2005) Methodology of H rosiglitazone-treated Zucker fatty rats: an
NMR spectroscopy of the human brain at very in vivo nuclear magnetic resonance study. Dia-
high magnetic fields. Appl Magn Reson 29 betes 51(7):2066–2073
BF03166960
Chapter 12
1
H Spectroscopic Imaging of the Rodent Brain
Rui V. Simões, Emma Muñoz-Moreno, Raúl Tudela, and Guadalupe Soria
Abstract
Proton MR spectroscopic imaging (MRSI) can provide a variety of “molecular images” from animal models
of human disease, which are useful for different research purposes. This chapter describes a protocol for
in vivo acquisition and analysis of MRSI data from the rodent brain.
Key words Magnetic Resonance Spectroscopic Imaging (MRSI), Chemical shift imaging (CSI), Field
of view (FOV), Volume of interest (VOI), Point resolved spectroscopy (PRESS), Acquisition-weight-
ing, Spectral analysis
1 Introduction
Proton MR spectroscopic imaging (MRSI), or chemical shift imag-

ing (CSI), is an imaging modality available in most preclinical MR
spectrometers. By combining the advantages of MRI and MRS,
MRSI provides spatially resolved metabolic profiles within specific
organs or tissues, such as the brain parenchyma. This information is
useful to generate simultaneously different types of “molecular
images” in vivo.
The preclinical applications of MRSI in the rodent brain are
quite diverse, including regional maps of metabolites [1], tempera-
ture [2], pH [3], tissue delimitation [4], response to therapy [5],
pattern perturbation [6], and response to stimulation [7]. These
techniques have been used to study several animal models of dis-
ease, such as Parkinson’s [8], ischemia [9], and cancer [10]. More
recently, we have also used MRSI in the neonatal rabbit brain
(Fig. 1).
Compared to MRI, the performance of proton MRSI in vivo is
hampered by low signal-to-noise ratios (SNR), due to the low
concentration of metabolites in the tissues, and low spatial resolu-
tion—a compromise to acquire data in a timely manner. Impor-
tantly, as the nominal voxel size decreases, spectra become noisier
(lower SNR) but better resolved [11]. Volume localization
189
190 Rui V. Simões et al.
Fig. 1 MRSI of the neonatal rabbit brain. Data acquired at two days post-delivery in the transversal plane, with
Point Resolved Spectroscopy (PRESS) localization and acquisition-weighted k-space sampling (Hanning filter),
as detailed previously [17]. Top left, reference T2-w image (RARE), acquired over a 24 24 mm field of view
(FOV) and 1 mm slice thickness, showing the volume of interest for MRSI (purple box, 9 9 mm—enlarged on
the right, showing the final spectral matrix within) and one specific voxel position (black box, showed enlarged
at the bottom). Additional parameters for MRSI: matrix (initial size, MTXi ¼ 8 8, with 9 μL nominal voxel
size; Fourier interpolated to final size, MTXf ¼ 32 32, with 0.6 μL voxel display size); repetition time (TR),
2500 ms; echo time (TE), 14 ms; 512 accumulation (12 in the center of k-space); water-suppression (VAPOR
module, 300 MHz); 4006.4 Hz sweep-width and 2k points; 21.5 min total acquisition time. Major peak
assignments: alanine (Ala), creatine and phosphocreatine (CrT pool), glutamine (Gln) and glutamate (Glu; Glx
pool), glycerophosphocholine and phosphocholine (ChoT pool), glycine (Gly); glutathione (GSH), lactate (Lac),
myo-inositol (Ino), N-acetylaspartate (NAA) and N-acetylaspartylglutamate (NAAT pool), phosphoetanolamine
(PE); taurine (Tau), and macromolecules (MM). Experiment performed on a Bruker BioSpec 7T running with
Paravision 5.1 and a single-channel quadrature RF coil for the rat brain. Matrix spectra processed and
displayed with Paravision 5.1 softwate (CSIvisualizer, CSIdash, and TopSpin). Enlarged single spectrum
displayed with jMRUI 5.2 software, after zero order phase adjustment and line-broadening with a 4 Hz
apodization filter
techniques, such as Point Resolved Spectroscopy (PRESS), and

non-Cartesian modalities of k-space sampling, e.g., acquisition-
weighting, have greatly improved the performance of MRSI. Spe-
cifically, accumulating more in the center of k-space than at its
edges improves the shape of the Spatial Response Function
(SRF), which in turn reduces signal contamination without
compromising sensitivity or spatial resolution [12].
Another important aspect of MRSI is post-processing. A large
amount of data is generated in each experiment, which can be
Spectroscopic Imaging 191
challenging to handle. After visual inspection of the data, with

available software tools such as jMRUI [13], the classical approach
for spectral analysis is quantification—either as metabolite ratios
(most frequent, using, e.g., total creatine as an internal reference)
or absolute, in which case an additional reference water signal is
typically used. The most popular software tools available are based
on linear fitting with metabolite basis sets: LC Model [14] and
QUEST module in jMRUI. An alternative to spectral quantification
is classification. Thus, methods based on pattern recognition analy-
sis are available, such as SpectraClassifier [15], which rely on the full
spectral pattern as a “metabolic fingerprint” to discriminate
between different groups [16], rather than estimations of individ-
ual metabolite contributions to that pattern.
A previous protocol is available, which describes the require-
ments for mouse brain MRSI on a Bruker 7 T spectrometer,
equipped with a single-channel surface coil and running with
ParaVision 4.0; and post-processing based on pattern recognition
analysis [17]. Since then, multi-channel coils became widely avail-
able for larger animals and the acquisition procedures changed
significantly. Here, we describe an MRSI protocol for the rodent
brain in vivo on a Bruker 7 T spectrometer, using a 4-channel
surface coil and ParaVision 6.0.1 for data acquisition, and jMRUI
and LC Model for spectral analysis.
2 Materials
This section describes the requirements to perform good quality

MRSI in the rodent brain in vivo. The specific configuration used at
our site is also detailed (information in parenthesis).
2.1 Hardware 1. High-field magnet, ideally horizontal and 7 Tesla or above

Requirements (Bruker BioSpec 70/30 equipped with an Avance3 console—
Bruker BioSpin, Ettlingen, Germany).
2. Robust gradient system and up to second order shims
(B-GA12 gradient coil inserted into a B-GA20S gradient sys-
tem, providing gradient amplitudes of 400 mT/m, with
integrated shim setup).
3. Proton radiofrequency coil for the rat brain, with good sensi-
tivity (4-channel receiving quadrature surface coil, actively
decoupled from the transmitting volume resonator).
2.2 Software 1. Animal physiology monitoring (PC-SAM 32 v8.02, Small Ani-

Requirements mal Instrument Inc.).
2. Manufacturer software for MR data acquisition and processing
(ParaVision 6.0.1).
3. Third party software for post-processing and analysis of MRSI

data (several options, including jMRUI [13], LC Model [14],
and SpectraClassifier [15]).
2.3 Animals 1. Rodent (we have tested this protocol with rats between
250 and 400 g; but is also valid for mice as long as the FOV
and VOI sizes are suitably adjusted).
2. Vaporizer and induction chamber, to anesthetize the animals
outside the MR spectrometer (isoflurane in O2/N2O gas
mixture).
3. Cradle to position the animal inside the MR spectrometer,
which includes: restraining system (stereotactic fixing with
three fixation points, for the ear cavities and teeth); mainte-
nance of anesthesia (isoflurane gas, delivered from vaporizer);
heating system (recirculating water blanket); sensors for respi-
ration and temperature monitoring (chest/abdominal pillow-
probe and rectal probe, respectively).
4. Monitoring unit to control animal physiology (control/gating
module, Small Animal Instruments Inc. Stony Brook, NY,
USA).
3 Methods
3.1 Animal Handling 1. Turn on the gas flow (30% O2, 70% N2O, 800 mL/min) and
anesthesia (3–4% isoflurane) in the induction chamber and put
the rat inside.
2. Once the animal is asleep (1–2 min), quickly move it to the MR
cradle, where anesthetic gas is already being delivered (30% O2,
70% N2O, 800 mL/min, 1–2.5% isoflurane) and the heating
system running (recirculating heated water system); position
the animal head, using the head fixation system (see Note 1).
3. Position the sensors for breathing and rectal temperature and
check the animal monitoring unit. The anesthesia and heating
system should be adjusted to control the breathing and rectal
temperature throughout the experiment, ideally between
60–80 BPM and 36.5–37.5 C, respectively.
4. At the end of the experiment, the animal should be removed
from the cradle and left to recover in a warm environment until
full recovery (2–3 min), and then returned to its cage.
3.2 Initial Reference 1. Once the animal is ready, put the 4-channel RF surface coil over
Images the head and position the cradle inside the magnet bore.
2. Syntonize the probe (Wobble) and acquire the scout localizers
in the three orthogonal planes, adjusting the animal position if
Fig. 2 Preparation for global shim adjustment. Automatic Map Shim adjustment is performed within an
ellipsoid adjusted to fit the brain volume, based on T2-w reference images: left, coronal view; center, sagittal
view; right, transversal view
necessary to have the brain in the isocenter of the magnet; then

acquire a localizer-multislice.
3. Acquire a B0 map of the whole brain: in the adjustments
platform, select “B0 map.”
4. Load standard T2-w images (standard turbo-RARE: e.g.,
TR/TE ¼ 6000/35 ms; 1 average, RARE factor 8) of the
whole brain, in each orthogonal plane (axial, coronal and sag-
ittal); these images will be used to define the geometry and
position parameters for MRSI, in Subsection 3.3.
5. Before acquiring the T2-w images, start by automatically
adjusting the global B0 field homogeneity. This can be per-
formed within an ellipsoid adjusted to fit the brain volume,
based on the localizer-multislice (Fig. 2): in the Setup tab of the
acquisition parameters, go to Auto Shim, select “Map Shim”
and “Shape Shim Volume ¼ Ellipsoid”; adjust the ellipsoid to
the brain volume, avoiding contact with the skull.
3.3 Preparation 1. Load an MRSI experiment (CSI) with volume localization

for MRSI (PRESS volume of interest, VOI), and use the T2-w images
acquired in each plane (Subheading 3.2) to define: FOV size
(typically 30 1.5 30 1.5 mm for the rat brain) and
position (e.g., transversal slice); slice thickness (usually
1.5 0.5 mm); and VOI geometry and position (usually
within 10 1.5 10 1.5 for the rat brain).
2. Turn on the outer volume suppression module (OVS), with a
0.5–1.0 mm gap to the VOI in each direction, and, e.g., 10 mm
slice thickness.
Fig. 3 Reference T2-w image for MRSI. Left side, high-resolution reference image (HRref), with matrix size
256 256. Right side, a low-resolution image, with the same matrix size as MTXf. The inner green square
represents the VOI, and the outer square indicates the region for the local homogeneity adjustments
3. Acquire two additional single-slice T2-w reference images

(RARE), with the same FOV geometry and position as those
defined for the MRSI experiment (imported parameters): a
high-resolution image (HRref: 256 256 matrix size) and a
low-resolution image (LRref: same matrix size as defined for the
output of the MRSI experiment, MTXf).
4. Use the T2-w reference images to confirm: HRref, the VOI
position within the FOV (ideally in the magnet isocenter);
LRref, the proper alignment of the VOI edges with the final
MRSI spectral matrix display (Fig. 3).
5. Load a PRESS water-line experiment and import the voxel
geometry and OVS definitions from the MRSI experiment;
then, perform automatic shimming (adjustments platform:
“local shim,” then Auto Shim and select “Map Shim”) in a
slightly larger volume than the MRSI VOI defined (e.g.,
þ0.5–1.0 mm in each plane—Fig. 3); a good performance
should generate a full-width at half maximum (FWHM) of
the water signal around 15–20 Hz; if above, e.g., 22 Hz,
consider repositioning the VOI and re-run Map Shim). Finally,
center the water resonance frequency (adjustments platform:
local frequency).
Fig. 4 B0 map after global and local shim adjustments. B0 map acquired in a
transversal plane, after local Map Shim adjustments. The larger square
represents the FOV for MRSI, and the inner ones the region for local
homogeneity adjustment and the VOI (smallest). The image shows that the B0
field is relatively homogeneous throughout the VOI region
6. Acquire a second B0 map of the whole brain (adjustments

platform: “B0 map”) and confirm the correct field homogene-
ity within the VOI region (Fig. 4); otherwise, repeat steps 5
and 6.
3.4 Acquisition 1. Before acquisition, the following parameters should be defined:

of MRSI Data (a) localization (voxel PRESS); k-space sampling (Acquisition-
weighted, Hamming window); number of averages
(NA 10—refers to the center of k-space in acquisition-
weighted mode); repetition time (TR, 1500–5000 ms); echo
time (TE: either long, 136 or 144 ms; or short, 12–20 ms);
sweep-width (SW, usually ~13 ppm); total number of points
(typically 2–4 k); number of dummy scans (e.g., 4); final matrix
size (MTXf); interpolation factor (defines the acquisition
matrix size, MTXi) (see Note 2).
2. With regard to radiofrequency pulses, we have used the follow-
ing configuration to acquire PRESS-MRSI data with 12 ms TE:
Hermite pulses (sharpness factor ¼ 3) for excitation (90 ) and

refocusing (180 ), with 1.0 and 0.8 ms, respectively
(~4000 Hz bandwidth), achievable with a standard
2.9–3.2 W reference power adjustment; Sech pulses for OVS,
with 1.5 ms spoiler duration and 25% strength.
3. Before acquiring the MRSI experiment, center the water fre-
quency and then adjust the water suppression with VAPOR
(Variable Pulse Power and Optimized Relaxation Delays mod-
ule, 300 Hz bandwidth), keeping a constant dB difference
between the two channels: start with high attenuation values
(low pulse power—only water peak detectable) and slowly
decrease until a spectral pattern can be distinguished in the
Acq/Reco display window, with a residual water peak signal;
then, acquire the water-suppressed MRSI experiment (see
Note 3).
3.5 Analysis 1. Once the MRSI acquisition is finished, spectra should be visu-
of MRSI Data ally inspected for quality assessment across the VOI (using
ParaVision CSIvisualizer). Thus, the spectral patterns should
be interpretable and without gross artifacts (extensively
reported in [18], e.g., outer-volume lipid contaminations due
VOI proximity to the scalp and/or suboptimal OVS adjust-
ment), and resolution (related shimming performance), SNR
(related to the number of averages), and water suppression
(related to the VAPOR pulse calibration) acceptable; if not,
the respective parameters should be revised (as indicated) and
the acquisition repeated.
2. For MRSI post-processing, third-party software is required.
Importantly, the individual raw data recorded from each chan-
nel are stored in the file “rawdata.job0,” whereas the “fid” file
is generated after reconstruction to the MTXf size defined, and
already combines the 4 channel data; if the data are analyzed
with Bruker TopSpin software, the “fid” file generated is
renamed to “ser,” keeping the same structure and size.
3. Before processing the MRSI data with available software tools,
such as jMRUI or LC Model, first make sure that the recon-
struction pipeline in ParaVision 6.0.1 and the output “acqp”
file generated are properly set to remove the voxel-to-voxel
180 flip filter and display the final matrix size (MTXf),
respectively (see Notes 4 and 5).
4. For spectral inspection and analysis of MRSI data, jMRUI 5.2 is
a suitable option; after loading the “fid” file, the voxel filling
order should be customized, as instructed in the manual
(matrix transposition required: e.g., for 32 32 MTXf size,
“0-0 1-32 2-64 (. . .) 1021-959 1022-991 1023-1023”);
finally, adjust zero order phase correction and apodization for
correct spectral display (effect of changing the MTXi size on
spectral resolution shown in Fig. 5) (see Note 5).
Fig. 5 MRSI spectra display and resolution. Two consecutive MRSI experiments (1 and 2) were acquired with
1500 ms TR, 12 ms TE, k-space sampled with a Hamming filter (10 averages in the center), 28.8 28.8 mm
FOV, 1.8 mm slice thickness, 10.8 10.8 mm VOI, and interpolated to a 32 32 MTXf size (1.5 μL display
voxel size): MRSI-1 was acquired with 16 16 MTXi size (5.8 μL nominal voxel size), 809 accumulations and
20 min acquisition time; whereas for MRSI-2 a 8 8 MTXi size was used (23.3 μL nominal voxel size), with
213 accumulations and 5 min acquisition time. (A) MRSI-2 data display with jMRUI 5.2: left-side, “mosaic
view” of the spectra selected within the VOI, after zero order phase adjustment and line-broadening with a
4 Hz apodization filter; right-side, respective HRref image showing the VOI region (12 12 matrix, in blue).
Spectra from the same voxel position (dark square) are shown enlarged for each MRSI experiment: (B) MRSI-
1; (C) MRSI-2. In smaller voxels (nominal size), SNR decreases but spectral resolution increases, as the tissue
volume becomes more homogeneous. Major peak assignments as indicated in Fig. 1, plus aspartate (Asp) and
gamma-aminobutyrate (GABA)
5. To quantify the MRSI data with metabolite basis sets, two

major software tools are available: jMRUI, with integrated
modules such as Quest; and LC Model. For the latter (LC
Model 6.3-IL), load the “fid” file and linearly fit the spectra
within the VOI as instructed in the manual. The results can be
displayed as regional color-coded maps (Fig. 6) using the fol-
lowing approach (or similar):
(a) Organize the individual quantification files generated by
LC Model for each voxel in matrix format (e.g., 12 11).
(b) Build an image with the same size as the HRref image, with
pixels set to zero everywhere but in the VOI region, which
is set as defined by the matrix generated in (a), and save in
nifti format—we performed this step with MATLAB
R2014a (Natick MA, USA).
(c) Visualize the image generated in (b) as a color-coded map
overlaid on the HRref image—for this we used the image
visualization software ITK-SNAP v3.4.0 (http://itksnap.
org/) (see Notes 6 and 7).
6. If classification is intended (rather than quantification), export
the MRSI data from jMRUI in text format and post-process
with additional software tools freely available, such as: DMPM
(http://gabrmn.uab.es/?q¼dmpm), for spectral normaliza-
tion; and SpectraClassifier [15], for pattern recognition analy-
sis. Additional details extensively reported in [17].
4 Notes
1. Additional material recommended for animal positioning in

the cradle: medical tape, for adequate restrainment; eye lubri-
cant, to prevent eyes from drying during the exam; and Vase-
line, to insert and position the rectal probe.
2. It is advisable to select a power of 2 MTXf size for MRSI;
otherwise, third-party software tools (such as LC Model and
jMRUI) are not able to read the data files.
3. Excessive pulse power should be avoided while adjusting the
water-suppression pulses (VAPOR module), which may lead to
an inversion of the residual water peak. This is not obvious in
the frequency-domain magnitude view of the Acq/Reco dis-
play window but can be detectable in the respective time-
domain section, as an inversion of the FID signal (real/imagi-
nary parts).
4. To process the MRSI data either with jMRUI or LC Model,
make sure that the MTX information in the “acqp” file is
correct (“##$ACQ_size ¼ (3)” describes MTXi and should
be modified to MTXf, if they have different sizes).
Fig. 6 MRSI quantification. Data acquired with 2500 ms TR, 12 ms TE, k-space sampled with a Hamming filter
(213 accumulations, with 10 averages in the center), 30.4 30.4 mm FOV, 1.9 mm slice thickness,
11.4 10.45 mm VOI, 8 8 MTXi size (27.4 μL nominal voxel size) interpolated to 32 32 MTXf (1.7 μL
display voxel size), and 8.5 min acquisition time. Top, linear fitting of an MRSI spectrum by LC Model 6.3-IL,
using artificial basis sets for macromolecules and 17 metabolites (those already detailed in Fig. 5 plus:
glucose, N-acetylaspartylglutamate and scyllo-inositol): MRS pattern (in black) shown with the fitting overlaid
(in red), including the respective baseline correction (in gray) and fitting residual (on top). Bottom, color-coded
maps of metabolite ratios (to total creatine, CrT) within the VOI area, as determined by LC Model fitting,
overlaid on the corresponding T2-w reference image: left, glutamine and glutamate pool (Glx/CrT); right,
N-acetylaspartate and N-acetylaspartylglutamate pool (NAAT/CrT); voxels displayed in black (set to zero)
indicate Cramér-Rao lower bounds (reliability indicator for the fitting) 20%; voxel position of the spectrum
displayed on top indicated by a black square
Fig. 7 Reference water-line MRSI. Spectral matrix corresponding to the PRESS-VOI region shown in Fig. 6,
indicating a homogeneous spatial distribution of the water peak in each voxel (water-suppression module off).
Data displayed with the ParaVision module CSI visualization tool
5. For jMRUI, additionally make sure that the Output Matrix

Type, in the reconstruction pipeline in ParaVision 6.0.1, was
set to “real” instead of “magnitude”; otherwise, a phase filter is
present which generates voxel-to-voxel 180 flips.
6. During post-processing of the MRSI data, it is often advisable
to discard from the analysis voxels at the edges of the VOI, due
to their low SNR [16]. This effect is particularly apparent in the
inferior part of the MRSI grid in Fig. 1 and in both color-coded
maps shown in Fig. 6 (voxels in black indicating metabolite
quantifications with error estimates above 20%).
7. For absolute metabolite quantifications, the following modifi-
cations to the protocol should be taken into account: 1. before
running the MRSI experiment, acquire a brain volumetric T1-
w and/or T2-w MRI, to determine the partial fractions of gray
and white matter, and cerebrospinal fluid in each voxel—useful
for partial-volume corrections; 2. acquire an additional MRSI
experiment without water suppression, to generate a water-line
spectral matrix (Fig. 7)—reference for quantification and also
useful for eddy current corrections; 3. consider acquiring
MRSI with longer TR (5 s), to reduce the interference of T1

saturation effects (although often incompatible with the total
acquisition time available for the study).
Acknowledgments
This work was sponsored by grants from: La Caixa Foundation;

AGAUR 2014 SGR, grant n 928; RD12/0036/0010, integrated
in Plan Nacional de IþDþI and co-funded by ISCIII-Subdireccion
General de Evaluacion and Fondo Europeo de Desarrollo Regional
(FEDER) (PI14/00595); Fundacio La Marato de TV3 (201441
10); and CERCA Programme/Generalitat de Catalunya. We are
indebted to the Experimental MRI 7T Unit of the IDIBAPS, for
the technical help.
References
1. Mlynarik V, Kohler I, Gambarota G, Vaslin A, 7. Seuwen A, Schroeter A, Grandjean J, Rudin M

Clarke PG, Gruetter R (2008) Quantitative (2015) Metabolic changes assessed by MRS
proton spectroscopic imaging of the neuro- accurately reflect brain function during drug-
chemical profile in rat brain with microliter induced epilepsy in mice in contrast to fMRI-
resolution at ultra-short echo times. Magn based hemodynamic readouts. NeuroImage
Reson Med 59(1):52–58 120:55–63
2. Parry-Jones AR, Liimatainen T, Kauppinen 8. Boska MD, Lewis TB, Destache CJ, Benner EJ,
RA, Grohn OH, Rothwell NJ (2008) Nelson JA, Uberti M, Mosley RL, Gendelman
Interleukin-1 exacerbates focal cerebral ische- HE (2005) Quantitative 1H magnetic reso-
mia and reduces ischemic brain temperature nance spectroscopic imaging determines thera-
in the rat. Magn Reson Med 59(6): peutic immunization efficacy in an animal
1239–1249 model of Parkinson’s disease. J Neurosci 25
3. Garcı́a-Martı́n ML, Herigault G, Rémy C, (7):1691–1700
Farion R, Ballesteros P, Coles JA, Cerdán S, 9. Alf MF, Lei H, Berthet C, Hirt L, Gruetter R,
Ziegler A (2001) Mapping extracellular pH in Mlynarik V (2012) High-resolution spatial
rat brain gliomas in vivo by 1H magnetic reso- mapping of changes in the neurochemical pro-
nance spectroscopic imaging: comparison with file after focal ischemia in mice. NMR Biomed
maps of metabolites. Cancer Res 61 25(2):247–254
(17):6524–6531 10. Delgado-Goni T, Martin-Sitjar J, Simões RV,
4. Ortega-Martorell S, Lisboa PJ, Vellido A, Acosta M, Lope-Piedrafita S, Arus C (2013)
Simoes RV, Pumarola M, Julia-Sape M, Arus Dimethyl sulfoxide (DMSO) as a potential
C (2012) Convex non-negative matrix factori- contrast agent for brain tumors. NMR Biomed
zation for brain tumor delimitation from MRSI 26(2):173–184
data. PLoS One 7(10):e47824 11. Juchem C, Merkle H, Schick F, Logothetis
5. Liimatainen TJ, Erkkila AT, Valonen P, NK, Pfeuffer J (2004) Region and volume
Vidgren H, Lakso M, Wong G, Grohn OH, dependencies in spectral line width assessed by
Yla-Herttuala S, Hakumaki JM (2008) 1H 1H 2D MR chemical shift imaging in the mon-
MR spectroscopic imaging of phospholipase- key brain at 7T. Magn Reson Imaging 22
mediated membrane lipid release in apoptotic (10):1373–1383
rat glioma in vivo. Magn Reson Med 59 12. von Kienlin M, Ziegler A, Le Fur Y, Rubin C,
(6):1232–1238 Décorps M, Rémy C (2000) 2D-spatial/2D-
6. Simões RV, Delgado-Goñi T, Lope-Piedrafita- spectral spectroscopic imaging of intracerebral
S, Arús C (2010) 1H-MRSI pattern perturba- gliomas in rat brain. Magn Reson Med 43
tion in a mouse glioma: the effects of acute (2):211–219
hyperglycemia and moderate hypothermia. 13. Naressi A, Couturier C, Devos JM, Janssen M,
NMR Biomed 23(1):23–33 Mangeat C, de Beer R, Graveron-Demilly D
(2001) Java-based graphical user interface for classification of brain tumors in mice with per-
the MRUI quantitation package. MAGMA 12 turbation enhanced (PE)-MRSI. Integr Biol
(2–3):141–152 (Camb) 4(2):183–191
14. Pfeuffer J, Tkac I, Provencher SW, Gruetter R 17. Simões RV, Candiota AP, Julià-Sapé M, Arús C
(1999) Toward an in vivo neurochemical pro- (2013) In vivo magnetic resonance spectros-
file: quantification of 18 metabolites in short- copy imaging (MRSI) and ex vivo quantitative
echo-time (1)H NMR spectra of the rat brain. J neuropathology by high resolution magic angle
Magn Reson 141(1):104–120 spinning proton magnetic resonance spectros-
15. Ortega-Martorell S, Olier I, Juliá-Sapè M, Arús copy (HR-MAS). In: Springer CS Jr
C (2010) SpectraClassifier 1.0: a user friendly, (ed) Animal models of brain tumors, Neuro-
automated MRS-based classifier-development methods volume series, vol 77. Springer,
system. BMC Bioinformatics 11:106 New York, pp 329–365
16. Simões RV, Ortega-Martorell S, Delgado- 18. Kreis R (2004) Issues of spectral quality in
Goni T, Le Fur Y, Pumarola M, Candiota AP, clinical 1H-magnetic resonance spectroscopy
Martin J, Stoyanova R, Cozzone PJ, Julia- and a gallery of artifacts. NMR Biomed 17
Sape M, Arus C (2012) Improving the (6):361–381
Part IV
Special MRI Techniques

Chapter 13
Susceptibility Weighted MRI in Rodents at 9.4 T

Ferdinand Schweser, Marilena Preda, and Robert Zivadinov
Abstract
Susceptibility Weighted Imaging (SWI) is an established part of the clinical neuroimaging toolbox and,
since its inception, has also successfully been used in various preclinical studies. Exploiting the effect of
variations of magnetic susceptibility between different tissues on the externally applied, static, homoge-
neous magnetic field, the method visualizes venous vasculature, hemorrhages and blood degradation
products, calcifications, and tissue iron deposits. The chapter describes in vivo and ex vivo protocols for
preclinical SWI in rodents.
Key words SWI, Preclinical, MRI, Phase imaging, Magnetic susceptibility, Rodents, Mice, Rat,
Gradient echo, Bold
1 Introduction
Susceptibility weighted imaging (SWI) [1–4] is a technique that

visualizes variations of magnetic susceptibility in biological tissues.
It is the most sensitive technique for the visualization of the venous
vasculature, hemorrhages and blood degradation products, calcifi-
cations [5–10] and tissue iron deposits [11–18]. Ogawa et al. [19]
carried out the first experiments with SWI in a rat at 8.5 T in their
1990 seminal work on functional MRI. Since then, SWI has
become an established part of the clinical neuroimaging toolbox
[10, 20, 21] and has successfully been used in preclinical studies
with rodents [22–28] and larger animals [29].
SWI exploits the effect of tissues’ magnetic susceptibility varia-
tions on the externally applied, static, homogeneous magnetic field.
Susceptibility variations can induce field perturbations on two dif-
ferent spatial scales, a microscopic and mesoscopic scale similar to
and below the voxel size and a macroscopic scale larger than the
voxel size [30]. Microscopic and mesoscopic field perturbations
result in spin dephasing, which is visible as a more rapid effective
transverse relaxation (increased R2* ¼ 1/T2* constant). Macro-
scopic field inhomogeneities may not necessarily lead to a
205
206 Ferdinand Schweser et al.
substantial change in R2*, but instead, affect the voxel-average

Larmor frequency. Both effects become apparent on strongly
T2*-weighted magnitude and phase images, respectively, i.e.,
MRI using radio-frequency spoiled gradient-echo recalled (GRE)
sequences with relatively long echo times [4].
The blood oxygen level dependence (BOLD) effect [19, 31]
describes the dependence of the magnetic susceptibility of blood on
its oxygenation level, a basis for the visualization of the venous
vasculature, blood deposits, and degradation products (such as
hemorrhages) with SWI. BOLD is related to a change in the
electron spin state of the heme’s ferrous iron from high (paramag-
netic) to low spin (diamagnetic) when binding oxygen. Since
plasma surrounding hemoglobin and parenchyma surrounding
veins are diamagnetic, the low oxygenation level of venous blood
is accompanied by inhomogeneous magnetic environments on
microscopic (intravascular) and macroscopic (extravascular) spatial
scales, a basis for the visualization of veins as small as 15–30 μm in
diameter with SWI at 9.4 T [32]. Another source of contrast in
SWI is non-heme tissue iron, which is predominantly stored in the
superparamagnetic ferrihydrite crystal core of the iron-storage pro-
tein ferritin [33], increasing the bulk magnetic susceptibility of
tissues with the load of iron. Also, calcifications, which appear to
be diamagnetic [34–38], can be seen on SWI.
To facilitate image inspection, clinical SWI typically combines
the phase and magnitude contrast into a composite image
[21, 39]. This combination is achieved by delineating
susceptibility-related field perturbations visible on the phase images
with a dedicated intensity filter and multiplying the resulting
so-called “phase mask” with the corresponding T2*-weighted mag-
nitude images [3]. Such image post-processing is particularly bene-
ficial at clinical field strengths (B0; TE·B0 60–85 ms [40]), where
the phase images show a much improved anatomical contrast com-
pared to magnitude images, which usually do not clearly visualize
smaller veins alone. In fact, phase images contain a high degree of
anatomical contrast, also allowing a direct inspection, which has
recently received increased interest [12–14, 41–45].
An explanation for the relatively low T2*-weighted magnitude
image contrast in the clinical setting is that, at lower field strengths,
the venous blood and surrounding parenchyma have relatively
similar R2* constants. However, R2* of blood increases exponen-
tially with B0 [40], compared to the R2* of surrounding paren-
chyma, which increases approximately linearly [46]. This difference
in field-strength dependence diminishes the benefit of combining
magnitude and phase images at higher field strengths. Hence, at
field strengths typically used for preclinical MRI (B0 7 T), the
intravascular contribution of blood to the magnitude image con-
trast is often sufficient to visualize veins [29, 32, 47]. In particular,
at higher field strengths, the benefit from a high voxel aspect ratio
SWI 207
commonly used in clinical SWI to increase mesoscopic spin dephas-

ing around veins, and the combination of the magnitude with the
phase mask seems to drastically diminish [29, 40]. Our experience
is that the typical combination of magnitude and phase images,
while enhancing the contrast of veins slightly, does not significantly
improve the conspicuity of veins in preclinical imaging. On the
contrary, the combination of phase and magnitude often leads to
additional susceptibility artifacts in the resulting hybrid images that
can render the interpretation of image contrast more difficult.
Hence, we usually do not employ phase-based filtering in preclini-
cal MRI at 9.4 T. However, for preclinical imaging at lower field
strengths the classical combination of phase and magnitude may
yield improved vessel conspicuity.
In this chapter, we describe our established in vivo (Fig. 1) and
ex vivo protocols (Fig. 2) for preclinical SWI in rodents at 9.4 T.
2 Materials
2.1 General Imaging 1. 20 cm diameter horizontal-bore 9.4 Tesla magnet (Bruker

Materials Biospin; BioSpec 92/20 USR) housed in a noise-dampened
Faraday room cage that shields all relevant imaging frequencies
(>60 dB shielding) (see Note 1).
2. Actively shielded gradient coil with 114 mm inner diameter
and integrated higher-order shims (Bruker Biospin; BGA-12S
HP; 440 mT/m gradient strength; 3440 T/m/s maximum
linear slew rate) (see Note 2).
3. Scanner control workstation running both ParaVision 5.1 and
ParaVision 6.0.1 (see Note 6).
4. Spoiled multi-echo gradient echo (GRE) pulse sequence
(called “MGE” in Bruker ParaVision) for in vivo experiments
(see Notes 3 and 5) and spoiled single-echo GRE pulse
sequence for postmortem experiments (called “FLASH” in
Bruker ParaVision).
2.2 In Vivo 1. Animals: Rats or mice without any metal parts in or attached to
Experiments their bodies, such as metal ear tags (see Note 7). We have been
using primarily adult female Swiss Jim Lambert (SJL/J) mice
for Multiple Sclerosis (MS) research and adult male Wistar rats
for Traumatic Brain Injury (TBI) research, but SWI should
work in any animal that allows a stable anesthesia.
2. Radio-frequency (RF) coil configuration for rat experiments
(see Note 10): Cross-coil configuration with an anatomically
shaped four-element receive-only1H RF array coil with 2 2
coil element topology (see Note 8) and a circularly polarized
volume RF coil with 86 mm inner diameter for transmission.
Fig. 1 Illustration of in vivo SWI at 9.4 T in the lateral fluid percussion rat model of TBI. Magnitude images of
the 3D multi-echo sequence were averaged, bias-field corrected to mitigate signal inhomogeneities due to the
use of a surface receiver coil. Finally, minimum intensity projection (mIP) images were calculated over 1 mm
in coronal (top left), sagittal (top right), and horizontal (bottom) slices. The filtered phase image shows
magnetic field perturbations due to susceptibility variations in the brain
The receive coil is attached to an animal cradle with integrated

water hoses for body temperature control, integrated provision
for inhalation of anesthesia, and an optimized three-point head
fixation using a tooth bar and ear plugs (see Note 9). The
animal cradle is attached to an automatic, motorized sample
positioning system with laser cross-hair sighting mounted on
the front of the magnet.
SWI 209
Fig. 2 Illustration of ex vivo SWI in the lateral fluid percussion rat model of TBI. Shown are mIPs of bias-field
corrected magnitude images over 50 μm. White arrows with straight tail point at TBI-related hemorrhages
(hypointense), arrows with circle-tails point at neurons in the caudate, and arrows with box-tail point at veins
3. RF coil configuration for mouse experiments (see Note 10):

Anatomically shaped, cryogenically cooled (20–30 K)1H trans-
ceiver coil with 1 2 coil element array structure (Bruker
BioSpin;1H MRI CryoProbe) mounted inside the magnet
bore with special coil mounting equipment (see Note 18).
Temperature control unit for the coil head, two additional
external preamplifiers, closed-cycle cryogenic platform consist-
ing of cooling unit, primary Helium compressor, and Helium
transfer line from the equipment room into the magnet room.
Animal sled with integrated water hoses for body temperature
control, integrated provision for inhalation of anesthesia, and
fixation using a tooth bar and custom-made, anatomically
shaped lateral and vertical head fixation parts as well as
custom-made head-holding cushions (see Note 11). Cryo-
Probe simulator to test positioning outside of the magnet.
4. MR-compatible fluid heating system with thermo-controlled
water reservoir connected to hoses embedded in the animal
cradles. Additional water heating pads for rats and mice.
5. MR-compatible small animal monitoring and gating system
(Model 1030; SA Instruments, Inc.; Stony Brook, NY) with
software for real-time display of respiration and temperature.
Fiber-optic connections from magnet room to laptop running
the software in the control room. Rectal thermistor probe and
pneumatic respiration pillow sensor are connected to a battery-

powered monitoring unit placed on the animal cradle (inside
the magnet during MRI). Lubricated protective plastic sleeves
for the thermistor probe to avoid corrosion. Pulse oximeter
with bidirectional fiber optic sensors extending from the mea-
surement module outside the magnet to the animal inside the
bore to measure arterial blood oxygen saturation (SpO2), car-
diac plethysmogram waveform, heart rate, and pulse
distension.
6. Calibrated small-animal inhalation anesthesia equipment
located outside the magnet room, including vaporizer, isoflur-
ane, and medical-grade oxygen (see Note 12).
7. Tygon tubing running from the anesthesia equipment in the
control room to the nose cone on the animal cradle.
8. Ducted certified chemical fume hood (see Note 13).
9. Transparent induction chamber with detachable lid; placed
under the fume hood. Lid of container is connected to the
vaporizer and has an additional outlet to avoid overpressure. A
thermostatically controlled electric heating pad, located under
the anesthesia chamber, avoids hypothermia during anesthesia
induction.
10. Surgical tape used to tape down the monitoring sensors and
the animal onto the animal cradle.
11. Ophthalmic ointment.
12. Chlorine dioxide disinfectant, 100 ppm.
13. Personal protective equipment. Latex or nitrile gloves and lab
coat may be sufficient. Eye protection for disinfection and
masks if allergic to rodent dander/urine.
2.3 Ex Vivo 1. Neutral buffered formalin (NBF).

Experiments ( See 2. 1 mmol/ml Gadobutrol (Gadavist, Bayer).
Note 14 )
3. Rat.
4. 20 ml Galden HT80 perfluorinated polyether (Solvay Specialty
Polymers, Alpharetta, USA) (see Notes 15 and 16).
5. 20 ml Falcon Centrifuge Tube with conical bottom.
6. Mounting putty, removable.
7. Soft foam to mechanically stabilize the tissue specimen inside
the Falcon tube.
8. RF coil configuration (see Notes 10 and 17 ): Cryogenically
cooled1H transceiver coil (see mouse experiments).
SWI 211
2.4 Data Analysis 1. MATLAB (R2013b; The MathWorks, Natick, MA).

2. Linux workstation (Ubuntu 12.04) with 48 cores (dual Intel
Xenon E5-2697v2 CPUs at 2.7 GHz) and 396 GB RAM (see
Note 19).
3. MATLAB code to load MRI data (see Note 20).
3 Methods
3.1 Animal Perfusion 1. Perfuse animal at 5 ml/min with 1.5 ml per gram body weight
for the Ex Vivo of 10 mM Gadobutrol (10 μl of Gadobutrol for every 1 ml
Experiment solution) phosphate-buffered saline (PBS) followed by 1.5 ml
per gram of 10 mM Gadobutrol NBF.
2. Dissect brain.
3. Post-fix brain for 24 h in NBF with 10 mM Gadobutrol. Keep
it at 4 C.
4. After 24 h, transfer brains to PBS with 10 mM Gadobutrol.
Store brain at 4 C.
5. Transfer sample from the cold room to MRI control room in
good time before the experiment to allow warming up to room
temperature.
6. Right before the MRI experiment, transfer the brain to the
Falcon tube (see Notes 21 and 22).
7. Fill the Falcon tube with Galden and screw on the cap (see Note
23).
8. Transfer the tube into the magnet room.
9. Gently tap the tube onto the bench several times to remove/
release air bubbles from the brain’s surface and from within the
foam. Place the tube on the animal sled with its conical tip in
the nose cone. Mechanically fix the tube with the mounting
putty. Use enough putty so that the end of the tube with the
cap is slightly higher than the end of the conical tip, trapping air
bubbles in the cap.
3.2 Pre-scan 1. Connect the additional animal heating blanket to the water
Preparations for In tubing.
Vivo Experiments 2. Activate the water heating mechanism of the MRI animal cra-
dle. Set the target temperature to 45 C (see Note 24).
3. Switch on the electric heating pad under the anesthesia
chamber.
4. For mice: Prepare surgical tape to align the two SpO2 sensor
pieces (without the clamp) such as to allow for connection to
the base of the mouse tail (see Note 25).
5. Place paper towel on the bottom of the induction chamber to

retain urine and feces.
6. Connect the anesthesia chamber with the vaporizer, but keep
oxygen flow shut off.
7. Mice and ex vivo: Make sure that the cryogenic coil head is
warmed to 38 C.
8. Ensure that the anesthesia tubing is connected to the nose cone
and the setup is otherwise ready for the animal to be scanned.
9. Confirm that oxygen and isoflurane quantity/volumes are suf-
ficient for the duration of the experiment.
10. Ensure that the monitoring system is installed and working
properly (check battery) (see Note 26).
11. Prepare pieces of surgical tape necessary for animal fixation
(feet, the rectal probe, respiration pad, pulse oxygenation
clamp, heating blanket, surface receiver coil) to speed up the
positioning process once the animal is anesthetized.
12. Place the custom-made head-holding cushion on the animal
sled and prepare other fixation mechanisms.
13. Enter the animal information into the scanner software, select
the desired protocol, and queue the first scout sequence (see
Note 27).
3.3 Induction 1. Transport the animal in its cage from the animal holding room
of Anesthesia to the MRI control room and place it under the fume hood
and Positioning for In until ready for the induction. Prevent the animal from becom-
Vivo Experiment ing stressed by bright light, loud noises, strong smells, etc.
Provide drinking water to the animal.
2. Using gloves, transfer the animal to the induction chamber. Set
oxygen flow to approximately 1.0 l/min and isoflurane con-
centration to 1%. After 1 min, increase isoflurane to 4–4.5%.
3. Monitor the animal closely; gently shake induction chamber
regularly and inspect response of animal. Once the animal
reaches a depth of anesthesia that allows transporting into the
magnet room (typically indicated by ~60 breaths per minute in
mice and ~50 breaths per minute in rats) (see Note 28), which
usually happens in less than 10 min.
4. Once anesthesia is deep enough, shut off the oxygen flow and
disconnect the vaporizer from the anesthesia chamber. Con-
nect the vaporizer with the tubing running into the magnet
room and set oxygen flow to 1.2 l/min and isoflurane to
2.5–3%.
5. Using gloves, open the anesthesia chamber and quickly trans-
port the animal into the magnet room. Cover the animal with
both hands minimizing heat loss. Inside the magnet room,
SWI 213
place the animal on the cradle prone and gently push the
animal’s nose into the nose cone. Monitor physiological signs
to ensure that animal stays under stable anesthesia. If anesthesia
became lighter during the transport, wait until animal is stable
again (see Note 29).
6. Place the respiration pad under the animal’s mid-abdominal
area (rat) or the animal’s upper abdominal area (mice), or atop
of the animal and tape it onto the cradle to avoid movement
during the exam.
7. Gently pull animal into nose cone using the tooth bar. Hold the
tooth bar with one hand while with the other hand gently push
the animal’s head down into the custom-made head support.
Pull animal’s tail gently to make sure the teeth are still hooked
up in the tooth bar. Verify the head is horizontal by looking at
the position of the eyes relative to the animal cradle. Hold the
tooth bar in that position with one hand and tighten the screw
with the other hand.
8. Continue with the lateral fixation of the head. For rats, push
the side ear bars gently into the rat’s ears. For mice, press the
side support squeeze gently (see Note 30) against the head
from the sides and tighten the screws into place. Avoid rotation
of the head and ensure that the animal’s head is centered in the
left-right direction (see Notes 31 and 32).
9. Place a generous amount of eye lubricant onto eyes to avoid
drying out.
10. Gently, stretch animal’s feet and tape them to the bottom of
the sled. This stretching will ensure that the animal will not
turn/twist if, by chance, wakes up during scanning.
11. Look at the animal from the side (horizontally). Your fixation is
right if you do not see any respiration-related motion of the
hair at the top of the head.
12. For mice: Position the custom-made soft tape cushion over the
top of animal’s head. Secure it with tape.
13. For rats: Place the surface coil atop of the animal’s head. Watch
for the ears not to get caught on the sides. The center of the
coil should coincide with the center of the brain. Move the
nose cone forward or backward if needed. Tape the coil in
place, pushing down slightly.
14. Insert the temperature probe into the lubricated sleeve and
very gently slide it into the rectum using circular motions,
deep enough to ensure correct reading throughout the experi-
ment (see Note 33). Tape the rectal probe wire to the sled (not
to the animal’s tail!). Stretch gently the animal’s tail and tape to
the sled as well.
15. Install the SpO2 sensor. For mice, slide animal’s tail into the
opening of the prepared pulse oxygenation sensor setup. Make
sure sensors are as far up at the base of the tail as possible to
ensure strong reading throughout the experiment. Tape the
cables in place onto the sled’s bottom and sides. For rats, put
the rat-clap on the hind foot of the rat. Ensure proper reading
of the sensor and reposition sensors if needed. Make sure that
the foot is held in the clamp tight enough to ensure strong
reading throughout the experiment. To stabilize it even more,
tape the clamp in place to the sled’s side. Watch the monitor for
a while to make sure the signal is strong and stable.
16. If the respiration pad is placed atop of the animal, position it
now. Tape it central to animal’s back above the lung area and
the cradle, snug but not tight. Verify that the respiration wave
is stable.
17. Cover the animal’s back with the heating blanket. Tape it in
place around the animal, gentle yet snug.
18. Watch the vitals to see how stable and strong they are.
19. Mice and postmortem: Slide the animal cradle into the cryo-
probe simulator to check and resolve mechanical obstructions,
if necessary. Lift the cradle against the coil head and closely
watch respiration rate. The respiration rate should remain the
same—if it changes, reposition!
20. Slide the sled into the magnet.
21. Adjust the concentration of Isoflurane to achieve a stable res-
piration rate ideally ~70 breaths/min.
22. Record the initial body temperature reading and adjust water
bath temperature throughout the scanning to maintain
it. General body temperature range is 36.5–37.5 C, and varies
from animal to animal.
3.4 MRI Scan 1. Acquire a standard three-plane scout with default, automated
system adjustments. Confirm that the brain is positioned at the
magnet isocenter. If this is not the case, modify the location of
the animal sled. When using the cryogenic coil, also inspect the
signal inhomogeneity profile and confirm that the coil is axially
centered at the magnet’s isocenter. Adjust location of animal
bed and coils and run scout again until positions are acceptable.
In addition, for ex vivo, ensure that the specimen is rotated
such that a homogeneous intensity profile will be obtained in at
least one slice orientation.
2. In vivo experiment: Run a quick sagittal, single-slice 2D MGE
scan to check if respiration-related movements risk to deterio-
rate the SWI images (see Note 34). For mice: TE1 ¼ 3.99 ms,
ΔTE ¼ 8 ms, 5 echoes with positive readout (flyback),
SWI 215
Fig. 3 Illustration of images obtained from the 2D motion-sensing pulse sequence in a mouse. The top row
shows typical images obtained when the positioning is insufficient for SWI (increasing TE from left to right).
The white arrows point at motion-related artifacts. After improving the fixation, no artifacts are discernible on
the images of the motion-sensing sequence (bottom row)
TR ¼ 90 ms, averages ¼ 1, 18 tip, 2 ms Hermite pulse, no

acceleration/zero-filling, 0 ms gradient stabilization time,
85 kHz spectral bandwidth, 33% echo position, 40% read
spoiler for 2 ms, 1.83 ms 2D phase gradient, FOV
27 14.3mm2, 0.286 mm slice thickness, matrix 252 180,
H-F read orientation, 10 dummy scans, fat suppression
(Gauss512, 1401 Hz), no gating, total acquisition time: 16 s.
3. Load the resulting images into the Image Display and inspect
the quality of the images at later echo times. Improve position-
ing if there are substantial motion artifacts or untypical signal
loss at later echoes (with a grainy/wavy pattern, see Fig. 3).
Repeat the motion-sensing scan and the re-positioning until
the image quality at later echoes is acceptable (see Note 36).
4. Tune and match coils (see Note 37).
5. Perform a field map-based higher-order shim to a volume
extended to cover the largest possible portion of the brain
while avoiding tissue outside of the brain (see Note 38).
6. Cryogenic coil configuration only: Adjust transmit pulse refer-
ence gain in a coronal, 2 mm thick slice. Position the slice to the
most upper top of the brain. For the ex vivo experiment, place
the calibration slab at about 1/3 from the upper part of the
brain during this step.
7. Prescribe the high-resolution 3D MGE sequence (see Note
44). For mice (ParaVision 5.1): TE1 ¼ 2.38 ms, ΔTE ¼ 4.04 ms
(see Note 48), 9 echoes with positive readout (flyback) (see
Note 40), TR ¼ 90 ms (see Note 46), averages ¼ 1, 18 tip
(see Note 47), 0.35 ms Hermite pulse, no acceleration/zero-
filling, 0 ms gradient stabilization time, 85 kHz spectral band-
width, 33% echo position, 40% read spoiler for 2 ms, linear
phase encoding order, 0.83 ms 2D phase gradient, FOV
27 14.3 8 mm3 (see Note 26), matrix 252 180 100

(80 80 80 μm3 nominal resolution) (see Note 41), coronal
slab orientation (see Note 39), H-F read orientation,
10 dummy scans, fat suppression (Gauss512, 1401 Hz), no
gating, total acquisition time 27 min.
For rats (ParaVision 5.1): TE1 ¼ 2.31 ms, ΔTE ¼ 3.10 ms
(see Note 48), 16 echoes with bipolar readout (see Note 40),
TR ¼ 100 ms (see Note 46), averages ¼ 1, 19 tip (see Note
47), 0.35 ms Hermite pulse, no acceleration/zero-filling, 0 ms
gradient stabilization time, 50 kHz spectral bandwidth, 50%
echo position, 40% read spoiler for 1.97 ms, linear phase
encoding order, 0.6 ms 2D phase gradient, FOV
30 30 14 mm3 (see Note 44), matrix 135 208 97
(222 144 144 μm3 nominal resolution) (see Note 41),
coronal slab orientation (see Note 39), H-F read orientation,
10 dummy scans, fat suppression (Gauss512, 1401 Hz), no
gating, total acquisition time 34 min.
Ex vivo (ParaVision 6.0.1 (see Note 42); single-echo
FLASH (see Note 43)): TE ¼ 8 ms, TR ¼ 37 ms (see Note
46), averages ¼ 20, 67 tip (see Note 47), 0.35 ms Hermite
pulse, no acceleration/zero-filling, 44.6 kHz spectral band-
width, 50% echo position, auto read/slice spoiler calculation,
linear phase encoding order, FOV 14 24 9 mm3, matrix
467 800 300 (30 μm isotropic) (see Note 41), coronal slab
orientation, X read orientation, three dummy scans, RF
spoiling, total acquisition time 49 h and 20 min.
8. Ex vivo experiment only: Check that your high-resolution
sequence is set up properly and all calibrations succeeded. To
this end, duplicate the high-resolution scan, increase the inter-
polation factor in both phase-encode directions, and reduce the
number of averages so that the acquisition does not take more
than a few minutes. Check the image quality on this
low-resolution scan and improve calibrations or prescription.
Repeat the low-resolution scans until everything is set up cor-
rectly. Pay particular attention to hypointense saturation bands
at the top of the brain. If a significant decline in image intensity
is visible in the upper part of the brain (close to the coil), move
the calibration slab of the transmit gain calibration slightly up
and repeat calibration (see Notes 49 and 35).
9. Deactivate the image reconstruction. The amount of data pro-
duced by high-resolution multi-echo imaging with multiple
receiver channels may exceed the system’s memory limits or
the image reconstruction takes relatively long, preventing con-
tinuation of scanning with other sequences. If your system has
sufficient memory and you can afford letting the system remain
idle while the images are being reconstructed, you may leave
the image reconstruction activated. In ParaVision the image
SWI 217
reconstruction may be deactivated by deactivating

GS/GO_online_reco and GS/GO_reco_display in the Single
Parameter Tab (ParaVision 6.0.1) and Edit GS/“online reco”
and “reco display” as well as Edit acqp/“online reco” and
“reco display” in GO class (ParaVision 5.1).
10. Start the high-resolution scan (see Note 50).
11. After the exam has completed, transfer the imaging data to
your local image processing workstation.
3.5 Maintain 1. Respiration should be as stable as possible throughout the

Anesthesia During MRI scanning process. However, variability may take place due to
imaging gradient noise, possible positioning discomfort, or
due to animal illness. It is important to remember that dramatic
changes in the level of Isoflurane will result in even more
instability. Thus, it is critical to monitor the respiration contin-
uously and make minute changes only when the instability
persists more than 30–40 s. If the respiration is very low and
variable with significant periods of no peaks recorded, termi-
nate the scan and allow the animal to recover. It is also impor-
tant to monitor the heart rate in tandem with the respiration
and body temperature. Animals may have a congenital heart
condition that gets exacerbated by anesthesia.
2. Closely monitor the animal’s body temperature to avoid hypo-
and hyperthermia, which would result in a less stable anesthesia
potentially associated with motion artifacts. Maintain the tem-
perature at (36 0.5) C by manually up- and down-regulating
the temperature of the water reservoir (see Note 51).
3. Closely monitor the animal’s respiration rate. If the respiration
rate is too low/high, decrease/increase the anesthetic (see
Note 53).
3.6 Recovery 1. Carefully remove all tape that was used to position the blood
and Preparation oxygenation sensor, the rectal probe, feet to the cradle, head
for Next Scan soft tape cushion, and so forth.
2. Release the head from the restraint devices.
3. Turn off anesthesia.
4. Remove animal from nose cone and recover in a well-ventilated
induction box (lid open) that was kept warm with a heating pad
from below. Optionally, use infrared heating lamp from top but
ensure that animal does not overheat. Monitor recovery. Pedal
reflex should be noticeable after 1 min. Pinch animal toe or the
very tip of their tail. An immediate jerk reaction will take place
if the animal has pain reflexes.
5. Disinfect (using MB-10) all components wearing gloves, eye
protection, and lab coat.
6. Unplug battery pack and plug into charger.

7. Power off water bath heating mechanism.
8. Refill Isoflurane.
9. Ex vivo: Remove the brain from the tube and rinse with saline.
Transfer back to the storage solution. Recover the used PFPE
for later experiments (see Note 52).
3.7 Post-processing 1. Load the MRI data with MATLAB (see Note 20).
and Data Analysis 2. Sinc-interpolate the spatial resolution of the images by a factor
of 1.5 or higher (see Note 55).
3. Reconstruct magnitude images from the raw multichannel
k-space data with the sum-of-squares technique [48] (see
Note 56).
4. Perform a bias field correction (e.g., using N4-ITK [49]) to
reduce the signal inhomogeneity of the surface coil (see Note
57).
5. Calculate a sliding minimum intensity projection (mIP) of the
magnitude image over at least 1 mm of consecutive slices (see
Notes 57 and 58).
6. Combine the multichannel phase images and remove large-
scale field inhomogeneities (background fields) with a high-
pass filter (see Notes 60 and 62).
7. Combine the phase from different echo times into one single
phase image that may be interpreted (see Note 60).
8. Export the resulting images to your preferred data format, e.g.,
NIfTI (see Note 59).
4 Notes
1. SWI may be performed at any magnetic field strength currently

used for preclinical MRI, including clinical field strengths
[28]. However, SWI benefits from higher magnetic field
strength. Higher field strengths provide comparable contrast
to lower field strengths at shorter echo times, hence, allowing a
reduction of the repetition time (TR). In combination with the
higher nuclear spin polarization, this enables shorter scan times
or higher spatial resolution. However, imaging at higher field
strengths is also prone to macroscopic field inhomogeneities at
the boundaries of the brain associated with phase gradients and
leading to signal dephasing.
2. SWI does not require particularly powerful imaging gradients.
The traditional single-echo SWI protocol employs a relatively
low readout bandwidth, posing minimal requirements on the
SWI 219
gradient system. The multi-echo sequence used in our protocol

requires more powerful gradients to enable a higher number of
echoes for a given TR.
3. In the clinical setting, SWI employs a fully flow compensated
spoiled 3D single-echo fast low angle shot (FLASH) pulse
sequence with a relatively long echo time (T2*-weighted) and
low readout bandwidth [3, 4]. The flow compensation miti-
gates magnitude signal loss due to incoherent addition of spin
phases (phase cancellation) as well as signal displacement
(ghosting) in both phase and magnitude [50]. While a similar
pulse sequence may in principle also be used in the preclinical
setting (see Note 4), we commonly use the multi-echo GRE
sequence with a high number of echoes and high readout
bandwidth. Although the increased readout bandwidth implies
that each individual echo image has reduced SNR compared to
images acquired with lower bandwidth, averaging of the echo
images in a post-processing step compensates for this SNR loss
[51]. More importantly, it is our experience that the acquisition
of multiple echoes with high bandwidth is less prone to
motion-related image artifacts compared to only one echo
acquired with lower bandwidth. The MGE sequence also
allows the quantification of R2* by fitting an exponential func-
tion to the magnitude signal in every voxel as well as sophisti-
cated echo-combination techniques may be applied to produce
phase images with improved CNR [52]. We routinely employ
such an R2*-based multi-echo phase combination in the course
of Quantitative Susceptibility Mapping (QSM), a post-
processing technique for GRE phase images that yields maps
of the tissue magnetic susceptibility distribution. QSM involves
a number of computationally relatively complex processing
steps to maintain the integrity of the phase signal and calculate
the susceptibility maps. A detailed discussion of the QSM
processing pipeline cannot be provided within this chapter
and may be found elsewhere [53–58].
4. For a single-echo protocol, set the echo time to 16 ms
(at 9.4 T). The benefit of longer echo times for visualizing
susceptibility variations is limited compared to the increasing
sensitivity to motion and field inhomogeneity artifacts. Set
readout bandwidth to the lowest possible value.
5. Conventional MGE sequences usually do not provide flow
compensation in all spatial directions [59]. However, it is our
experience that flow compensation has a relatively limited
benefit in preclinical SWI. The phase cancellation in the
absence of flow compensation increases the conspicuity of
veins on the magnitude images, and ghosting effects are mini-
mal, probably due to the high flow speed.
6. We use ParaVision 5.1 for in vivo SWI due to historical reasons

(instead of the newer version 6.0.1). However, the technique
may equally well be employed using newer versions of ParaVi-
sion. In particular, ParaVision 6.0.1 with the SWI Extension
Package can produce filtered phase images and perform the
common magnitude-phase combination.
7. All metal parts, even if they are not ferromagnetic (i.e., not
noticeably attracted in a magnetic field gradient), must be
removed before the experiment to avoid image artifacts due
to field perturbations.
8. A multichannel receive coil requires multiple1H receiver
channels.
9. Improve conventional needle ear bars by placing a bubble of
hardened glue on their tip. The larger surface reduces the
pressure on the skull, resulting in a more stable anesthesia,
and avoids accidental rupture of the eardrums.
10. Coils should generally be chosen such that the sensitivity is
high in the region of interest. SWI does not require a special
coil configuration and may also be performed with a transceiver
volume coil, as the simplest realization. Multichannel surface-
coils generally provide higher SNR in the brain compared to
volume coils, but at the detriment of spatial signal inhomoge-
neity associated with a decreasing signal-to-noise ratio (sensi-
tivity) with increasing distance from the coil (both magnitude
and phase). However, depending on the coil and the study
objective, the image inhomogeneity may not hamper the inter-
pretation of imaging findings.
11. Fixation of the mouse head using a conventional three-point
fixation with needle-like ear bars often places too much pres-
sure onto the skull, compressing nerves and leading to unstable
anesthesia. Our lateral fixation part has an L shape. It consists
of a square (approx. 8 8 mm) or rectangular (approx.
8 15 mm) side that is padded with few layers of Nexcare
cushions tape (3M). This side will be in contact with animal’s
jaws. The other side of the lateral fixation is engineered to fit in
the animal cradle groves designated for the regular ear fixation
parts, and it is held down by the same plastic screws. The lateral
head fixation parts are attached to the screws of the original
vendor parts on the animal sled. The lateral fixation parts were
designed to prevent head motion without exposing the head to
excessive pressure by flexibly adjusting to the mouse head and
neck anatomy. Figure 4 shows the 3D model (designed with
FreeCAD, www.freecadweb.org), printed with a fused deposi-
tion modeling printer and padded with several layers of Nex-
care cushions tape. The vertical head fixation consists of a
U-shaped piece of foam that is placed on the bottom of the
SWI 221
Fig. 4 Illustration of the setup for mice. Panel (a) shows the 3D models of the custom side fixation bars. Panel
(b) illustrates the setup with SpO2 sensor (straight tail black arrow), respiration sensor (white arrow), and
temperature probe (circle-ended arrow). Panel (c) shows the same setup with a mouse. The arrows mark the
custom 3D printed fixation bars shown in panel (a). Panel (d) illustrates the vertical fixation using a soft tape
cushion
cradle near the nose cone and can be considered the fourth
fixation point. The animal’s head is cushioned on the sides and
on the bottom by this foam providing support below animal’s
neck without obstructing its respiration. The sides of this
U-shaped foam add even more protection when the lateral
fixation parts squeeze the animal’s head. The vertical fixation
is particularly important for smaller mice for which the stan-
dard animal sled does not provide sufficient vertical support.
12. The BOLD signal depends on cerebral metabolic oxygen con-
sumption, blood flow, and oxygenation level of the arterial
blood. Hence, the inhaled gas mixture and, in particular, the
anesthetic agent has a considerable impact on the vessel con-
trast in preclinical SWI [31]. Shen et al. [60] recommended
intraperitoneal injection of a mixture of ketamine, xylazine,
and atropine. Inhalation anesthesia allows for faster induction
and quicker recovery than injectable agents, and the dose is
easily regulated. However, Isoflurane increases the cerebral
blood flow (CBF) resulting in a reduced deoxygenation level
of venous blood and, hence, reduced SWI vein contrast. A
reduced visibility of veins is disadvantageous for venography,
but may be considered as beneficial for the detection of hemor-
rhages, iron storage, and calcium. Halothane reduces the oxy-
gen consumption without increased blood flow, but has poorer
analgesia and muscle relaxation qualities and is associated with
toxic metabolic products. The use of agents for preclinical
BOLD fMRI experiments (alpha-Chloralose, Medetomidine,

Etomidate; urethane for terminal procedures) may produce
better visualization of veins. Some researchers use Isoflurane
with an oxygen-nitrogen carrier gas, instead of pure oxygen or
oxygen-air. Nitrous Oxide has mild analgesic properties, but is
associated with human health hazards. If the objective of the
study is to compare the visualization of veins or detect changes
thereof, a robust reproducible anesthesia mechanism needs to
be chosen.
13. A Biological Safety Cabinet is not sufficient because HEPA
filters do not scavenge or absorb chemical vapors so that
waste isoflurane is simply recirculated in the room. If a fume
hood is not available, it is recommended to utilize a laminar
flow hood/negative pressure system with activated carbon
filtered exhaust chamber or an active charcoal filter may be
attached to the outlet of the induction chamber.
14. Arbitrary tissue specimens may be scanned with this setup,
including fixed human tissue samples. Perfusion of rodents
with Gadolinium is a procedure to reduce the T1-time of the
tissue (more efficient MRI acquisition) and increase neuronal
contrast [61].
15. Galden HT80 is a perfluoropolyether with a magnetic suscep-
tibility similar to biological tissues and a low viscosity. The
absence of protons in this liquid implies that no signal is
obtained from outside the specimen with1H-MRI, which
allows a more beneficial receiver gain setting and a smaller
FOV. The similar susceptibility reduces field inhomogeneities
compared to placing the sample in air, where the high suscep-
tibility difference between air-tissue creates strong field gradi-
ents inside the brain. The low viscosity allows the fluid to
penetrate all cavities of the specimen, expelling air and, thus,
preventing susceptibility artifacts around air bubbles.
16. Other perfluoropolyether may be used as well, such as Fomblin
(which has a higher viscosity than Galden HT80) [62–65],
Galden HS240 [66], Galden D40 [67], and Galden SV80
[68, 69].
17. Any coil may be used for ex vivo experiments. The reduced
sensitivity compared to the cryogenic coil may be compensated
for by longer scan times (e.g., a greater number of averages).
18. A transceiver cryo-coil has the disadvantage of a rather high
signal inhomogeneity, implying reduced SNR with increasing
distance from the coil. However, these intensity inhomogene-
ities may be mitigated using bias field correction algorithms in
the post-processing step.
19. A workstation with lower computational performance may be
sufficient to process the images. The limiting factor when
SWI 223
Fig. 5 Illustration of the ex vivo setup. The specimen is fixed within a PFPE-filled tube using soft foam. The
sagittal and horizontal perspectives illustrate the orientation of the brain in the tube. A similar set up was used
to acquire the images in Fig. 2
working with high-resolution multi-echo data acquired with

phased-array coils is usually the available RAM. However,
physical memory limitations may partially be addressed by
slice-wise processing of the imaging data.
20. Bruker provides a MATLAB package for handling ParaVision
data (pvmatlab) to its customers on request.
21. Place the brain on top of the foam and slide both into the
Falcon tube up to, but not into the conical part of the tube.
Ensure that the tissue is pressed only gently against the internal
tube wall and does not deform. Reduce the amount of foam if
required. The brain should be touching the tube’s wall, and the
desired slice orientation on the images should be horizontal if
the brain is facing up. See Fig. 5 for an example of the coronal
slice imaging setup.
22. The location and orientation of the brain in the tube are only
relevant if a surface coil is used with an inhomogeneous sensi-
tivity profile.
23. The viscosity of Galden HT80 is so low that it flows out of a
syringe’s needle. Pour the liquid from the storage container
into the Falcon tube, ensuring that spills will be caught and can
be transferred back into the storage container.
24. Do not exceed the maximum temperature specified for any of
the components of the heating circuit.
25. The clamps for the hind foot or the tail are very bulky in an
already small space, and the reading is not as robust and con-
sistent as when you attach the sensors directly at the base of
the tail.
26. Turn on the laptop display software and verify that all instru-
ments are reading. If they do not, check that all the tools are
powered, and fiber connections (red laser light visible?) are
intact. The error messages will guide you through what’s
wrong.
27. Ensure that the slices of the scout are prescribed without spatial
offsets, i.e., they are centered at the isocenter. With this slice
location, it is simple to confirm that the brain of the animal is
centered at the isocenter of the magnet.
28. A sufficient depth of anesthesia can be tested by placing a slight
pressure on the animal’s hind limb. The pedal withdrawal reflex
is absent in deeply anesthetized animals. However, never let the
animal get too deep because this can result in instability later
during the anesthesia!
29. Some strains/models (in particular smaller animals) cool off or
recover from anesthesia very quickly. When working with such
animals, the anesthesia chamber may be transported into the
magnet room and placed right next to the heated animal cradle
before opening it and transferring the animal onto the cradle.
However, this procedure exposes the operator to a higher level
of isoflurane compared to opening the chamber under the
fume hood. Furthermore, the anesthesia chamber must not
contain any ferromagnetic parts.
30. If the eyes popped up you squeezed too much!!
31. You may need to move the nose cone forward or backward to
match the ear bars with the ears’ location.
32. The fixation must not cause pain to the animal, because this
could wake the animal up or render the anesthesia unstable.
Avoid hard, strong, body-spasm-type, gasping respirations,
which indicates respiration difficulty. The respiration should
appear shallow enough that the whole body does not move
with each breath (only the abdominal area should pulse with
each breath). If you see these gasps, you’ve probably packed
the animal in too tight or isoflurane level is too high. Loosen
up the animal and/or decrease the isoflurane concentration in
small increments. Wait for the respiration to become smoother
yet stable and uniform. Do not try to save time during the
positioning. This is one of the most important steps of the
experiment!
33. Do not push if you feel it does not slide in. Rather not use the
probe instead of rupturing the rectum.
34. Motion is the most critical factor for image quality in preclini-
cal SWI, in particular when 3D sequences are employed. Later
echoes are particularly motion sensitive, which is why we rec-
ommend the use of a motion-sensing multi-echo scan with
long echo times. Unstable animals with grasping respiration
show particularly high level of artifacts that can render images
at later echo times unusable. Neither respiration nor cardiac
triggering significantly reduces the motion-related artifacts in
SWI at later echo times. The artifacts are probably related to
changes in the susceptibility-related magnetic field (phase) due
SWI 225
to abdominal motion outside the FOV, e.g., of the lungs,

rather than a rigid movement of the head. There is no need
to run the main SWI scan if the later echoes in the motion-
sensing scan have an acceptable quality. If those images are
degraded, every effort should be made to improve anesthesia
and the positioning of the animal.
35. A re-calibration of the gradient amplifiers may be necessary if
the magnitude image signal of a multi-echo sequence with
flybacks (mono-polar readout) decays so quick over the echo
number that the later echoes do not show any image contrast
and consist only of noise, even in a homogeneous spherical
phantom. In this case, the k-space acquisition window drifts
with every flyback, eventually missing the center of the k-space.
If only one or two gradient amplifiers are affected, the artifact
may be reduced by switching the read direction (try all three
directions).
36. Periodic stripes or point-like artifacts with very high image
intensity relative to the tissue intensity when using the cryo-
probe may be caused by eye lube on the coil head. Carefully
clean the coil head and continue.
37. Some coils are pre-tuned and do not need to be tuned and
matched.
38. Field inhomogeneities may lead to signal loss or reduce SNR in
later echoes. Field map-based local higher-order shimming to
the brain theoretically improves the image quality. However,
while this step is included in our routine protocols, our experi-
ence is that linear global shimming is usually sufficient for SWI.
Inspect the raw phase images (without filtering) of later echoes
for semi-parallel wraps, which indicate a slowly varying field
throughout the brain that could be minimized by higher-order
shimming.
39. To maximize the dephasing-related signal decay around veins
at clinical field strengths, SWI employs axially prescribed
sequences with anisotropic voxels that have a higher voxel
edge length in the axial direction (aspect ratio of 2–4 [40]).
Our experience is that the slab orientation does not have to be
strictly axial at 9.4 T because intravascular contrast mechanisms
prevail at higher field strengths.
40. Echo times exceeding 30 ms at 9.4 T often show substantial
motion-related artifacts in phase-encode direction, in both 3D
and 2D applications. For phase imaging, the optimal CNR is
obtained when the echo time equals T2* of the tissue
[70]. With a T2* between 25 and 40 ms in rat brain, it would
appear that the optimal echo time is around 30 ms. However,
in practice, the CNR benefit from choosing echo times higher
than 20 ms is often outweighed by the occurrence of localized
Fig. 6 Illustration of the temporal phase evolution (top row, from left to right), associated artifacts, and the
benefit of a magnitude-weighted combination (bottom row; from left to right more echoes used). The middle
row shows magnitude images corresponding to the phase images. The circle-ended arrow marks the field
perturbation due to a hemorrhage. The straight-ended arrows indicate phase artifacts. In the top row, the
arrow points at an intrinsic wrapping artifact due to a high susceptibility difference. Such artifacts are difficult
to eliminate using filtering and propagate into the combined phase images (bottom row), requiring an
appropriate cut-off for the echo time. The straight-ended arrow in the bottom row points at magnetic field
inhomogeneities close to the surface of the brain. Due to their spatial localization, such inhomogeneities
cannot be mitigated by field shimming and propagate into typical magnitude-phase SWI composite images,
complicating the interpretation of signal changes in the cortex. The box-ended arrow points at the improved
gray-white matter contrast and reduced noise level in the magnitude-weighted phase average image
wrapping artifacts in the brain that result from the brain tissue
susceptibility variation itself. Since these fields are not slowly
varying, such wrapping artifacts cannot be suppressed easily
(Fig. 6).
41. For QSM, relatively isotropic voxels and whole brain coverage
are beneficial. Hence, our 3D-MGE standard SWI protocol
covers the entire brain with a nearly isotropic resolution. If
smaller veins need to be detected, a 2D protocol with partial
coverage should be utilized and measurement time should be
invested into higher in-plane resolution instead. Depending on
the goal of the study, the in-slice resolution should be isotropic
to ensure that vessel visibility is independent of the orientation
of vessels. The resolution in read direction may be increased
substantially without additional measurement time by a simul-
taneous increase of the readout bandwidth (however, at the
expense of SNR). Measurement time may be decreased by
reducing the resolution in phase-encode direction.
42. We use ParaVision 6.0.1 for ex vivo MR microscopy measure-
ments because we observed inter-slice intensity variations on
magnitude images when using large matrices in PV5.1.
SWI 227
43. Perfusion with Gadolinium drastically increases the tissue’s

R2* constant. The rapid relaxation renders the acquisition of
multiple echoes difficult with our gradient system, because
when imaging with high spatial resolution the de- and rephas-
ing gradient pulses require a considerable time. We hence
employ a single echo sequence for perfused tissue specimens.
For fixed tissues without Gadolinium perfusion, multi-echo
sequences similar to those used in vivo are possible.
44. If a short acquisition time is required, employ a 2D sequence
that covers only part of the brain. With partial brain coverage
the 3D sequence often suffers from a higher artifact level
compared to the 2D acquisition mode, in particular close to
the edges of the imaging slab. Generally, 2D sequences seem to
perform better in small-animal SWI than 3D sequences. At the
same measurement time and with otherwise comparable para-
meters, the latter often suffer from more severe magnitude
signal inhomogeneity and lower vein contrast. This may be
explained by a higher sensitivity of 3D sequences to motion.
However, the slice thickness is limited with 2D sequences and
whole brain coverage becomes inefficient due to very long TRs.
When using a 2D sequence we employ ventral-dorsal as read
direction, which allows faster scanning by reducing the FOV in
left-right direction (phase-encode direction). Define the FOV
in read-direction such that all signal-generating tissue is cov-
ered. Set the number of slices and the slice gap to achieve the
required spatial coverage of the brain. This may require increas-
ing the repetition time temporarily. A reasonable slice thickness
is 500 μm. Use Interlaced slice ordering to avoid cross-talk
effects. Set TR to minimum possible value (see Note 46) and
adjust the tip angle to the Ernst Angle (see Note 47). Reduce
the excitation pulse length to the minimum that is possible
with the chosen slice thickness.
45. With a primary interest in the brain, some left-right fold-over
of muscle tissue may be acceptable as long as the fold-over does
not affect the brain.
46. Set TR to minimum possible value. While measurements with-
out averaging can yield a reasonable quality in stable animals,
motion (respiration, cardiac) often results in artifacts in phase-
encoding direction that degrade image detail and homogene-
ity. Averaging usually mitigates these artifacts considerably. It is
hence recommended to always reduce TR to the minimum
value instead and increase the number of averages such that
the measurement time hits the time constraints.
47. Set the tip angle to the Ernst Angle, which is the tip angle
that provides
the maximum
magnitude SNR for a given TR:
=
TR
α ¼ cos 1 e T 1 . The T1 time of the brain may be
measured with a T1-mapping sequence (e.g., T1map_RARE in

ParaVision). A reasonable estimate for in vivo rat brain tissue at
9.4 T is T1 ¼ 1700 ms.
48. The inter-echo time determines the maximum number of ech-
oes for a given TR. Increase the readout bandwidth to reduce
the inter-echo time. However, check that each individual echo
has sufficient signal. Noise in phase images increases exponen-
tially for magnitude SNR below 3 [71].
49. With a surface transmit coil, the nominal tip angle is obtained
only in a thin slice, with higher tip angles closer to the coil and
lower angles farther away from the coil. If the tip angle
becomes too high, a drastic decrease in signal intensity is
observed. Place the calibration slab as far away from the coil
as possible without saturation in the brain. Try to meet the
Ernst angle condition as close to the center of the brain (or the
region of interest within the brain) as possible.
50. The MGE sequence with short TR has a relatively high gradi-
ent duty cycle. Confirm that the duty cycle does not exceed
your system’s limitations with a duty cycle simulation before
running the sequence for the first time. If the maximum duty
cycle is exceeded, either increase TR or increase the inter-echo
interval.
51. Feedback systems are commercially available that automatically
adjust the heating temperature according to the measured
animal temperature.
52. Do not leave specimen in the perfluoropolyether longer than
required, because long time exposure may affect the MRI
properties of the specimen.
53. The ideal respiration rate under anesthesia is between 100 and
140/min. However, the high noise level of (multi-echo) GRE
imaging usually requires a deeper anesthesia with a respiration
rate between 50 and 90/min to avoid motion. Stimulation of
muscle nerves associated with rapid gradient switching may be
another reason why a deeper anesthesia is required to keep the
animal still. The respiration rate should be as high as possible
and as low as necessary. The optimal range also depends on the
strain and model and should be determined in pilot experi-
ments before the start of the study. If anesthesia is too light, the
animal may move head and limbs during the acquisition and
even free itself from the stereotactic fixation, requiring repeti-
tion of the experiment. If the respiration rate is too low, the
animal starts grasping for air, causing significant contractions
of the whole abdomen that result in severe artifacts on SWI
(probably due to associated macroscopic magnetic field
changes). Respiration rate below 50/min often results in
unstable anesthesia and may even lead to sudden death. If a
SWI 229
low respiration rate is needed to keep the animal still, check the
animal’s temperature and improve fixation.
54. ParaVision 5.1 acquires the specified “matrix dimension” in
read direction. If 33% echo position is specified, this effectively
results in a shift of the k-space readout window and must not
be confused with the “Partial Fourier” feature on other plat-
forms, that reduces the effectively acquired matrix size and fills
the data matrix elements that were not acquired with zeros. A
set echo position below 50% needs to be accounted for when
working with the raw data, e.g., by placing the raw data matrix
in a larger array of zeros with a centered k-space origin in that
matrix. In addition, if isotropic nominal spatial resolution is
desired, the acquisition matrix size specified in ParaVision must
be decreased accordingly.
55. In MATLAB, use the command padarray to add zeros to the
first, second, and third dimension of the k-space data. Make
sure that the parity of the matrix dimensions remains the same,
i.e., increase the matrix size from 256 to 512, but not from
256 to 515. Do not add zeros to the echo- or coil-dimensions.
Apply an 3D Fast Fourier Transform to the k-space data
(data ¼ fftshift(fft(fft(fft(ifftshift(data),[],1),[],2),[],3)); in
matlab) to obtain the image space data.
56. Square the magnitude signal in every matrix element, sum over
all coil elements and take the square root of the result
(in MATLAB: sqrt(sum(abs(data).^2,5), if the coil elements
are stored in the 5th dimension of the data matrix). If a single-
channel receive coil is used (compared to a multichannel coil),
it is sufficient to directly calculate the absolute value (abs(data))
after the inverse Fourier transform to obtain the magnitude
images.
57. This processing step may also be performed on images recon-
structed by ParaVision. The image may be found in the sub-
folder pdata/.
58. For every pixel in a certain slice jS, determine the minimum
intensity value of the equivalent voxel in the adjacent N slices
(In MATLAB: dataMIP(:,:,jS) ¼ min(data(:,:,jS-N:jS+N),[],3).
The calculation of the mIP may be performed for all slices
using a for-loop over jS. However, make sure that the mIP is
calculated over less slices at the boundaries of the imaging slab.
59. For NIfTI export from MATLAB use the Matlab package by
Jimmy Shen (https://www.mathworks.com/matlabcentral/
fileexchange/8797-tools-for-nifti-and-analyze-image).
60. Several methods can be employed to combine multichannel
phase images [72] and remove background phase contribu-
tions [73]. A simple homodyne phase combination of the
multichannel phase images may be performed by creating a
2D Hanning filter with a kernel size of, e.g., 64 64 voxels.

Apply the filter slice-wise to the complex-valued 3D image data
of every echo and channel. Weigh the phase of the homodyne
filtered complex-valued images in each voxel with the original
magnitude in that voxel, sum the weighted phase from differ-
ent channels in each voxel, and divide the resulting value in
each voxel by the sum of the magnitude values of that voxel in
the channels. In MATLAB, this may be achieved by combined-
PhaseCh ¼ sum(abs(data)).*angle(dataHom),5)./ sum(abs
(data),5) (see Note 61).
61. The high-pass filter may not properly eliminate phase wrapping
if the large-scale magnetic field inhomogeneities are too
strong. In this case, the filtered phase may suffer from residual
wrap artifacts (Fig. 6).
62. In the case of 2D imaging or if the MRI system permits recon-
struction of multi-echo images, filtered phase images may also
directly be obtained from within ParaVision 6.0.1 with the SWI
extension package. Without SWI license, unfiltered phase images
may only be obtained directly from ParaVision when a single-
channel receiver coil is used. To this end, activate “Prototype
Reconstruction” in the RECO Parameters tab of the Processing
Platform and set the “Output Image Type” to Phase with
“Output Mapping” to “User Scale”. Set Output Slope and
Offset to 1 and 0, respectively. With multichannel coils, a proper
phasing of the individual coil phase images is required [72],
which is not supported by current versions of ParaVision.
63. A simple multi-echo combination may be achieved by multi-
plying each individual filtered phase image with the
corresponding sum-of-squares magnitude image, summing the
various echoes and dividing the result by the sum of the sum-of-
squares magnitude at the different echo times. In MATLAB, this
may be achieved by combinedPhaseChTe ¼ sum(magnSos
(:,:,:,1:nE).* combinedPhaseCh (:,:,:,1:nE),4)./ sum(magnSos
(:,:,:,1:nE),4), where nE ¼ 6 is the last echo to be used for the
combined phase (see Note 40). For later echo times the homo-
dyne filter is often not powerful enough to eliminate phase
wraps completely. The value of nE, which determines the latest
echo to be used for the combination, needs to be adjusted such
that echoes with residual phase wraps are excluded from the
combination (Fig. 6).
Acknowledgements
We are grateful to Drs. David Poulsen (Department of Neurosur-

gery, University at Buffalo) and Claire Modica (Buffalo Neuroim-
aging Analysis Center, Department of Neurology, University at
SWI 231
Buffalo) for support with the ex vivo experiments. Research

reported in this publication was funded by the National Center
for Advancing Translational Sciences of the National Institutes of
Health under award Number UL1TR001412. The content is
solely the responsibility of the authors and does not necessarily
represent the official views of the NIH.
References
1. Haacke EM, Xu Y, Cheng YC, Reichenbach JR hemorrhage and calcium on susceptibility-

(2004) Susceptibility weighted imaging (SWI). weighted images. Neurographics 6
Magn Reson Med 52(3):612–618. https:// (3):123–126
doi.org/10.1002/mrm.20198 10. Ong BC, Stuckey SL (2010) Susceptibility
2. Reichenbach JR, Barth M, Haacke EM, weighted imaging: a pictorial review. J Med
Klarhofer M, Kaiser WA, Moser E (2000) Imaging Radiat Oncol 54(5):435–449.
High-resolution MR venography at 3.0 Tesla. https://doi.org/10.1111/j.1754-9485.2010.
J Comput Assist Tomogr 24(6):949–957 02208.x
3. Reichenbach JR, Haacke EM (2001) High- 11. Chen X, Zeng C, Luo T, Ouyang Y, Lv F,
resolution BOLD venographic imaging: a win- Rumzan R, Wang Z, Li Q, Wang J, Hou H,
dow into brain function. NMR Biomed 14 Huang F, Li Y (2012) Iron deposition of the
(7–8):453–467 deep grey matter in patients with multiple scle-
4. Reichenbach JR, Venkatesan R, Schillinger DJ, rosis and neuromyelitis optica: a control quan-
Kido DK, Haacke EM (1997) Small vessels in titative study by 3D-enhanced susceptibility-
the human brain: MR venography with deox- weighted angiography (ESWAN). Eur J Radiol
yhemoglobin as an intrinsic contrast agent. 81(4):e633–e639. https://doi.org/10.1016/
Radiology 204(1):272–277. https://doi.org/ j.ejrad.2012.01.003
10.1148/radiology.204.1.9205259 12. Du S, Sah SK, Zeng C, Wang J, Liu Y,
5. Ahmed BA, Kumar NA, Vijayan D, Magudees- Xiong H, Li Y (2015) Iron deposition in the
waran PK, Ganesan P (2016) Magnetic reso- gray matter in patients with relapse-remitting
nance imaging susceptibility- weighted multiple sclerosis: a longitudinal study using
imaging is more reliable to detect hemorrhage three-dimensional (3D)-enhanced T2*-
and calcification than magnetic resonance weighted angiography (ESWAN). Eur J Radiol
imaging T2*-weighted gradient echo in brain 84(7):1325–1332. https://doi.org/10.1016/
imaging. Int J Sci Study 4(3):203–205. 10. j.ejrad.2015.04.013
17354/ijss/2016/352 13. Hagemeier J, Weinstock-Guttman B,
6. Barbosa JH, Santos AC, Salmon CE (2015) Bergsland N, Heininen-Brown M, Carl E,
Susceptibility weighted imaging: differentiat- Kennedy C, Magnano C, Hojnacki D, Dwyer
ing between calcification and hemosiderin. MG, Zivadinov R (2012) Iron deposition on
Radiol Bras 48(2):93–100. https://doi.org/ SWI-filtered phase in the subcortical deep gray
10.1590/0100-3984.2014.0010 matter of patients with clinically isolated syn-
7. Berberat J, Grobholz R, Boxheimer L, drome may precede structure-specific atrophy.
Rogers S, Remonda L, Roelcke U (2014) Dif- AJNR Am J Neuroradiol 33(8):1596–1601.
ferentiation between calcification and hemor- https://doi.org/10.3174/ajnr.A3030
rhage in brain tumors using susceptibility- 14. Ning N, Zhang L, Gao J, Zhang Y, Ren Z,
weighted imaging: a pilot study. AJR Am J Niu G, Dai Y, EX W, Guo Y, Yang J (2014)
Roentgenol 202(4):847–850. https://doi. Assessment of iron deposition and white matter
org/10.2214/AJR.13.10745 maturation in infant brains by using enhanced
8. Deistung A, Mentzel HJ, Rauscher A, T2 star weighted angiography (ESWAN): R2*
Witoszynskyj S, Kaiser WA, Reichenbach JR versus phase values. PLoS One 9(2):e89888.
(2006) Demonstration of paramagnetic and https://doi.org/10.1371/journal.pone.
diamagnetic cerebral lesions by using suscepti- 0089888
bility weighted phase imaging (SWI). Z Med 15. Wang D, Li WB, Wei XE, Li YH, Dai YM
Phys 16(4):261–267 (2012) An investigation of age-related iron
9. Li C, Imbesi S, Lee R, Haacke E, Chen J deposition using susceptibility weighted imag-
(2016) Potential pitfalls when differentiating ing. PLoS One 7(11):e50706. https://doi.
org/10.1371/journal.pone.0050706
16. Wang D, Zhu D, Wei XE, Li YH, Li WB (2013) autoimmune encephalomyelitis mouse model
Using susceptibility-weighted images to quan- of multiple sclerosis using susceptibility MRI
tify iron deposition differences in amnestic mild and hyperoxygenation. PLoS One 10(5):
cognitive impairment and Alzheimer’s disease. e0127033. https://doi.org/10.1371/journal.
Neurol India 61(1):26–34. https://doi.org/ pone.0127033
10.4103/0028-3886.107924 26. Shen Y, Kou Z, Kreipke CW, Petrov T, Hu J,
17. Xu X, Wang Q, Zhang M (2008) Age, gender, Haacke EM (2007) In vivo measurement of
and hemispheric differences in iron deposition tissue damage, oxygen saturation changes and
in the human brain: an in vivo MRI study. blood flow changes after experimental trau-
NeuroImage 40(1):35–42. https://doi.org/ matic brain injury in rats using susceptibility
10.1016/j.neuroimage.2007.11.017 weighted imaging. Magn Reson Imaging 25
18. Yu J, Qi F, Wang N, Gao P, Dai S, Lu Y, Su Q, (2):219–227. https://doi.org/10.1016/j.
Du Y, Che F (2014) Increased iron level in mri.2006.09.018
motor cortex of amyotrophic lateral sclerosis 27. Verma SK, Kan EM, Lu J, Ng KC, Ling EA,
patients: an in vivo MR study. Amyotroph Lat- Seramani S, Kn BP, Wong YC, Tan MH, Velan
eral Scler Frontotemporal Degener 15 SS (2015) Multi-echo susceptibility-weighted
(5–6):357–361. https://doi.org/10.3109/ imaging and histology of open-field blast-
21678421.2014.906618 induced traumatic brain injury in a rat model.
19. Ogawa S, Lee TM, Nayak AS, Glynn P (1990) NMR Biomed 28(9):1069–1077. https://doi.
Oxygenation-sensitive contrast in magnetic org/10.1002/nbm.3351
resonance image of rodent brain at high mag- 28. Yang SH, Lin J, Lu F, Dai YY, Han ZH, CX F,
netic fields. Magn Reson Med 14(1):68–78 FL H, HC G (2016) Contrast-enhanced sus-
20. Krishnan AS, Lansley JA, Jager HR, Mankad K ceptibility weighted imaging with ultrasmall
(2015) New vistas in clinical practice: superparamagnetic iron oxide improves the
susceptibility-weighted imaging. Quant Imag- detection of tumor vascularity in a hepatocellu-
ing Med Surg 5(3):448–452. https://doi.org/ lar carcinoma nude mouse model. J Magn
10.3978/j.issn.2223-4292.2015.03.03 Reson Imaging 44(2):288–295. https://doi.
21. Liu S, Buch S, Chen Y, Choi HS, Dai Y, org/10.1002/jmri.25167
Habib C, Hu J, Jung JY, Luo Y, Utriainen D, 29. Kim SG, Park SH (2011) High-resolution
Wang M, Wu D, Xia S, Haacke EM (2017) venographic BOLD MRI of animal brain at
Susceptibility-weighted imaging: current status 9.4 T: implications for BOLD fMRI. Suscepti-
and future directions. NMR Biomed 30(4): bility weighted imaging in MRI: basic concepts
e3552. https://doi.org/10.1002/nbm.3552 and clinical applications. Wiley-Blackwell,
22. Bir C, Vandevord P, Shen Y, Raza W, Haacke Hoboken, pp 637–647
EM (2012) Effects of variable blast pressures 30. Ogawa S, Lee TM (1990) Magnetic resonance
on blood flow and oxygen saturation in rat imaging of blood vessels at high fields: in vivo
brain as evidenced using MRI. Magn Reson and in vitro measurements and image simula-
Imaging 30(4):527–534. https://doi.org/10. tion. Magn Reson Med 16(1):9–18
1016/j.mri.2011.12.003 31. Ogawa S, Lee TM, Kay AR, Tank DW (1990)
23. Blasiak B, Landry J, Tyson R, Sharp J, Iqbal U, Brain magnetic resonance imaging with con-
Abulrob A, Rushforth D, Matyas J, Ponjevic D, trast dependent on blood oxygenation. Proc
Sutherland GR, Wolfsberger S, Tomanek B Natl Acad Sci U S A 87(24):9868–9872
(2014) Molecular susceptibility weighted 32. Park SH, Masamoto K, Hendrich K, Kanno I,
imaging of the glioma rim in a mouse model. Kim SG (2008) Imaging brain vasculature with
J Neurosci Methods 226:132–138. https:// BOLD microscopy: MR detection limits deter-
doi.org/10.1016/j.jneumeth.2014.01.034 mined by in vivo two-photon microscopy.
24. Nathoo N, Agrawal S, Wu Y, Haylock-Jacobs S, Magn Reson Med 59(4):855–865. https://
Yong VW, Foniok T, Barnes S, Obenaus A, doi.org/10.1002/mrm.21573
Dunn JF (2013) Susceptibility-weighted imag- 33. Harrison PM, Arosio P (1996) The ferritins:
ing in the experimental autoimmune encepha- molecular properties, iron storage function and
lomyelitis model of multiple sclerosis indicates cellular regulation. Biochim Biophys Acta
elevated deoxyhemoglobin, iron deposition 1275(3):161–203
and demyelination. Mult Scler 19 34. Buch S, Liu S, Ye Y, Cheng YC, Neelavalli J,
(6):721–731. https://doi.org/10.1177/ Haacke EM (2015) Susceptibility mapping of
1352458512460602 air, bone, and calcium in the head. Magn Reson
25. Nathoo N, Rogers JA, Yong VW, Dunn JF Med 73(6):2185–2194. https://doi.org/10.
(2015) Detecting deoxyhemoglobin in spinal 1002/mrm.25350
cord vasculature of the experimental
SWI 233
35. Chen W, Zhu W, Kovanlikaya I, Kovanlikaya A, (2013) Iron deposition in the precentral grey
Liu T, Wang S, Salustri C, Wang Y (2014) matter in patients with multiple sclerosis: a
Intracranial calcifications and hemorrhages: quantitative study using susceptibility-
characterization with quantitative susceptibility weighted imaging. Eur J Radiol 82(2):
mapping. Radiology 270(2):496–505. e95–e99. https://doi.org/10.1016/j.ejrad.
https://doi.org/10.1148/radiol.13122640 2012.09.006
36. Deistung A, Schweser F, Wiestler B, Abello M, 44. Wang D, Li YY, Luo JH, Li YH (2014)
Roethke M, Sahm F, Wick W, Nagel AM, Age-related iron deposition in the basal ganglia
Heiland S, Schlemmer HP, Bendszus M, of controls and Alzheimer disease patients
Reichenbach JR, Radbruch A (2013) Quanti- quantified using susceptibility weighted imag-
tative susceptibility mapping differentiates ing. Arch Gerontol Geriatr 59(2):439–449.
between blood depositions and calcifications https://doi.org/10.1016/j.archger.2014.04.
in patients with glioblastoma. PLoS One 8(3): 002
e57924. https://doi.org/10.1371/journal. 45. Hagemeier J, Weinstock-Guttman B,
pone.0057924 Heininen-Brown M, Poloni GU,
37. Schweser F, Deistung A, Lehr BW, Reichen- Bergsland N, Schirda C, Magnano CR,
bach JR (2010) Differentiation between dia- Kennedy C, Carl E, Dwyer MG, Minagar A,
magnetic and paramagnetic cerebral lesions Zivadinov R (2013) Gray matter SWI-filtered
based on magnetic susceptibility mapping. phase and atrophy are linked to disability in
Med Phys 37(10):5165–5178. https://doi. MS. Front Biosci (Elite Ed) 5:525–532
org/10.1118/1.3481505 46. Peters AM, Brookes MJ, Hoogenraad FG,
38. Straub S, Laun FB, Emmerich J, Jobke B, Gowland PA, Francis ST, Morris PG, Bowtell
Hauswald H, Katayama S, Herfarth K, Schlem- R (2007) T2* measurements in human brain at
mer HP, Ladd ME, Ziener CH, Bonekamp D, 1.5, 3 and 7 T. Magn Reson Imaging 25
Rothke MC (2017) Potential of quantitative (6):748–753. https://doi.org/10.1016/j.
susceptibility mapping for detection of pros- mri.2007.02.014
tatic calcifications. J Magn Reson Imaging 45 47. Bourekas EC, Christoforidis GA, Abduljalil
(3):889–898. https://doi.org/10.1002/jmri. AM, Kangarlu A, Chakeres DW, Spigos DG,
25385 Robitaille PM (1999) High resolution MRI of
39. Rauscher A, Sedlacik J, Deistung A, Mentzel the deep gray nuclei at 8 Tesla. J Comput Assist
HJ, Reichenbach JR (2006) Susceptibility Tomogr 23(6):867–874
weighted imaging: data acquisition, image 48. Roemer PB, Edelstein WA, Hayes CE, Souza
reconstruction and clinical applications. Z SP, Mueller OM (1990) The NMR phased
Med Phys 16(4):240–250 array. Magn Reson Med 16(2):192–225
40. Deistung A, Rauscher A, Sedlacik J, Stadler J, 49. Tustison NJ, Avants BB, Cook PA, Zheng Y,
Witoszynskyj S, Reichenbach JR (2008) Sus- Egan A, Yushkevich PA, Gee JC (2010)
ceptibility weighted imaging at ultra high mag- N4ITK: improved N3 bias correction. IEEE
netic field strengths: theoretical considerations Trans Med Imaging 29(6):1310–1320.
and experimental results. Magn Reson Med 60 https://doi.org/10.1109/TMI.2010.
(5):1155–1168. https://doi.org/10.1002/ 2046908
mrm.21754 50. Frank LR, Crawley AP, Buxton RB (1992)
41. Duyn JH, van Gelderen P, Li TQ, de Zwart JA, Elimination of oblique flow artifacts in mag-
Koretsky AP, Fukunaga M (2007) High-field netic resonance imaging. Magn Reson Med 25
MRI of brain cortical substructure based on (2):299–307
signal phase. Proc Natl Acad Sci U S A 104 51. Denk C, Rauscher A (2010) Susceptibility
(28):11796–11801. https://doi.org/10. weighted imaging with multiple echoes. J
1073/pnas.0610821104 Magn Reson Imaging 31(1):185–191.
42. Hagemeier J, Dwyer MG, Bergsland N, https://doi.org/10.1002/jmri.21995
Schweser F, Magnano CR, Heininen-Brown M, 52. Wu B, Li W, Avram AV, Gho SM, Liu C (2012)
Ramasamy DP, Carl E, Kennedy C, Melia R, Fast and tissue-optimized mapping of magnetic
Polak P, Deistung A, Geurts JJ, Reichenbach susceptibility and T2* with multi-echo and
JR, Zivadinov R (2013) Effect of age on MRI multi-shot spirals. NeuroImage 59
phase behavior in the subcortical deep gray (1):297–305. https://doi.org/10.1016/j.
matter of healthy individuals. AJNR Am J Neu- neuroimage.2011.07.019
roradiol 34(11):2144–2151. https://doi.org/
10.3174/ajnr.A3569 53. Duyn J (2013) MR susceptibility imaging. J
Magn Reson 229:198–207. https://doi.org/
43. Rumzan R, Wang JJ, Zeng C, Chen X, Li Y, 10.1016/j.jmr.2012.11.013
Luo T, Lv F, Wang ZP, Hou H, Huang F
54. Haacke EM, Liu S, Buch S, Zheng W, Wu D, 65. van Duijn S, Nabuurs RJ, van Rooden S, Maat-
Ye Y (2015) Quantitative susceptibility Schieman ML, van Duinen SG, van Buchem
mapping: current status and future directions. MA, van der Weerd L, Natte R (2011) MRI
Magn Reson Imaging 33(1):1–25. https:// artifacts in human brain tissue after prolonged
doi.org/10.1016/j.mri.2014.09.004 formalin storage. Magn Reson Med 65
55. Liu C, Li W, Tong KA, Yeom KW, Kuzminski S (6):1750–1758. https://doi.org/10.1002/
(2015) Susceptibility-weighted imaging and mrm.22758
quantitative susceptibility mapping in the 66. Hakkarainen H, Sierra A, Mangia S,
brain. J Magn Reson Imaging 42(1):23–41. Garwood M, Michaeli S, Grohn O, Liimatai-
https://doi.org/10.1002/jmri.24768 nen T (2016) MRI relaxation in the presence of
56. Liu C, Wei H, Gong NJ, Cronin M, Dibb R, fictitious fields correlates with myelin content
Decker K (2015) Quantitative susceptibility in normal rat brain. Magn Reson Med 75
mapping: contrast mechanisms and clinical (1):161–168. https://doi.org/10.1002/
applications. Tomography 1(1):3–17. 10. mrm.25590
18383/j.tom.2015.00136 67. Meijer FJ, Goraj B, Bloem BR, Esselink RA
57. Reichenbach JR, Schweser F, Serres B, Deis- (2017) How I do it: clinical application of
tung A (2015) Quantitative susceptibility brain MRI in the diagnostic work-up of Parkin-
mapping: concepts and applications. Clin Neu- sonism. J Parkinsons Dis 7:211. https://doi.
roradiol 25(Suppl 2):225–230. https://doi. org/10.3233/JPD-150733
org/10.1007/s00062-015-0432-9 68. Birkl C, Langkammer C, Krenn H, Goessler W,
58. Wang Y, Liu T (2015) Quantitative susceptibil- Ernst C, Haybaeck J, Stollberger R, Fazekas F,
ity mapping (QSM): decoding MRI data for a Ropele S (2015) Iron mapping using the tem-
tissue magnetic biomarker. Magn Reson Med perature dependency of the magnetic suscepti-
73(1):82–101. https://doi.org/10.1002/ bility. Magn Reson Med 73(3):1282–1288.
mrm.25358 https://doi.org/10.1002/mrm.25236
59. Xu B, Liu T, Spincemaille P, Prince M, Wang Y 69. Dal-Bianco A, Grabner G, Kronnerwetter C,
(2014) Flow compensated quantitative suscep- Weber M, Hoftberger R, Berger T, Auff E,
tibility mapping for venous oxygenation imag- Leutmezer F, Trattnig S, Lassmann H,
ing. Magn Reson Med 72(2):438–445. Bagnato F, Hametner S (2017) Slow expansion
https://doi.org/10.1002/mrm.24937 of multiple sclerosis iron rim lesions: pathology
60. Shen Y, Kou Z, Haacke EM (2011) Suscepti- and 7 T magnetic resonance imaging. Acta
bility weighted imaging in rodents. Susceptibil- Neuropathol 133(1):25–42. https://doi.org/
ity weighted imaging in MRI: basic concepts 10.1007/s00401-016-1636-z
and clinical applications. Wiley-Blackwell, 70. Schweser F, Deistung A, Reichenbach JR
Hoboken, pp 649–667 (2016) Foundations of MRI phase imaging
61. Kim S, Pickup S, Hsu O, Poptani H (2009) and processing for Quantitative Susceptibility
Enhanced delineation of white matter struc- Mapping (QSM). Z Med Phys 26(1):6–34.
tures of the fixed mouse brain using https://doi.org/10.1016/j.zemedi.2015.10.
Gd-DTPA in microscopic MRI. NMR Biomed 002
22(3):303–309. https://doi.org/10.1002/ 71. Gudbjartsson H, Patz S (1995) The Rician
nbm.1324 distribution of noisy MRI data. Magn Reson
62. Aggarwal M, Mori S, Shimogori T, Med 34(6):910–914
Blackshaw S, Zhang J (2010) Three- 72. Robinson SD, Bredies K, Khabipova D,
dimensional diffusion tensor microimaging Dymerska B, Marques JP, Schweser F (2017)
for anatomical characterization of the mouse An illustrated comparison of processing meth-
brain. Magn Reson Med 64(1):249–261. ods for MR phase imaging and QSM: combin-
https://doi.org/10.1002/mrm.22426 ing array coil signals and phase unwrapping.
63. Andrews TJ, Osborne MT, Does MD (2006) NMR Biomed 30(4):e3601. https://doi.org/
Diffusion of myelin water. Magn Reson Med 10.1002/nbm.3601
56(2):381–385. https://doi.org/10.1002/ 73. Schweser F, Robinson SD, de Rochefort L,
mrm.20945 Li W, Bredies K (2017) An illustrated compari-
64. Shmueli K, van Gelderen P, Li T, Duyn J son of processing methods for phase MRI and
(2008) High resolution human brain suscepti- QSM: removal of background field contribu-
bility maps calculated from 7 Tesla MRI phase tions from sources outside the region of inter-
data. 16th Annual Meeting of the International est. NMR Biomed 30(4):e3604. https://doi.
Society for Magnetic Resonance in Medicine, org/10.1002/nbm.3604
p 642
Chapter 14
Biomedical 19F MRI Using Perfluorocarbons

Tuba G€uden-Silber, Sebastian Temme, Christoph Jacoby,
and Ulrich Flögel
Abstract
Background-free fluorine (19F) MR imaging exhibits an excellent degree of specificity, and facilitates among
others the in vivo visualization of inflammatory processes. Merging19F MR images with morphologically
matching1H MR images enables the exact anatomic localization of the observed19F signal. Biochemically
inert nanoemulsions of perfluorocarbons, which are known to be taken up by the macrophage/monocyte
system, are widely used as contrast agents for preclinical applications. Herein, the most common protocols
are described to obtain high-resolution and artifact-free19F MR images even for compounds with com-
plex19F MR spectra. In addition, we report on the utilization of perfluorocarbons with individual spectral
identities and targeting approaches to specifically visualize thrombi by19F MRI.
Key words Fluorine MRI, Perfluorocarbons, Sterol-based post-insertion, Active targeting, Chemical
shift imaging
1 Introduction
19
F MRI is a promising noninvasive imaging technique for transla-
tional approaches, as it exhibits a great degree of specificity due to
the lack of any background signal in the body. Merging of match-
ing1H and19F datasets enables the anatomic localization of the
observed19F MR signal as “hot spot” [1–3]. The19F signal intensity
is directly proportional to the number of accumulated19F nuclear
spins which enables the convenient quantification of19F MR data.
As fluorinated contrast agents with a high payload of19F nuclei,
biochemically inert perfluorocarbon nanoemulsions (PFCs) are
promising candidates [2, 4, 5]. In particular, perfluoro-15-crown-
5 ether (PFCE; Fig. 1a) with 20 magnetically equivalent19F nuclei
per molecule is one of the most widely used PFCs for preclinical
applications. However, since PFCE displays a long biological half-
life (>250 days) [6], it is not suitable for prospective clinical
Tuba G€
uden-Silber and Sebastian Temme contributed equally to this work.
235
236 €den-Silber et al.
Tuba Gu
Fig. 1 Scheme of PFC application for in vivo tracking of biomedical processes by19F MRI using passive and
active targeting. (a) Chemical structures of PFCE, PFOB, and F-44E. (b) High-pressure homogenization with
phospholipids yields stable perfluorocarbon nanoemulsions (PFCs). The fluorophil PFC core (gray sphere) is
shielded by a lipid monolayer with the polar phosphate group (pink spheres) directed to the hydrophilic phase.
Preformed PFCs can be functionalized with targeting ligands (green spheres) through insertion of a cholesterol
anchor (red) into the lipid layer (SPIT, see text for details). After intravenous injection of PFCs (c), neat PFCs are
phagocytized by macrophages and monocytes and carried to inflammatory foci (d, left), whereas functiona-
lized PFCs can be directed to specific targets independent of phagocytosis (d, right, here thrombi)
approaches. Alternatives are perfluorooctyl bromide (perflubron,

PFOB) and trans-bis(perfluorobutyl) ethylene (F-44E) (Fig. 1a)
with much shorter biological half-lives of about 12 days (PFOB)
and 28 days (F-44E), respectively [6]. After intravenous injection
of the PFCs, the particles are phagocytized by macrophages or
19
F MRI 237
monocytes and, hence, “transported” into inflamed areas (Fig. 1d

left). For active targeting of specific surface epitopes, PFCs can be
functionalized with distinct binding ligands (Fig. 1b). Examples for
passive and active targeting approaches using neat and specifically
targeted PFCs are given herein. Of note, the utilization of different
PFCs requires specific imaging techniques in order to receive arti-
fact-free19F MR images. The single resonance19F MR spectrum of
PFCE allows the application of conventional procedures, but
PFOB and F-44E exhibit complex multipeak MR spectra, where
standard MR sequences would lead to chemical shift artifacts in the
acquired images. Thus, for PFOB and F-44E mainly chemical shift
imaging (CSI) and sequences with narrow spectral bandwidths are
required for artifact-free imaging, which will be described in detail
in the following sections (Fig. 1).
2 Materials
2.1 MR Equipment 1. MR investigations described herein were performed using a

vertical Bruker AVANCEIII 9.4 T wide-bore (89 mm) NMR
spectrometer operating at frequencies of 400.21 MHz for1H
and 376.54 MHz for19F measurements using Bruker micro-
imaging units (Micro2.5 and Mini0.5; Bruker, Rheinstetten,
Germany).
2. Data are acquired with the use of actively shielded gradient sets
(40 mm inner diameter (ID) capable of 1.5 T/m maximum
gradient strength or 57 mm ID with 0.3 T/m). The birdcage
resonators (Bruker, Rheinstetten, Germany) used for the
experiments (25, 30, and 38 mm ID) are tunable to both1H
and19F (see Note 1).
3. Experiments are operated by the software ParaVision 5.1 and
TopSpin 2.0 (Bruker, Rheinstetten, Germany).
2.2 Data 1. Merging1H and19F MR images is carried out with an in-house

Visualization developed software module (see Note 2) on a standard Win-
and Quantification dows PC. For superimposing the images of both nuclei, a “hot
iron” color lookup table was applied to19F images.
2. The visualization as well as the quantification of CSI data is also
handled with an in-house developed software module (see Note
3) on a standard Windows PC.
2.3 Microfluidization 1. For the preparation of PFCs, an Ultra Turrax (TP 18/10;
of PFCs IKA-Werke, Staufen, Germany) is used for the primary high-
performance dispersion process, and a Microfluidizer (M110P,
Microfluidics, Newton, MA, USA) is applied for the
subsequent homogenization process at high pressure.
Tuba Gu
2. Compounds: perfluoro-15-crown-5 ether (PFCE), perfluor-

ooctyl bromide (PFOB), trans-bis(perfluorobutyl)ethylene
(F-44E), phospholipid E80S (kindly provided from Lipoid
AG, Ludwigshafen, Germany), 1-(perfluoro-n-hexyl)decane
(diblock, F6H10), phosphate buffer (see Note 4).
3. The PFCs are characterized on a Zetatrac NPA152 (Microtrac
Inc., North Largo, Florida, USA) for particle size distribution
and ζ-potential.
4. Emulsions:
(a) PFCE nanoemulsion formation: 2.4% (w/w) phospholipid
(Lipoid E80S) is dispersed in 77.6% (w/w) phosphate
buffer (10 mM) isotonized with glycerol (2.5%) (see
Note 4). Thereafter, 20% (w/w) PFCE (see Note 5) is
added and a crude emulsion is formed by high shear
mixing (Ultra Turrax). High-pressure homogenization is
performed in 10 cycles at 1000 bar (Microfluidizer). The
nanoemulsion is heat-sterilized in glass vials under stan-
dard conditions (121 C, 2 bar, 20 min).
(b) PFOB nanoemulsion formation: The formation of PFOB
nanoemulsion is essentially accomplished as described
above. 2.9% (w/w) phospholipid (Lipoid E80S) is dis-
persed in 44.2% (w/w) 10 mM phosphate buffer (see
Note 4). For stabilization of the emulsion, a semi-
fluorinated fluorocarbon/hydrocarbon diblock alkane
compound (C6F13C10H21, F6H10; equimolar with
E80S) is added (see Note 6). After addition of 50%
(w/w) PFOB, the emulsion is pretreated with the high-
performance disperser (Ultra Turrax), and directly after-
wards, homogenized at 1000 bar in 10 cycles (see above).
The PFOB nanoemulsion is heat-sterilized under standard
conditions (121 C, 2 bar, 20 min).
(c) F-44E nanoemulsion formation: The formation of F-44E
nanoemulsion is performed as described for the formation
of PFOB nanoemulsion except for the composition of
3.0% (w/w) phospholipid and 53% (w/w) F-44E.
2.4 Animal 1. For animal narcosis, isoflurane is used as inhalation anesthetic

Anesthesia (see Notes 7 and 8). Isoflurane (1.5 Vol.%) is mixed with 30%
and Handling oxygen O2 in nitrogen N2. The water-saturated gas mixture is
controlled through a gas flow control unit with flow meters
(V-100 from Voegtlin, Aesch, Switzerland) for O2 and N2 (see
Note 9).
2. For applying the anesthesia gas mixture, our setup includes an
in-house made nose cone for mice connected with a gas inlet
and, of importance, additionally with a gas outlet (see Note 10).
19
F MRI 239
3. A small animal monitoring system providing a pneumatic pil-

low, a rectal temperature probe, and ECG electrodes is used to
monitor the animals’ vital functions. This system can also be
utilized to synchronize data acquisition with cardiac and respi-
ratory motion in the course of specific MR measurements.
4. To avoid hypothermia of the mice during MR imaging experi-
ments, the animals’ body temperature is kept at 37 C using a
tempering unit (see Note 11).
5. The animal handling system to insert anesthetized animals into
the magnet is custom-made.
3 Methods
3.1 Perfluorocarbon 1. The nanoemulsions are stored at 4 C, but should be

Nanoemulsion pre-warmed to body temperature shortly before application.
Preparation The nanoemulsions are stable over years, but over time there
might be some sedimentation of the emulsion due to the high
density of the perfluorocarbons, wherefore they should be
gently shaken before injection.
2. Characterization: The resulting nanoemulsions are character-
ized by photon correlation spectroscopy (Zetatrac NPA152) to
determine the hydrodynamic diameter, and the polydispersity
index (PDI). The ζ potential is measured with the same device
(see Notes 12 and 13). The average particle diameter and ζ
potential for unmodified nanoemulsions are about 100 nm and
35.6 mV for PFCE, 256 nm and 42.1 mV for PFOB, as well
as 237 nm and 41.7 mV for F-44E. The PDI should be in the
range of 0.10–0.20 to ensure adequate homogeneity of the
nanoemulsions.
3.2 Surface Currently, different ways to modify lipid surfaces of PFC nanopar-
Functionalization ticles are in use. One of these methods is the sterol-based post-
of PFCs by Sterol- insertion technique (SPIT) which enables the modification of pre-
Based Post-insertion formed PFCs under very mild conditions using ligands coupled to
cholesterol moieties. The cholesterol moiety acts as an anchor
which spontaneously inserts into the lipid layer of the PFCs at
ambient temperature. Thus, the SPIT technique is particularly
suitable for labile targeting ligands which might be destroyed by
the harsh process of high-pressure homogenization or microfluidi-
zation [7, 8].
1. Before SPIT can be performed, a cholesterol containing com-
pound has to be prepared. The compound consists of the
cholesterol moiety (chol), a targeting ligand as well as a linker,
mostly a polyethylene glycol molecule of 1–2 kDa (PEG1000-
Tuba Gu
PEG2000). However, several other possibilities depending on

the application purpose are conceivable.
2. For specific targeting of acute thrombus formation, we used a
cholesterol-PEG2000-α2AP anchor (see Note 14) for
subsequent SPIT. As control, a chol-PEG2000-Q3A anchor
was prepared (see Note 14).
3. To perform SPIT, chol-PEG2000-α2AP or chol-PEG2000-Q3A
is added to the preformed PFCs in a molar ratio of 10:1
cholesterol:phospholipid. The incubation is conducted on a
rotary shaker at 17 C (room temperature is also possible) for
5–6 h (see Note 15), during which the cholesterol moiety of
the targeting as well as non-targeting compounds inserts spon-
taneously into the phospholipid layer of the PFCs.
4. Non-inserted anchors can be separated from the surface-modified
PFC particles by a size exclusion chromatography step. In this
case, it should be considered that the nanoemulsion is highly
diluted and should be concentrated prior to use. Since we apply
the cholesterol anchor in small amounts (5 mol% of the total
lipid), we assume that the cholesterol conjugates are inserted
completely, and therefore, in most cases skip this procedure.
5. Characterization: (Note 13) PCS measurement resulted in
values of about 165 nm (size), 11.7 mV (ζ potential), and a
PDI of 0.1–0.2 for targeted PFCs, respectively.
3.3 Disease Models Herein, we give an overview of models which were successfully
monitored by1H/19F MRI for immune cell tracking or thrombus
formation (see Subheading 4). Since the detailed description of
these models would be beyond the scope of this chapter, we refer
the interested reader to the according references. For the visualiza-
tion of inflammatory patterns or thrombus formation, mice
received in each case an intravenous bolus application of PFCs
(3 mM/kg body weight) at least 24 h prior the imaging session
to ensure adequate19F loading of circulating monocytes/macro-
phages and efficient binding to the thrombus, respectively.
1. Otitis model: Ear clipping is routinely used for animal identifi-
cation. We used ear clipping as a simple and reproducible
inflammation model as described by Jacoby et al. [6].
2. Induction of glomerulonephritis: The induction of anti-GBM
nephritis (GBM ¼ glomerular basement membrane) was per-
formed essentially as accomplished by Rosenkranz et al.
[9]. Three days after a preimmunization step, 0.25 ml of anti-
GBM antibodies were injected intravenously through the tail
vein. Control animals received 0.25 ml medium.
3. Subcutaneous inflammation model: We elaborated a simple
inflammation model injecting lipopolysaccharide (LPS) loaded
19
F MRI 241
matrigel (50 μg LPS in 50 μl matrigel) subcutaneously into the

neck of the animal (see Note 16). Details concerning this kind
of inflammation model can be found in Temme et al. [10].
4. Human ex vivo and murine in vivo thrombus model: For ex vivo
formation of thrombi, 5 ml of human blood was collected by
puncture of the vein in citrate tubes. The blood sample was
centrifuged at 150 g for 10 min, and thereafter, about 1 ml of
platelet rich plasma (PRP) was collected and supplemented
with 100 μl thrombin (5 U/ml), 100 μl CaCl2 (0.4 mM),
and 5 μl ADP (1 mM). The PRP was dispensed on a 96-well
plate (round-bottom) with 100 μl PRP per well, and 25 μl of
targeted or non-targeted PFCs were added and incubated for
60 min at 37 C. The formed thrombi were harvested from
each well and washed with tyrode solution.
In vivo induction of thrombi in the murine vena cava was
carried out as described by Temme et al. [8]. In brief, the top of
the inferior vena cava was treated with a small filter paper
soaked with a 10% ferric chloride solution (FeCl3) for 4 min.
To allow labeling of freshly formed thrombi by α2AP-PFCs, the
targeted nanoemulsions were injected 5 min prior to thrombus
induction via the tail vein.
3.4 Starting the MR 1. The animal is anesthetized with 1.5% isoflurane in a water-
Experiment saturated gas mixture of 30% oxygen in nitrogen applied at a
rate of 75 ml/min by manually restraining the animal and
placing its head in an in-house built nose cone (see Note 10).
Thereafter, it is fixed within a custom-made animal handling
system, and then, inserted into the magnet. Afterwards, the
coil is tuned and matched on the1H resonance. A pilot scan is
carried out to verify the correct positioning of the animal
within the resonator, and, if necessary, the position of the
mouse is corrected (see Note 17).
2. Before any experiment is started, it is important to adjust the
shim and basic frequency and to determine the reference atten-
uation as well as the optimal receiver gain.
3. The anatomical reference images (1H) of the region of interest
are acquired by performing either FLASH (fast low-angle shot)
or RARE (rapid acquisition with relaxation enhancement)
experiments (see other chapters in this book for detailed
description of1H MR techniques).
3.5 Adjusting 1. After the acquisition of the anatomical reference images, the
Parameters for19F MRI coil is tuned and matched to the19F resonance frequency.
2. Since the19F spin density will be quite sparse (even after injec-
tion of PFCs), most of the routinely applied automatic adjust-
ments will fail. Therefore, reference pulse gains for the distinct
Tuba Gu
resonators have to be determined in preliminary experiments

with undiluted perfluorocarbon emulsions. From these phan-
toms, also the basic frequencies suitable for imaging of the
individual PFCs have to be identified by19F MR spectroscopy.
3. Depending on the fluorinated compound used, the appropriate
frequency is set on-resonance (see Note 18).
4. In experiments with high fluorine load, the adequate setting
may be verified with a non-volume selective19F MR
spectroscopy scan.
5. For19F MRI, the geometry of the previous1H scans is imported
to ensure the anatomical corresponding localization of images
from both nuclei (see Note 19).
6. Depending on the administered contrast agent, conventional,
narrow spectral bandwidth or chemical shift imaging (CSI) are
performed (see following sections).
3.6 Conventional imaging is performed for compounds with simple

Conventional19F MRI MR spectra such as PFCE with a single resonance peak (Fig. 2a,
top). With regard to pulse sequences, the use of a RARE sequence
has proved to be advantageous for a variety of reasons. On the one
hand, relaxation times of PFC emulsions (T1 ~1 s; T2 ~0.5 s) are
well suited for RARE (long T2 with only twice T1). Routinely, a
RARE factor of 32 results in an appropriate SNR with a resolution
(voxel size 0.2 μl) close to the anatomical1H MR image within an
acceptable acquisition time (20 min). Another major advantage of
the RARE over gradient echo sequences is that it results in a signal
void of flowing blood particles, since they are not refocused. There-
fore, detected signals can be attributed unequivocally to bound
PFCs at the target site without contamination from19F signals of
circulating PFCs, which is of particular importance for specific
detection of targeted PFCs bound to the thrombus surface within
the vessels.
1. Parameters routinely used are: FOV 25.6 25.6 mm2, Matrix
64 64, ST 1–2 mm (see Note 19), RARE factor ¼ 32,
TR ¼ 2500 ms, TE ¼ 4.4 ms, 256 averages, acquisition
time ¼ 21 min (see Note 20).
2. For PFCE, an effective spectral acquisition bandwidth of
25 kHz is used. (Fig. 2).
3.7 Adjusting For compounds with complex19F MR spectra the spectral band-
the Spectral width used for conventional RARE imaging (cf. Subheading 3.6)
Bandwidth leads to chemical shift artifacts. However, due to the relatively large
for Multipeak MR chemical shift range of the19F nuclei (from 300 to 400 ppm)
Spectra these artifacts can be avoided by adjusting the19F RARE sequence
and aligning the effective spectral bandwidth to lower values. This
method is applicable for compounds like PFOB and F-44E with
19
F MRI 243
Fig. 2 Suitability of various PFCs for in vivo19F MR inflammation imaging in an otitis model. (a)19F MR spectra
of PFCE (top), PFOB (middle), and F-44E (bottom). PFCE generates a single resonance peak, whereas for PFOB
and F-44E multipeak MR spectra are obtained due to the presence of nonequivalent19F nuclei. Red brackets
indicate the spectral bandwidths used for the individual compound of 25 kHz for PFCE, and 10 kHz for PFOB
and F-44E, respectively. (b) In vivo visualization of inflammation after ear clipping:1H/19F MR images of the ear
acquired 24 h after clipping and intraveneous injection of PFCE (top), PFOB (middle), and F-44E (bottom),
respectively. In the1H MR images (left) the clipping areas are clearly evident. Anatomically corresponding
artifact-free19F images (middle) could be obtained for all PFCs with the spectral acquisition bandwidths given
in (a). The area of inflammation is clearly delineated by each PFC with almost comparable sensitivity.
Interestingly,19F signals were predominantly found at the proximal, highly perfused base of the ear as can
be recognized in the merged images (right)
MR spectra in which at least one strong signal is well resolved from

the other resonance frequencies (Fig. 2a).
1. The most suitable19F resonance frequency for PFOB and
F-44E is selected and set on-resonance, respectively. For both
compounds, the signal originating from the CF3 group is used.
Tuba Gu
2. Again, in experiments with high fluorine load, the adequate

setting may be verified with a non-volume selective19F MR
spectroscopy scan.
3. To avoid contaminations from the other resonances the effec-
tive spectral bandwidth is reduced to 10 kHz (see Note 21).
4. Parameters are the same as in Subheading 3.6 with exception of
TE which rises to 8.2 ms due to reduction of the spectral
bandwidth.
3.8 Chemical Shift As already pointed out above, conventional gradient or spin echo
Imaging for Multipeak sequences for the imaging of PFCs with complex MR spectra
MR Spectra produce chemical shift artifacts along the readout direction. At
the same time, slice selection is impaired by displacement errors,
due to its frequency dependence. Using chemical shift imaging
(CSI) all spatial dimensions are phase encoded, wherefore chemical
shift artifacts are absent. Compared to the narrow bandwidth19F
RARE sequence (cf. see Subheading 3.7), in which only one reso-
nance frequency is used for imaging, the CSI method allows the
selection of all arising19F resonance frequencies for reconstruction
of the19F image.
1. For 2D CSI, spectra are acquired within a slice, spatially
arranged in a two-dimensional matrix. 2D CSI can be per-
formed with the following parameters: FOV 20 20 mm2,
Matrix 256 25 25 (first dimension is spectroscopic), ST
2 mm, sinc3 excitation pulse length ¼ 150 μs, flip angle ¼ 23 ,
TR ¼ 30 ms, TE ¼ 739 μs, sine-bell acquisition weighting with
a maximum of 28 averages, acquisition time ¼ 2.5 min.
2. It is also possible to examine the distribution of spectral com-
ponents within a volume of interest by performing 3D CSI
experiments using the following parameters: FOV
30 30 32 mm3, Matrix 256 33 33 17 (first
dimension is spectroscopic), sinc3 excitation pulse
length ¼ 150 μs, flip angle ¼ 23 , TR ¼ 30 ms, TE ¼ 732 μs,
averages ¼ 1, acquisition time ¼ 5.1 min.
3.9 Applications The high affinity of the macrophage/monocyte system for PFC
uptake (see Fig. 1) can be exploited to monitor the infiltration of
3.9.1 Immune Cell
PFC-loaded immune cells into inflammatory foci by19F MRI
Tracking with PFCs
in vivo [2, 6, 11, 12]. This enables the noninvasive visualization
of regions affected by inflammatory processes in several clinically
relevant disease models [2], such as viral myocarditis, ischemic
heart disease, stroke, atherosclerosis, glomerulonephritis, and
others.19F MRI is performed depending on the administered PFC
with a conventional, narrow bandwidth, or CSI sequence.
1. Conventional19F MRI using PFCE in a glomerulonephritis
model
19
F MRI 245
Fig. 3 Visualization of enhanced renal inflammation in CD73/ mice by1H/19F MRI. (a) Anatomically
corresponding sections of1H and19F MR images from individual WT and CD73/ mice acquired 10 days
after induction of nephritis showing an increased accumulation of19F signal in kidneys of the CD73-deficient
mice as compared to WT. Red arrows: cortex and medulla renalis; yellow arrows: subcutaneous fat; SC: spinal
cord; LK and RK: left and right kidney. (b) Quantification of19F signal intensity within the kidney 10 and 17 days
after immunization (n ¼ 5, *p < 0.05)
Based on a glomerulonephritis model in CD73-deficient

(CD73/) mice, we could use19F MRI to demonstrate the
crucial role of extracellularly formed adenosine as an innate
protective mechanism for the kidneys [13]. For monitoring
of ongoing inflammation, mice were imaged at days 10 and
17 after induction of nephritis. In RARE1H MR images
kidneys could be easily identified with contrast between
medulla and cortex (Fig. 3a, top). The corresponding19F MR
images clearly show enhanced infiltration of monocytes/
macrophages into kidneys of CD73/ mice as compared to
WT controls 10 days after induction (Fig. 3a, middle). Note,
that the19F signal was predominantly detected within the renal
cortex (which contain most of the glomeruli) in both WT and
CD73/ mice (Fig. 3a, bottom). Integration of19F signal
intensities revealed an almost two-fold increased accumulation
of PFCs within the kidney of CD73-deficient mice already
10 days after induction of nephritis and these alterations
became statistically significant 1 week later (Fig. 3b). There
were no differences in kidney volumes between WT and
CD73/ mice during the observation period as determined
from multislice1H MR data sets (data not shown).
Tuba Gu
Furthermore, no differences were observed between the two

groups in absence of nephritis (data not shown).
2. Narrow spectral bandwidth19F MRI for PFCs with complex MR
spectra in an otitis model
Ear clipping is routinely used for animal identification. As
this process is associated with the development of inflammation
around the clipped area, we used this as a simple and reproduc-
ible inflammation model to compare the performance of the
clinically applicable PFOB and F-44E with the preclinical gold
standard PFCE. The latter was imaged in a conventional man-
ner, while for PFOB and F-44E, the effective spectral band-
width was reduced to 10 kHz. As can be recognized from
Fig. 2, the area of inflammation could be clearly visualized by
each of the PFC utilized. Also there are no chemical shift
artifacts for PFOB and F-44E, which delineated the inflamed
areas in the ear with similar sensitivity like PFCE. Interestingly,
in each case the inflammatory patterns were most pronounced
at the proximal, highly perfused base of the ear.
3. 3D19F CSI of F-44E in a model of subcutaneous inflammation
To induce subcutaneous inflammation, we injected lipopoly-
saccharide (LPS) loaded matrigel (see Note 16) into the neck of
the animals (Fig. 4a). Twenty-four hours after matrigel implan-
tation, F-44E was applied via the tail vein and 3D19F CSI was
utilized to monitor the development of subcutaneous inflam-
mation. Whereas with the approach described above, only
the19F signal arising from the CF3 groups (six19F nuclei out
of 18) is used, the CSI sequence can utilize all signals from the
18 fluorine for image reconstruction (see Fig. 6 + Note 3).
Merging of1H and19F images revealed that PFCs are located
predominantly in the border zone of the matrigel plug
(Fig. 4b). No19F signal was found in control plugs (data not
shown) indicating that matrigel itself does not induce inflam-
mation. As matrigel is of murine origin, no rejection reaction
was expected. Subsequent histology confirmed the predomi-
nant localization of immune cells at the periphery of the plug,
however the exact reason for this is yet unclear. Most likely,
high local chemokine concentrations or distinct stop signals
could account for the restriction of the infiltrated cells to the
periphery (Fig. 4).
3.9.2 Active Targeting Due to the PFC uptake in macrophages and monocytes,19F MR
of Thrombi imaging with neat PFCs can be used for tracking of these cells in a
with Functionalized PFCs variety of settings. However, the functionalization of PFCs with
binding ligands enables a more specific targeting, and, therefore,
allows the visualization of structures which would not be labeled by
non-functionalized PFCs [14, 15].
19
F MRI 247
Fig. 4 Monitoring of immune cell infiltration in a subcutaneous inflammation model. (a) Anatomical location of
the LPS-doped matrigel for induction of subcutaneous inflammation. Sagittal (top) and axial slices (bottom)
show the site of injection (yellow arrows). The red, dashed rectangle corresponds to the position of the axial
slices shown on the right. (b)1H/19F MRI of mice injected with LPS-doped matrigel (1 μg/μl). F-44E was
injected 24 h after matrigel implantation, and animals were scanned by1H/19F MRI 24 h later. Yellow arrows
indicate the location of the matrigel plug.19F 3D CSI images were reconstructed with the CSI tool (cf. Fig. 6;
top:1H; middle:19F; bottom: merge). The inflamed area can be clearly located next to the matrigel
19
1. F MRI using targeted PFCs in an ex vivo as well as in vivo
thrombus model
During the early phase of thrombus formation, factor XIIIa
crosslinks α2-antiplasmin with fibrin [16]. Thus, peptides
derived from α2-antiplasmin (α2AP) can be used for specific
labeling of freshly developed thrombi. By means of SPIT, α2AP-
decorated PFCs with PFCE as core were formed (see Note 14)
and applied for active targeting and visualization of acute
thrombi by19F MRI. As negative control, a low affinity peptide
for factor XIIIa, i.e. Q3A [8, 17], was coupled to PFCs.
Figure 5a shows ex vivo human thrombi in vials incubated
with Q3A-PFCs (left vial) and α2AP-PFCs (right vial). As
expected, the application of Q3A-PFCs did not result in any
Tuba Gu
Fig. 5 Specific targeting of thrombi with α2AP-functionalized PFCs. (a)1H/19F MRI of vials containing ex vivo
formed human thrombi:1H MR reference (left),19F RARE image (middle), and merged images (right). The19F MR
signal only arises from thrombi exposed to α2AP-PFCs (right vial). Q3A-PFCs (left vial) were used as negative
control. (b) Top: In vivo visualization of thrombus formation in the murine vena cava with anatomical1H MR
reference (left),19F RARE image (middle), and merged images (right). Bottom: Enlargement of the area around
the vena cava. The formed thrombus can be vaguely recognized in the1H MR image (left), while the hot spot in
the19F MR image (middle) clearly indicates its presence within the vena cava when merged with the
anatomical reference (right). SC spinal cord, LK and RK left and right kidney, VC vena cava, AA aorta
abdominalis
significant19F signal, whereas the active targeting approach

with α2AP-PFCs successfully labeled the formed thrombi
resulting in an SNR of >100.
Furthermore, this approach was also suitable for unequivocal
detection of thrombi in the in vivo setting. In a murine model
of deep venous thrombosis, we were able to show that actively
targeted PFCs could be successfully employed to detect
thrombi with a diameter of less than 1 mm and very high
signal-to-noise ratio (SNR > 70). While the newly formed
19
F MRI 249
thrombus can only be dimly recognized in1H MRI at the

ventral side of the vena cava as a dark gray structure (arrows
in Fig. 1b, bottom), α2AP-PFCs clearly delineated the thrombus
in the19F MR image (Fig. 5b middle + right) (Fig. 5).
4 Notes
1. The birdcage resonators used in our setup are tuneable to

both1H and19F, which can be technically realized due to the
relative small frequency gap between1H and19F (~24 MHz at
9.4 T, 400.21 vs. 376.54 MHz). Despite less comfortable, this
coil setup has turned out to be preferable over double-tuned
resonators in terms of homogeneity and sensitivity for both
nuclei.
2. The ParaVision software (PV 5.1) provides an overlay of images
acquired from different nuclei with rudimentary functionality
only. Therefore, we developed a tool (written in LabVIEW v6,
National Instruments, Austin, TX, USA) which allows the
merging of multislice as well as 3D image stacks even for
different in-plane resolutions. Furthermore, it provides several
reconstruction options like the application of zero filling and
window functions. All images showing overlays of1H and19F
datasets in this chapter were prepared with this software mod-
ule. Of note, in principle all reconstructed Bruker image data-
sets (2dseq) can also be evaluated with the free available
software ImageJ using the plugin “Import Bruker NMR files.”
3. In addition, the evaluation of acquired CSI data by ParaVision
5.1 is quite uncomfortable and inefficient. The in-house devel-
oped CSI analysis tool (LabVIEW v6, see above) enables the
reconstruction of 2D and 3D CSI datasets with several options,
such as line broadening, zero filling, spatial window functions
etc., and, thereafter, the visualization and quantification of the
data. In Fig. 6a the graphical user interface (GUI) of the CSI
tool is shown. For each pixel in the1H image (green grid) the
correlated19F spectrum of the19F CSI dataset can be selected
(Fig. 6a, top). Thereafter, integration gates can be defined
manually (Fig. 6a, bottom), and within these gates, all individual
spectral intensities are summed up in a pixel-wise manner
which yields in a19F image free of chemical shift artifacts
(Fig. 6b). The resulting images are quantified by means of
signal-to-noise or intensity measurements (Fig. 6).
4. Phosphate buffer: 0.08% (w/w) sodium dihydrogen phos-
phate, 0.85% (w/w) disodium hydrogen phosphate, 0.40%
(w/w) sodium chloride, 2.5% (w/w) glycerol, demineralized
water (millipore grade). The pH is adjusted to 7.0.
Tuba Gu
Fig. 6 In-house developed software module for visualization and quantification of CSI datasets. (a) GUI of the
CSI analysis tool: Images loaded into the tool were acquired after injection of an emulsion mixture of PFCE and
PFOB into a healthy mouse. Top left:1H MR reference image with overlaying grid representing the in-plane
matrix. The red cross depicts the selected voxel. Top right: 2D19F CSI dataset with the individual19F MR
spectra in each voxel (FOV 2.56 2.56 cm2, matrix 64 64, zero-filling to 128 128, ST 1 mm, acquisition
time 16 min). Bottom:19F MR spectrum of the selected voxel (red cross) indicating the presence of both PFOB
and PFCE in the liver. The enlargement of this spectrum given beneath (red frame) shows the integration gates
for resonance peaks originating from PFOB. (b) Reconstruction of the19F MR image (middle) from the pixel-
wise summation of spectral intensities in the integration gates displayed above and merge (right) with the
anatomical1H MR image (left) demonstrating an equal distribution of PFOB over the entire liver
19
F MRI 251
5. Of note, perfluorocarbons are volatile substances, hence,

handling of the perfluorocarbons during the emulsification
processes need to be performed quickly.
6. Perfluorocarbons possess, due to their high fluorine payload,
lipophobic as well as hydrophobic properties and are in partic-
ular fluorophil. Emulsification with phospholipids is therefore a
challenge, since the hydrophilic phosphate group is strongly
rejected by the perfluorocarbons. On the other hand, they also
lack a strong affinity for the lipophilic phase. Therefore, semi-
fluorinated alkanes (fluorocarbon/hydrocarbon diblocks) are
added in small amounts during perfluorocarbon nanoemulsion
formation. While the lipophilic hydrocarbon part of the
diblock inserts into the lipophilic area of the phospholipid
layer, the fluorophilic part penetrates the fluorocarbon phase.
This kind of cross-linking facilitates the stabilization of the
particles as well as a more precise control of the particle size [5].
7. Inhalation anesthesia using isoflurane is preferred to injection
anesthesia due to its convenient controllability. Furthermore,
both the induction of and recovery from isoflurane anesthesia
are rapid and display only a negligible cardiorespiratory depres-
sion, resulting in murine physiological heart (about 600 bpm)
and respiration rates (about 100/min), respectively. A clinical
vaporizer can be used but should be equipped with a gas flow
control unit suitable for mice (see Note 7). For long-term
experiments, it should be considered that isoflurane accumu-
lates in lipid-rich areas due to its lipophilicity (see Note 6).
8. Isoflurane tends to accumulate over time within regions of high
lipid content due to its lipophilic properties. However, during
long-term experiments with control mice that did not receive
any PFC injections, we found19F MR signals from isoflurane in
the lipid-rich area of the thorax (subcutaneous fat) not detect-
able before 1 h of anesthesia (Fig. 7), when routine19F imaging
parameters were used (see Subheading 3.6) (Fig. 7).
9. The anesthesia gas mixture consisting of isoflurane, oxygen,
and nitrogen is moisturized to prevent the dehydration of
mucous membranes by passing the gas mixture through a
gas-washing bottle filled with water. Notably, for controlling
the gas flow, the use of clinical standard flow meters capable to
regulate gas flows >1–2 l/min should be avoided, since for
mice with a physiological respiration rate of 100/min and a
murine tidal volume of 0.1–0.2 ml a maximum required flow of
20 ml/min is calculated. The flow rate of 50–75 ml/min used
in our setup is more than adequate to support normal mouse
ventilation. Higher flow rates will facilitate dehydration and
result in non-physiological conditions. The V-100 units used
Tuba Gu
Fig. 7 Appearance of isoflurane signals in19F MR images over time. (a) Chemical structure of isoflurane. (b)
Long-term MR investigation of a control mouse under isoflurane anesthesia (1.5%) without PFC injection.
Left:1H MR images of the mouse thorax (GE ¼ gradient echo; T1-weighted; RARE; T2-weighted) with slices
from the base of the heart (top) down to the apex (bottom) and the upper part of the liver. Right:19F MR images
(ST 2 mm, matrix 64 64) of the same locations at different time points (acquisition time 20 min) showing the
slow appearance of19F signals from isoflurane which is known to accumulate in lipid-rich areas, here in
particular in the subcutaneous fat. Of note, up to 1 h of anesthesia isoflurane-related signals are almost
negligible
in our setup are calibrated for gas flows of 10–100 ml/min

(N2) and 5–50 ml/min (O2).
10. Due to its fluorine content (see chemical structure in Fig. 7a),
isoflurane itself gives rise for signals in19F MR images, which
may be misinterpreted as false-positive signals from PFCs (see
Note 6 and Fig. 7). To avoid the accumulation of gaseous
isoflurane in the resonator, our setup (Fig. 8a) includes an
in-house made nose cone for mice (Fig. 8b), which has the
necessary gas inlet for delivery of the anesthetic to the animal,
and, of importance, a gas outlet connected. The latter prevents
the accumulation within the probe head (as well as the labora-
tory), and, therefore, the appearance of disturbing19F MR
signals from isoflurane (Fig. 8).
19
F MRI 253
Fig. 8 Setup for inhalation anesthesia. (a) The in-house-built nose cone for mice
is connected to an isoflurane vaporizer and a gas flow control unit with flow
meters (V-100) for O2 and N2. This setup enables the convenient control of
anesthesia throughout the experiment. (b) The magnification of the nose cone
shows the elastic mask, which tightly clasps around the nose of the mouse, the
inlet and the offlet system for the inhalation anesthesia to avoid accumulation of
isoflurane within the probe head
11. In our setup, the temperature is kept at 37 C by regulating the

temperature of the gradient cooling device (BCU 20, Bruker,
Germany). However, several other possibilities of temperature
regulation are conceivable.
12. Photon correlation spectroscopy (PCS) is a light scattering
technique that is used for the determination of particle sizes
and other properties. The particle size is given in terms of the
hydrodynamic diameter, which represents the diameter of a
sphere that has the same translational diffusion coefficient as
the particle. The size of the PFC droplets can be controlled
among others through the ratio of emulsifier/perfluorocarbon
and the number of high-pressure homogenization cycles. The
polydispersity index (PDI) indicates the heterogeneity or
homogeneity of the PFC mixture. The smaller the size distri-
bution, the smaller the PDI value, and the higher the
Tuba Gu
homogeneity. In general, PDI values smaller than 0.05 are

ordinarily seen in highly monodisperse standards only. PDI
values greater than 0.5 indicate a very broad size distribution
of the particles, and thus, a large heterogeneity. The formation
of the PFCs in our setup leads to PDI values of 0.10–0.20,
indicating a small size distribution and therefore a good homo-
geneity of the nanoemulsions. Another important parameter
when characterizing PFCs is the ζ potential. The ζ potential is a
surface potential that exists between the particle surface and
the dispersing liquid. More precisely, it is the potential differ-
ence between the dispersion medium and the stationary layer
of the particle surface after shearing off the diffuse layer of ions
by an external electric field. An adequate ζ potential prevents
the formation of a contact phase between nanoparticles, and
therefore, stabilizes the particle size. A value of approximately
30 mV or lower seems to be sufficient to ensure long-term
stability of the PFCs. Of note, not all PCS devices allow for
concomitant determination of both the particle size and the ζ
potential. The Zetatrac used for our experiments is capable of
the estimation of the ζ potential as well as the determination of
the particle size.
13. PCS performance: After equilibration of the laser, the sample
chamber is washed with millipore grade water and checked for
purity with 1 ml of water. Subsequently, the water is removed,
and the background signal is determined against 1 ml phos-
phate buffer (see Note 4). For measurement of PFCs, 10 μl of
the prepared nanoemulsions is diluted in 1 ml phosphate buffer
and placed in the sample chamber. The final measurement runs
over 10 cycles and values are averaged over three samples from
each nanoemulsion.
14. First, a cholesterol-PEG anchor with maleimide as reactive
group is synthetized. To this end, an equimolar mixture of
maleimide-PEG2000-NH2 and cholesteryl chloroformate is
reacted with the activator triethylamine for 24 h under exclu-
sion of light in a nitrogen atmosphere. For the purification of
the product chol-PEG2000-maleimide, a chromatography step
is conducted using a sephadex LH20 column.
For specific targeting of acute thrombi, a 14 amino acid peptide
derived from α2-antiplasmin (α2AP) is used. During thrombus
formation, α2-antiplasmin is cross-linked to fibrin at the glutamine
Q3 side by factor XIIIa [16, 17]. As control, glutamine Q3 is
converted into alanine Q3A which is known to show low affinity
for factor XIIIa. The α2AP—as well as the Q3A—peptide is func-
tionalized with a cysteine residue at the amino acid position 13, at
which the coupling reaction with the maleimide group of chol-
19
F MRI 255
PEG2000-maleimide takes place for the formation of chol-PEG2000-

α2AP and chol-PEG2000-Q3A [8].
15. Shorter incubation times might be feasible, but have to be
determined empirically.
16. Matrigel is fluid at 4 C, and should, therefore, be kept cool
until injection. At body temperature, the fluid converts into
a gel.
17. The applied1H/19F resonators exhibit a limited FOV in the
xy-plane (determined by the ID) as well as in z-direction
(in dependence of the coil length). To ensure the greatest
possible magnetic field homogeneity, it is important to posi-
tion the area of interest as close as possible to the isocenter of
the resonator.
18. It is recommended to set the basic frequencies (BF) for the
individual PFCs via a macro which makes use of the initial
adjustments carried out for the first1H MR scan of a study.
Usage of the automatically determined1H BF for the object of
the current study will allow the calculation of the correct19F
BF from the absolute gap of1H water and19F resonance fre-
quencies of the PFC, which has to be previously determined
from1H/19F MR spectra of the individual perfluorocarbon
emulsions. For Bruker systems running PV5.1 the current1H
BF can be readout from the user-specific file “non_volatile_-
vals,” and after calculation of the appropriate19F BF this value
can be set with “exec pvcmd -set pvScan BF1 $value.” Listing
of the entire macro would be beyond the scope of this chap-
ter, but can be obtained from the authors.
19. As19F nuclear spins are quite sparse compared to1H nuclei in a
physiological environment, the slice thickness (ST) in19F MR
experiments may be set to higher values to increase the voxel
size and to include more19F nuclei for enhancement of the
signal intensity.
20. The field-of-view (FOV) should cover the whole area of inter-
est and the entire cross section of the animal to avoid aliasing.
TR stands for the repetition time, TE for the echo time.
21. Different19F resonance frequencies will lead to chemical shift
artifacts along the frequency encoding direction in19F MR
images. By decreasing the effective spectral acquisition band-
width, one well resolved resonance frequency can be isolated
for conventional imaging purposes. This will avoid chemical
shift artifacts, but—since the acquisition window prolongs
reciprocally with bandwidth reduction—will strongly increase
TE and thereby lead to a loss in SNR.
Tuba Gu
Acknowledgements
The authors would like to thank Prof. J€ urgen Schrader

(D€usseldorf) for his continuous support and encouragement as
well as Prof. Rolf Schubert and Dr. Christoph Grapentin (Freiburg)
for their enormous help in the development of the perfluorocarbon
emulsions (PFCs). Furthermore, we would like to thank Prof.
Cornelia Blume (Hannover) for providing us with the glomerulo-
nephritis model. The work shown herein was supported financially
by the Deutsche Forschungsgemeinschaft (DFG), the subproject
Z2 of the SFB 612, subprojects B2 and B5 of the SFB 1116, and
grants SCHR 154/13-1þ2.
References
1. Bartusik D, Aebisher D (2014)19F applications 8. Temme S, Grapentin C, Quast C, Jacoby C,
in drug development and imaging—a review. Grandoch M, Ding Z, Owenier C, Mayenfels F,
Biomed Pharmacother 68(6):813–817. Fischer JW, Schubert R, Schrader J, Flögel U
https://doi.org/10.1016/j.biopha.2014.07. (2015) Noninvasive imaging of early venous
012 thrombosis by 19F magnetic resonance imag-
2. Flögel U, Ding Z, Hardung H, Jander S, ing with targeted perfluorocarbon nanoemul-
Reichmann G, Jacoby C, Schubert R, Schrader sions. Circulation 131(16):1405–1414.
J (2008) In vivo monitoring of inflammation https://doi.org/10.1161/
after cardiac and cerebral ischemia by fluorine CIRCULATIONAHA.114.010962
magnetic resonance imaging. Circulation 118 9. Rosenkranz AR, Mendrick DL, Cotran RS,
(2):140–148. https://doi.org/10.1161/ Mayadas TN (1999) P-selectin deficiency
CIRCULATIONAHA.107.737890 exacerbates experimental glomerulonephritis:
3. Bulte JW (2005) Hot spot MRI emerges from a protective role for endothelial P-selectin in
the background. Nat Biotechnol 23 inflammation. J Clin Invest 103(5):649–659.
(8):945–946. https://doi.org/10.1038/ https://doi.org/10.1172/JCI5183
nbt0805-945 10. Temme S, Jacoby C, Ding Z, Bönner F,
4. Zhong J, Mills PH, Hitchens TK, Ahrens Borg N, Schrader J, Flögel U (2014) Technical
(2013) Accelerated fluorine-19 MRI cell track- advance: monitoring the trafficking of neutro-
ing using compressed sensing. Magn Reson phil granulocytes and monocytes during the
Med 69(6):1683–1690. https://doi.org/10. course of tissue inflammation by noninvasive
1002/mrm.24414 19F MRI. J Leukoc Biol 95(4):689–697.
5. Krafft MP (2001) Fluorocarbons and fluori- https://doi.org/10.1189/jlb.0113032
nated amphiphiles in drug delivery and bio- 11. Flögel U, Su S, Kreideweiss I, Ding Z,
medical research. Adv Drug Deliv Rev 47 Galbarz L, Fu J, Jacoby C, Witzke O, Schrader
(2–3):209–228 J (2011) Noninvasive detection of graft rejec-
6. Jacoby C, Temme S, Mayenfels F, Benoit N, tion by in vivo (19) F MRI in the early stage.
Krafft MP, Schubert R, Schrader J, Flögel U Am J Transplant 11(2):235–244. https://doi.
(2014) Probing different perfluorocarbons for org/10.1111/j.1600-6143.2010.03372.x
in vivo inflammation imaging by 19F MRI: 12. Temme S, Bönner F, Schrader J, Flögel U
image reconstruction, biological half-lives and (2012) 19F magnetic resonance imaging of
sensitivity. NMR Biomed 27(3):261–271. endogenous macrophages in inflammation.
https://doi.org/10.1002/nbm.3059 Wiley Interdiscip Rev Nanomed Nanobiotech-
7. Molnar D, Linders J, Mayer C, Schubert R nol 4(3):329–343. https://doi.org/10.1002/
(2016) Insertion stability of poly(ethylene gly- wnan.1163
col)-cholesteryl-based lipid anchors in lipo- 13. Grenz A, Zhang H, Eckle T, Mittelbronn M,
some membranes. Eur J Pharm Biopharm Wehrmann M, Kohle C, Kloor D, Thompson
103:51–61. https://doi.org/10.1016/j.ejpb. LF, Osswald H, Eltzschig HK (2007) Protec-
2016.03.023 tive role of ecto-50 -nucleotidase (CD73) in
renal ischemia. J Am Soc Nephrol 18
19
F MRI 257
(3):833–845. https://doi.org/10.1681/ASN. 16. Robinson BR, Houng AK, Reed GL (2000)

2006101141 Catalytic life of activated factor XIII in
14. Ahrens, Zhong J (2013) In vivo MRI cell track- thrombi. Implications for fibrinolytic resistance
ing using perfluorocarbon probes and fluorine- and thrombus aging. Circulation 102
19 detection. NMR Biomed 26(7):860–871. (10):1151–1157
https://doi.org/10.1002/nbm.2948 17. Tung CH, Ho NH, Zeng Q, Tang Y, Jaffer FA,
15. Ahrens, Flores R, Xu H, Morel PA (2005) In Reed GL, Weissleder R (2003) Novel factor
vivo imaging platform for tracking immuno- XIII probes for blood coagulation imaging.
therapeutic cells. Nat Biotechnol 23 Chembiochem 4(9):897–899. https://doi.
(8):983–987. https://doi.org/10.1038/ org/10.1002/cbic.200300602
nbt1121
Chapter 15
Rodent Abdominal Adipose Tissue Imaging by MR

Bhanu Prakash KN, Jadegoud Yaligar, Sanjay K. Verma,
Venkatesh Gopalan, and S. Sendhil Velan
Abstract
Rodents including rats and mice are important models to study obesity, diabetes, and metabolic syndrome
in a preclinical setting. Translational and longitudinal imaging of these rodents permit investigation of
metabolic diseases and identification of imaging biomarkers suitable for clinical translation. Here we
describe the imaging protocols for achieving quantitative abdominal imaging in small animals followed
by segmentation and quantification of fat volumes.
Key words Magnetic Resonance Imaging, Abdomen, Rats, Mice, Segmentation, Visceral fat, Subcu-
taneous fat, Obesity, Quantification
1 Introduction
Obesity is identified as a global epidemic and chronic disease with

multiple comorbidities. In spite of several studies on obesity and its
relation to adverse health effects, the mechanisms of obesity and its
complications is not fully understood. Understanding of fat accumu-
lation/distribution mechanisms, composition/heterogeneity of adi-
pose tissue, functionalities of the adipose tissue, fat metabolism, etc.
are essential for designing interventions. Many studies have eluci-
dated the heterogeneity of adipose compartments and its metabo-
lism with associated health risks [1–4]. Accumulation of adipose
tissue around the abdominal organs like kidney, liver, pancreas, etc.
is called abdominal or central/ visceral obesity. Studies have shown
that among the fat compartments, visceral fat results in adverse
effects on insulin resistance, higher risk of cancer, decreased or
altered metabolism of lipid and glucose including other pathologies
[1, 2]. This in turn leads to higher mortality. Increased abdominal
obesity also elevates the associated risk factors. Hence, (i) to under-
stand obesity related anatomy and physiology changes and (ii) to
plan effective treatment; accurate identification and quantification of
adipose tissue is important. Preclinical imaging supports longitudinal
259
260 Bhanu Prakash KN et al.
assessment to understand the evolution of fat partitioning and trans-

lation of biomarkers to a clinical setting.
Preclinical imaging provides a pathway for evaluating metabolic
interventions and identification of imaging markers. Rodent mod-
els are utilized to investigate fat partitioning, evaluate effectiveness
of metabolic interventions on fat accumulation/mobilization, drug
discovery, study of metabolic pathways and other applica-
tions. Quantification of abdominal fat volumes is important for
studies related to obesity, nutrition and metabolic syndrome and
associated comorbidities. Several imaging methods have been uti-
lized to investigate abdominal adipose tissues [5–7]. Magnetic Res-
onance Imaging (MRI) is a noninvasive technique and permits
longitudinal assessment of abdominal fat volumes. Recent studies
have summarized different fat-imaging techniques using MR and
other modalities [8–10]. In this study protocol we will elaborate
the procedures for abdominal imaging in rodents using Dixon
based quantitative chemical shift imaging followed by segmenta-
tion of abdominal adipose tissues.
2 Materials
2.1 MR Equipment 1. In vivo imaging was performed using a 7 T MRI/MRS Bruker

ClinScan scanner (Bruker BioSpin GmbH, Ettlingen, Ger-
many) equipped with actively shielded gradient and shim coils
(maximum absolute gradient strength of 650 mT/m)
operating on Siemens VB 17 platform.
2. For high resolution abdominal imaging, a 72 mm diameter,
volume transmit/receive (Tx/Rx) coil was used.
2.2 Other Equipment 1. Respiration and body temperature was monitored using a phys-
iological monitoring system (ML880 16/30 power lab system,
AD Instruments, Spechbach, Germany).
2. Dedicated animal chamber for anesthetizing the animals.
3 Methods
Animal preparation and coil selection is crucial for adipose tissue MR

imaging. The following sections describe in detail various steps
involved in animal preparation, coil selection, and protocol for MR
imaging.
3.1 Animal 1. Anesthesia : Place the animal in the induction chamber and
Preparation anesthetize with 2–2.5% isoflurane mixed with medical grade
oxygen and air. Monitor the vital signs including breathing and
heart rate while the animal is anesthetized. Ensure that animal is
completely anesthetized before imaging. Maintain the isoflurane
MR Imaging of Adipose Tissue 261
level with 1.5–2% during the imaging. In addition to isoflurane

other anesthetics including Ketamine and Xyalzine can also be
used depending on the study and condition of animals. The
prescribed dosage of Ketamine and Xyalzine for mice is
~150 mg/kg, 10 mg/kg body weight and for rat ~75 mg/kg,
10 mg/kg body weight. For initial induction of mouse use
0.1 ml/10 g of body weight and maintain with 0.05 ml/
10 g/30 min. For rat, use 0.1 ml/10 g body weight for initial
induction followed by 0.05 ml/10 g/30 min for maintenance.
2. Animal Positioning: Transfer the anesthetized animal to appro-
priate (rat or mouse) animal holder which will be mounted in
the imaging instrument. Place the animal in supine position
with tooth bar locked by animal’s upper teeth. Tooth bar is a
hollow tube which supplies the isoflurane mixed with air and
oxygen to keep the animal anesthetized during the imaging
process. Position the centeral part of the abdominal area along
the z direction and at the isocenter of the magnet by using the
laser alignment (depending on the availability).
3. Physiological monitoring: Place the respiration pad below the
abdomen of the animal and secure it. Insert dedicated rectal
temperature probe into the rectum of the animal. Monitor the
respiration and rectal temperature of the animal by using a physi-
ological monitor. During the imaging process, maintain the
respiration rate of 60 and 90 cycles/min for rat and mouse,
respectively.
3.2 MR Coil Selection 1. The transmit/receive volume coils are more suitable for imag-
ing of abdomen. The dimensions of the coils can vary depend-
ing on the system and vendors. It is preferable to use a coil
which provides maximum filling factor to achieve good signal-
to-noise ratio along with superior spatial resolution.
3.3 In Vivo MR 1. Localizer: Rapid gradient echo imaging sequences can be uti-
Imaging lized to localize the abdominal area. Localizer will provide
multiple images in three orthogonal planes including axial
(transverse), sagittal, and coronal directions. Initially utilize a
larger field of view (FOV) of around 45–70 mm for rats and
relatively smaller FOV of 25–35 mm for mice. Following para-
meters can be utilized for a rat weighing around 250 g. FOV
70 mm, Repetition time (TR) 300–350 ms, Echo time (TE)
~4 ms, Flip angle (FA) 25 , matrix size 128 128, in plane
resolution 0.430 mm 0.430 mm, slice thickness
(ST) ~1 mm, slice mode- interleaved, number of averages
(NA) ~1, 15 slices in each imaging plane including axial, sagit-
tal, and coronal directions. Depending upon the animal size the
parameters can be optimized for maximum coverage and
resolution.
Fig. 1 Localization of abdominal compartment with the help of coronal T2 TSE imaging
2. Localizing the lumbar location: Anatomically the abdominal

cavity is located between lumbar 1 (L1) to lumbar 5 (L5).
Lumbars L1 to L5 can be utilized as two reference points for
transverse imaging of abdominal region (Fig. 1). To locate the
lumbar, perform additional T2-weighted Turbo spin echo
(TSE) based coronal imaging of the spine region by using
following parameters: Field of view 78 mm, TR 2800 ms, TE
24 ms, FA 180 , matrix size 192 192, resolution
0.406 mm 0.406 mm, slice thickness 1 mm, slice mode
interleaved, NA 1, imaging plane coronal, number of slices in
coronal plane 15, respiratory gating—external trigger, turbo
factor 8, echo trains per slice 24, echo spacing 7.9 ms. Choose
the phase encoding direction along the shortest anatomical
dimension to avoid fold-over artifacts. Care must be exercised
in selection of frequency and phase encoding directions to
avoid for motion, fold-over, blood flow artifacts during imag-
ing. These parameters can be optimized depending on the
animal size and instrument software and hardware.
3. Planning the imaging slices for Dixon imaging: Images
acquired by the procedure described above (T2-weighted cor-
onal images of the spine) will serve as reference for the precise
Fig. 2 In-phase, out-phase (top row), fat and water (bottom row) images acquired using 2point Dixon
sequence on a 7 T scanner
selection of the imaging slices in the transverse direction of the

abdomen region. Load the Dixon imaging sequence from the
acquisition protocol. Place a 3D slab having multiple slices
covering from L1 to L5. Note that few slices around L1 region
are more susceptible for respiratory motion. To ensure good
quality of the imaging slices, it is recommended to have two or
three additional slices beyond L1 and L5 regions. Depending
on the quality and structural information these additional slices
can be excluded from the analysis.
4. Fat–Water Imaging: Select a suitable TE on the scanner to get
in-phase and out-of-phase signals (see Note 1).
5. For the 7 T MRI scanner (Fig. 2), two point Dixon images were
acquired from the abdomen of a rat using a FOV-68 mm, ST
1.0 mm, slice mode- interleaved, TR 8 ms, TE 1 ms (in-phase),
2.5 ms (out-of-phase), FA 8 , NA 1, Slices per slab 28, matrix
size 256 256, resolution 0.266 mm 0.266 mm, bandwidth
(BW): 1090/1500, respiratory gating—external trigger with
phase encoding along left to right and frequency encoding
along anterior to posterior directions, respectively.
3.4 Post-processing In the following section we describe a general procedure for quan-
tification of subcutaneous fat (SAT) and visceral adipose tissue
(VAT). The SAT and VAT compartments are separated by the
abdominal muscular wall where the fat exterior to the abdominal
wall is referred as SAT, whereas the interior compartment to the
Fig. 3 Block diagram representation of overall post-processing for SAT and VAT quantification used in our
study
abdominal wall is VAT (see Note 2). Figure 3 shows the different
stages of processing and quantification.
1. Collect the data from the MR scanner in DICOM format. If
DICOM format is not available then use any software that
converts the format of images to DICOM or to Analyze/
Nifti format. In the proposed framework we use the DICOM
format data.
2. Combine the individual DICOM files into a volume data,
where rows represent X, and columns are Y and slices are
represented as Z.
3. Pre-processing is an important step of segmentation. In this
stage remove information like noise, background, gantry, tub-
ing, etc. which influence the quantification. In our study we
used anisotropic diffusion filter based on Perona–Malik equa-
tions to remove the noise and enhance the edges of the abdom-
inal wall [12].
4. Bias field correction to reduce the B0 inhomogeneity was per-
formed after filtering using biased fuzzy C-means
algorithm [13].
5. Dixon imaging provides fat, water, in-phase, and out-phase
images. In the fat volume only fat is highlighted and signals
coming from water are completely suppressed. The range of
intensities in each slice might be different due to inhomogene-
ity effects (B0, RF, receive coils, etc.). Normalize the intensity
range of every slice in the volume data.
Fig. 4 A sample slice based representation of results at different stages of SAT and VAT segmentation
6. If the edges are blurred due to contrast stretching and filtering

using a suitable edge enhancement method (e.g., Sobel,
Roberts, Canny, LOG) accentuate the edges [14].
7. Derive the binary mask by thresholding the edge-enhanced
image. Using the edge information of the binary image one
can automatically construct the initial contour of the geodesic
active contour [15]. Else one can manually place the initial
contour closer to the abdominal wall as shown in Fig. 4.
8. After generating the initial contour allow the active contour to
expand/contract to converge at the abdominal wall. This sepa-
rates the abdomen into SAT and VAT regions.
9. Perform fuzzy C-means clustering separately on the SAT and
VAT regions to obtain the fat tissues. In our study we used
3 class segmentation of SAT and 4 class segmentation on VAT.
Perform region merging of different classes based on the simi-
larity index between the classes.
10. After performing the segmentation of all the slices, clean the
segmented data for outliers.
11. Validate the results obtained by automatic/semi-automatic

technique with manual segmentation to calculate the accuracy,
precision, repeatability, and dice index [16] (see Note 3).
4 Notes
1. The chemical shift of water (~4.7 ppm) and the n-methylene

protons of the fat signal (~1.2 ppm) are well separated in most
of the field strengths utilized for small animal imaging. Fat/-
water imaging can be performed with different MRI methods
including inversion recovery based approaches, selective exci-
tation of fat or water, or chemical shift imaging. Dixon imaging
utilizes the chemical shift difference of water and n-methylene
protons. Having different resonance frequencies, the water and
fat spins are at in- and out-of-phase with a period of 1/Δf,
where Δf is the frequency offset which can be calculated for a
given field strength B0 using the chemical shift Δσ, γ the
gyromagnetic ratio of protons as Δf ¼ B02π γ Δσ
. One can derive
the fat and water images from the prior knowledge of fat/water
frequency shifts for different echo times. Various algorithms are
available which exploit the reconstruction of fat and water from
two echoes, multiple echoes with either in-phase, out-of-phase
spaced echo times or variables echo times. The most commonly
utilized two point Dixon method produces water-only and
fat-only images using a dual-echo acquisition. Using three or
larger number of echoes improves the accuracy to separate the
water and fat and can also be utilized to derive additional
information on relaxation (ex. T2* map for water and fat, B0
inhomogeneity map). For a given magnetic field strength of
7 Tesla (T) the in- and out-of-phase echo times can be calcu-
lated as follows. The frequency shift difference between fat and
water is 4.7–1.2 ¼ 3.5 ppm and at 7 T, with proton resonance
frequency at 300 MHz results in a frequency separation Δf
¼ 300∗ 3.5 ¼ 1050 Hz. Hence for every Δt ¼ 1/Δf ¼ 1/
1050 Hz ¼ 0.9524 ms the fat and water will be either in-phase
or out-of-phase. Using this formula the echo times for in-phase
and out-of-phase can be estimated for any given field strengths,
i.e., in 7 T for every TE ¼ 0.9524 * (2n) ms we get the in-phase
signal of fat and water while at TE ¼ 0.9524 * (2n þ 1) we get
the out-of-phase signal, where n ¼ 0, 1, 2, 3,.... Similarly for
9.4 T the system frequency is 400 MHz and for 11.7 T it is
500 MHz. For 9.4 T the values of Δf ¼ 400∗3.5 ¼ 1400 Hz
and Δt ¼ 1/Δf ¼ 1/1400 Hz ¼ 0.7142 ms. To calculate the
TE of in-phase and out-of-phase signals we use 0.7142 * (2n)
and 0.7142 * (2n þ 1), respectively [11].
2. The abdominal fat can be broadly classified as SAT and intra-

abdominal fat (IAT), where IAT further can have multiple
compartments including intraperitoneal and retroperitoneal
compartments. The intraperitoneal fat can be further classified
as omental and mesenteric compartments, whereas the retro-
peritoneal is made up of gonadal, pararenal, perirenal, and
other compartments. Few publications also refer IAT as VAT.
The intraperitoneal fat is about 67% of VAT and retroperitoneal
contributes to the remaining 33% of VAT.
3. Quantitation of SAT/VAT volumes is essential for understand-
ing fat accumulation/mobilization, etc. during obesity and
anti-obesity interventions.
4. ISMRM fat-water toolbox can be used to compute the fat and
water images [17].
5. There are different methods available for segmenting the SAT
and VAT regions like thresholding, region growing, fuzzy clus-
tering, graph cut, watershed [18–23].
6. The algorithms used in different methods have their own
advantages and limitations. In this work, we have illustrated
the fuzzy C-means based algorithm for quantifying abdominal
fat volumes in rodents.
References
1. Brochu M, Poehlman, Ades PA (2000) Obe- Horiguchi T, Nishimoto T, Velan SS (2016)
sity, body fat distribution, and coronary artery Effect of exercise and calorie restriction on tissue
disease. J Cardpulm Rehabil 20(2):96–108 acylcarnitines, tissue desaturase indices, and fat
2. Colberg SR, Simoneau JA, Thaete FL, Kelley accumulation in diet-induced obese rats. Sci Rep
DE (1995) Skeletal muscle utilization of free 6:26445. https://doi.org/10.1038/srep26445
fatty acids in women with visceral obesity. J 7. Sadananthan SA, Prakash B, Leow MK, Khoo
Clin Invest 95(4):1846–1853. https://doi. CM, Chou H, Venkataraman K, Khoo EY, Lee
org/10.1172/JCI117864 YS, Gluckman PD, Tai ES, Velan SS (2015)
3. Evans DJ, Hoffmann RG, Kalkhoff RK, Kisse- Automated segmentation of visceral and sub-
bah AH (1984) Relationship of body fat cutaneous (deep and superficial) adipose tissues
topography to insulin sensitivity and metabolic in normal and overweight men. J Magn Reson
profiles in premenopausal women. Metab Clin Imaging 41(4):924–934. https://doi.org/10.
Exp 33(1):68–75 1002/jmri.24655
4. Vague J (1956) The degree of masculine differ- 8. Marzola P, Boschi F, Moneta F, Sbarbati A,
entiation of obesities: a factor determining pre- Zancanaro C (2016) Preclinical in vivo imaging
disposition to diabetes, atherosclerosis, gout, for fat tissue identification, quantification, and
and uric calculous disease. Am J Clin Nutr 4 functional characterization. Front Pharmacol
(1):20–34 7:336. https://doi.org/10.3389/fphar.2016.
5. Kn BP, Gopalan V, Lee SS, Velan SS (2014) 00336
Quantification of abdominal fat depots in rats 9. HH H, Kan HE (2013) Quantitative proton
and mice during obesity and weight loss inter- MR techniques for measuring fat. NMR
ventions. PLoS One 9(10):e108979. https:// Biomed 26(12):1609–1629. https://doi.org/
doi.org/10.1371/journal.pone.0108979 10.1002/nbm.3025
6. Gopalan V, Michael N, Ishino S, Lee SS, Yang 10. Schick F (1998) Simultaneous highly selective
AY, Bhanu Prakash KN, Yaligar J, Sadananthan MR water and fat imaging using a simple new
SA, Kaneko M, Zhou Z, Satomi Y, type of spectral-spatial excitation. Magn Reson
Hirayama M, Kamiguchi H, Zhu B, Med 40(2):194–202
11. Reeder SB, McKenzie CA, Pineda AR, Yu H, of volumetric and single-slice MRI adipose
Shimakawa A, Brau AC, Hargreaves BA, Gold analysis using a novel fully automated segmen-
GE, Brittain JH (2007) Water-fat separation tation method. J Magn Reson Imaging 41
with IDEAL gradient-echo imaging. J Magn (1):233–241. https://doi.org/10.1002/jmri.
Reson Imaging 25(3):644–652. https://doi. 24526
org/10.1002/jmri.20831 19. Joshi AA, HH H, Leahy RM, Goran MI, Nayak
12. Pernona P, Malik J (1990) Scale-space and edge KS (2013) Automatic intra-subject registra-
detection using anisotropic diffusion. IEEE Trans tion-based segmentation of abdominal fat
Pattern Anal Mach Intell 12(7):629–639. from water-fat MRI. J Magn Reson Imaging
https://doi.org/10.1109/34.56205 37(2):423–430. https://doi.org/10.1002/
13. Ahmed MN, Yamany SM, Mohamed N, Farag jmri.23813
AA, Moriarty T (2002) A modified fuzzy 20. Ranefall P, Bidar AW, Hockings PD (2009)
C-means algorithm for bias field estimation Automatic segmentation of intra-abdominal
and segmentation of MRI data. IEEE Trans and subcutaneous adipose tissue in 3D whole
Med Imaging 21(3):193–199. https://doi. mouse MRI. J Magn Reson Imaging 30
org/10.1109/42.996338 (3):554–560. https://doi.org/10.1002/jmri.
14. Peli T, Malah D (1982) A study of edge detec- 21874
tion algorithms. Comput Graph Image Process 21. Shen J, Baum T, Cordes C, Ott B, Skurk T,
20(1):1–21. https://doi.org/10.1016/0146- Kooijman H, Rummeny EJ, Hauner H, Menze
664X(82)90070-3 BH, Karampinos DC (2016) Automatic seg-
15. Lankton S, Tannenbaum A (2008) Localizing mentation of abdominal organs and adipose
region-based active contours. IEEE Trans tissue compartments in water-fat MRI: applica-
Image Process 17(11):2029–2039. https:// tion to weight-loss in obesity. Eur J Radiol 85
doi.org/10.1109/TIP.2008.2004611 (9):1613–1621. https://doi.org/10.1016/j.
16. Zou KH, Warfield SK, Bharatha A, Tempany ejrad.2016.06.006
CM, Kaus MR, Haker SJ, Wells WM 3rd, 22. Tang Y, Sharma P, Nelson MD, Simerly R,
Jolesz FA, Kikinis R (2004) Statistical valida- Moats RA (2011) Automatic abdominal fat
tion of image segmentation quality based on a assessment in obese mice using a segmental
spatial overlap index. Acad Radiol 11 shape model. J Magn Reson Imaging 34
(2):178–189 (4):866–873. https://doi.org/10.1002/jmri.
17. HH H, Bornert P, Hernando D, Kellman P, 22690
Ma J, Reeder S, Sirlin C (2012) ISMRM work- 23. Wald D, Teucher B, Dinkel J, Kaaks R,
shop on fat-water separation: insights, applica- Delorme S, Boeing H, Seidensaal K, Meinzer
tions and progress in MRI. Magn Reson Med HP, Heimann T (2012) Automatic quantifica-
68(2):378–388. https://doi.org/10.1002/ tion of subcutaneous and visceral adipose tissue
mrm.24369 from whole-body magnetic resonance images
18. Addeman BT, Kutty S, Perkins TG, Soliman suitable for large cohort studies. J Magn Reson
AS, Wiens CN, McCurdy CM, Beaton MD, Imaging 36(6):1421–1434. https://doi.org/
Hegele RA, McKenzie CA (2015) Validation 10.1002/jmri.23775
Chapter 16
Cardiac MRI in Small Animals

Min-Chi Ku, Till Huelnhagen, Thoralf Niendorf, and Andreas Pohlmann
Abstract
Cardiac magnetic resonance (MR) imaging of mice is a valuable tool for the precise in vivo diagnosis and
prognosis of heart defects. This detailed protocol describes the method of cardiac MR imaging in mice step
by step. A series of MR images captures the contractile function of the mouse heart and post-processing of
the image data yields morphometric parameters (myocardial mass, myocardial wall thickness, ventricular
end-systolic and end-diastolic volume) as well as functional parameters (stroke volume and ejection
fraction). This protocol may also serve as a starting point for MR imaging of rats, by using larger image
dimensions (field-of-view) and MR hardware suitable for larger animals.
Key words Magnetic resonance imaging (MRI), Cardiac magnetic resonance imaging (CMR),
Mouse, Heart, Function
1 Introduction
Since decades, cardiovascular diseases (CVDs) remain the global

number one cause of death (WHO 2016). Because of the complex-
ity of CVDs, the use of animal models in cardiovascular research
may help to discover the underlying molecular mechanisms. Fur-
ther development of novel therapies requires testing of the putative
therapeutic strategies in appropriate animal models such as mouse
models for human CVDs [1, 2].
Characterization, phenotyping, and evaluation of established
mouse models (genetically modified, surgically or pharmacologi-
cally induced) are valuable but still challenging in the longitudinal
manner. Noninvasive, in vivo monitoring of the rapidly moving
mouse heart is especially difficult. Cardiac magnetic resonance
imaging (CMR) provides excellent spatial and temporal resolution
and is suitable for longitudinal studies. CMR methods have been
developed not only for determining cardiac morphology, but also
for functional assessments as well as measuring the extent of myo-
cardial injury, remodeling [3, 4], perfusion [5], and edema
[6, 7]. CMR applications in both preclinical and clinical settings
269
270 Min-Chi Ku et al.
enable transferring knowledge gained in preclinical setups to be

translated from bench to the clinical bedside.
Here we describe the most frequently used CMR technique
and functional heart assessment for monitoring heart function in
mice [8, 9]. For applications that demand particularly high image
quality two optional methodological improvements are provided.
The first option addresses increasing the signal-to-noise ratio
(SNR), which in turn can be used to enhance spatial/temporal
resolution or reduce scan durations. Signal averaging is the simplest
and most commonly used approach for improving SNR, but comes
at the cost of prolonged measurement times. An alternative is to use
a cryogenic radiofrequency (RF) coil [10], which reduces the elec-
tronic background noise. Significant gains in SNR (approx. factor
2.5–4.0) for MRI of the mouse heart with a cryogenic RF coil have
been shown in comparison with a dedicated mouse heart RF coil
array operating at room temperature [9]. The second option
addresses image artifacts, caused by the high velocity blood flow.
These can be largely reduced by employing an imaging technique
called ultra-short echo time (UTE) pulse sequence [11, 12].
2 Materials
2.1 Animals Mice are housed under standard conditions according to the animal
welfare guidelines and regulations. Here we describe cardiac MRI
of C57BL/6J mice with a body weight of 20–30 g. Other strains
and ages may require some adaptations of the experimental setup
and protocol.
2.2 Preparation 1. Anesthesia: Mice are anesthetized by an initial inhalation nar-

of Animals cosis with 3% isoflurane in an oxygen/air mixture (2:1) with a
for Magnetic flow rate of 500 mL/min for induction for about 3 min and
Resonance Imaging then maintained at 0.5–1.5% isoflurane during MRI.
(MRI) 2. Vital sign control: Core body temperature should be main-
tained at 37 C with a small animal electrical warming pad
while in preparation. Respiration rate and temperature can be
monitored using a remote monitoring system (Model 1025,
SA Instruments Inc., Stony Brook, NY, USA).
3. Temperature measurement: An interferometric measurement
system (ACS-P4-N-62SC, Opsens, Quebec City, Canada),
including a fiber-optical temperature probe (OTP-M, Accu-
Sens, Opsens), is used (see Note 1).
2.3 Magnetic An MRI system including suitable accessories for the MR acquisi-
Resonance Imaging tion such as radio frequency (RF) antennas; equipment for animal
positioning, inhalation anesthesia, warming, and monitoring
Cardiac MRI 271
physiological parameters of the animals; and trained personnel for

operating the MRI system.
1. MRI system: A dedicated small animal MR system with a
magnetic field strength of 7.0 Tesla or higher (e.g., 7.0 T,
9.4 T, 11.7 T) is recommended. Here we describe the use of
a 9.4 T 20 cm bore system (Biospec 94/20, Bruker Biospin,
Ettlingen, Germany) equipped with a gradient system
integrated with shim set (B-GA12S2, Bruker Biospin, Ettlin-
gen, Germany; gradient amplitude 440 mT/m, max. slew rate
3440 T/m/s). This protocol describes cardiac MR using the
ParaVision 6.0 software by Bruker Biospin (software version
5.1 should suffice) and the self-gated acquisition method
Intragate FLASH, which requires a separate license.
2. Radio frequency (RF) coils: Use RF coils suitable for mouse
heart imaging, such as a receive-only mouse heart RF coil array
(e.g., 4-element surface coil; model: T11426_V3, Bruker Bios-
pin, Ettlingen, Germany) combined with a volume resonator
for RF transmission (model: T9361_V3, Bruker Biospin,
Ettlingen, Germany). Alternative for improved image qual-
ity/speed (“Option CP”): For high spatially/temporally
resolved imaging the use of a cryogenically cooled RF coil
(Mouse CryoProbe, Bruker Biospin, Ettlingen, Germany) is
recommended [9, 10] (see Note 2).
3. Gases: O2 and compressed air, as well as a gas-mixing system
(FMI Föhr Medical Instruments GmbH, Seeheim-Ober Beer-
bach, Germany).
4. Device for warming of animal while scanning: Use a circulating
warm-water based heating system, consisting of a plastic cover
or rubber mat with integrated tubing connected to a conven-
tional warm water bath. For alternative coil setups (e.g., Mouse
CryoProbe), water pipes may be built into the animal holder.
5. Monitoring of physiological parameters: For monitoring of
respiration and core body temperature throughout the entire
MR experiment, use a small animal monitoring system (Model
1025, Small Animal Instruments, Inc., Stoney Brook, NY,
USA), including a rectal temperature probe and pneumatic
pillow. Maintain body temperature at 37 C by water heater/
circulator set up to 42–45 C (see Note 3).
6. Data analysis: Heart function analysis of the MR data requires a
personal computer and software that permits manual segmen-
tation of the myocardium and ventricles. This can be achieved
by programming a tool in MATLAB (The Mathworks, Natick,
MA, USA) or ImageJ (Rasband, W.S., ImageJ, U.S. National
Institutes of Health, Bethesda, Maryland, USA, http://imagej.
nih.gov/ij/). Alternatively, there are commercial software tools
such QMass (Medis Leiden, The Netherlands) and CMR42

(Circle Cardiovascular Imaging Inc., Calgary, Canada). Here
we used CMR42.
3 Methods
3.1 Preparation 1. Start the MR scanner’s ParaVision software and—only once

of MRI prior to the first experiment—create and store the following
MR protocols:
(a) Protocol-1_Localizer (pilot scan for checking the mouse
heart position): FLASH sequence with 1 slice in each
direction (axial, coronal, and sagittal), repetition time
(TR) ¼ 14 ms, effective echo time (TE) ¼ 2.1 ms, flip
angle (FA) ¼ 30 , field of view (FOV) ¼ 40 40 mm2,
matrix size ¼ 88 128, zero-filled to 256 256 using a
partial Fourier acceleration of 1.68 in read direction,
16 averages, motion averaging “on,” scan time ¼ 28 s.
(b) Protocol-2_Multislice_Localizer: FLASH sequence,
TR ¼ 110 ms, TE ¼ 2.5 ms, FOV ¼ 35 35 mm2, matrix
size ¼ 88 128, zero-filled to 256 256 using a partial
Fourier acceleration of 1.68 in read direction, 5 slices with
a thickness of 0.7 mm, 7 averages, scan time ¼ 1 min 38 s.
(c) Protocol-3_Coronal_Localizer: FLASH sequence,
TR ¼ 28 ms, TE ¼ 2.09 ms, the geometry defined as an
axial FOV ¼ 25 25 mm2, matrix size ¼ 88 128, zero-
filled to 256 256 using a partial Fourier acceleration of
1.68 in read direction, three slices with a thickness of
0.6 mm, 15 averages, scan time ¼ 53 s.
(d) Protocol-4_Sagittal_LAX_Ig (for adjusting scan position
in a heart long axis view): Intragate (Ig) FLASH sequence,
TR ¼ 5.3 ms, TE ¼ 2.41 ms, FOV ¼ 35 35 mm2,
matrix size ¼ 96 128, zero-filled to 256 256 using a
partial Fourier acceleration of 1.6 in read direction, thick-
ness ¼ 1 mm, oversampling factor ¼ 60, scan time ¼ 40 s,
navigator slice of 5 mm thickness with “in-slice” geometry
(same slice orientation as imaging slice) (see Note 4).
(e) Protocol-5_Coronal_LAX_Ig (for adjusting scan position
in a heart long axis view): Ig FLASH sequence (see Note
4), TR ¼ 5.3 ms, effective echo time (TE) ¼ 2.41 ms,
FOV ¼ 30 30 mm2, matrix size ¼ 96 96, zero-filled
to 256 192 using a partial Fourier acceleration of 1.6 in
read direction, thickness ¼ 1 mm, oversampling fac-
tor ¼ 60, scan time ¼ 30 s, navigator slice of 5 mm
thickness with “in-slice” geometry.
Cardiac MRI 273
(f) Protocol-6_SAX_CINE_Ig: Ig FLASH sequence (see

Note 4), TR ¼ 8.5 ms, TE ¼ 1.58 ms, FA ¼ 20 ,
FOV ¼ 11 22 mm2, matrix size ¼ 109 192, zero-
1.55 in read direction, movie frame ¼ 16, slice thick-
ness ¼ 0.8 mm, oversampling factor ¼ 400 for room
temperature RF coil and 100 for cryogenic probe, scan
time ¼ 10 min 52 s for room temperature RF coil and
2 min 43 s for cryogenic RF coil, navigator slice of 2 mm
thickness with FA ¼ 2 , perpendicular to imaging slice.
(g) Alternative scan protocol that helps to reduce flow arti-
facts (“Option UTE”): Protocol_SAX_UTE_Ig: Ig-UTE
sequence (see Note 5), TR ¼ 9 ms, TE ¼ 0.4 ms,
FOV ¼ 25 25 mm2, matrix size ¼ 256 256, read
out matrix ¼ 256 256, projections ¼ 804, movie
frames ¼ 14, thickness ¼ 0.8 mm, oversampling fac-
tor ¼ 80, scan time ¼ 9 min 38 s, navigator slice of
2 mm thickness with FA ¼ 2 , which is parallel to imaging
slice.
(h) Protocol-7_LAX_CINE_Ig: Ig FLASH sequence (see
Note 4), TR ¼ 8.5 ms, TE ¼ 1.58 ms, FA ¼ 20 ,
FOV ¼ 11 22 mm2, matrix size ¼ 109 192, zero-
1.55 in read direction, movie frames ¼ 16, slice thick-
ness ¼ 0.8 mm, oversampling factor for cryogenic RF
coil ¼ 150, scan time ¼ 4 min 4 s, navigator slice of
2 mm thickness with FA ¼ 2 , in “arbitrary” orientation.
2. Install the RF coil(s) in the MR scanner and/or animal bed. For
the described setup: place the volume resonator at the center of
the magnet and fix the surface RF coil array on the animal bed
with tape.
3. Switch on the gradient amplifiers of the MR system. This will
also switch on the automatic animal positioning system
AutoPac.
4. Connect the animal holder to the animal positioning system
(AutoPac).
5. Attach the facemask unit (either commercial or custom-made)
to the animal holder and connect it to the inspiratory gas
providing system (luer tubing).
6. Place the plastic mat (warm-water based heating system) on top
of the mouse and then connect it to the warm-water circulation
system (water bath).
7. Switch on the water bath. Adjust the water bath temperature to
approximately 42–45 C (see Note 3).
8. Attach the rectal temperature probe and pneumatic pillow to

the small animal monitoring system and place the probes on the
animal bed, approximately at the lower abdominal position of
the mouse. Attach all tubes and cables along the length of the
animal bed, fixing it with masking (or autoclave marker) tape.
3.2 Animal 1. Switch on the small animal monitoring system.

Transportation, 2. Carefully transfer the animal into the MR scanner room and
Fixation, Positioning, anesthetize it in an appropriate induction box.
Masking
3. Place the anesthetized mouse on the MRI animal bed and
move the mouse so that the heart is positioned at the center
of the mouse heart RF coil.
4. Place a respiratory mask around the muzzle of the spontane-
ously breathing mouse.
5. Insert the rectal temperature probe after cleaning with alcohol
and coating with Vaseline™.
6. Cover the animal with the warming plastic mat. Watch the
respiration trance on the monitor of the small animal monitor-
ing system. Press the “Out” button of the AutoPac system to
make sure the animal bed is in the reference position. Switch on
the Laser position marker and drive the animal bed until the
anatomical region of interest (in this case the heart) is aligned
with the laser position. Switch off the laser (this point is now
stored as the point of interest).
7. Double check the entire animal bed to make sure nothing
protrudes.
8. Press the Work Position button of the AutoPac system to drive
the point of interest (point on the animal bed marked by the
Laser) to the isocenter of the magnet.
3.3 MRI Pre-scans 1. Register a new subject and study. For the first scan load the
Protocol-1_Localizer.
2. Tune and match the volume resonator using the Wobble
function.
3. Start the 1_Localizer scan and verify on the acquired images
that the heart is close to the isocenter of the magnet and well
within the field of view. If the heart is too far away from the
center of the magnet and the RF coil (heart not visible or in
region where signal intensity starts to tail off), it may be neces-
sary correct the animal or/and Laser marker position. In that
case, repeat the tuning/matching and this pilot scan.
4. Load the Protocol-2_Multislice_Localizer. Using the 1_Loca-
lizer scan as the reference edit the geometry of the 2_Multi-
slice_Localizer such that the slices packages are centered
around the heart (Fig. 1) and run the scan.
Cardiac MRI 275
Fig. 1 Image from 1_Localizer scans displaying the mouse thorax and upper abdomen in axial, sagittal, and
coronal views. The overlaid slice contours show how to position the slice packages for 2_Multislice_Localizer
centered on the heart
Fig. 2 (a) Representative slice for axial view of the thorax provided by the 2_Multislice_Localizer. The overlaid
slice contours show how to position the slice packages for running 3_Coronal_Localizer. The slice package is
centered on the heart. (b) Representative slice for coronal view of the thorax provided by the 3_Coronal_Lo-
calizer and the overlaid yellow box shows slice contour for running the 4_Sagittal_LAX_Ig
5. Load the Protocol-3_Coronal_Localizer. Use the 2_Multisli-

ce_Localizer scan as the reference and pick the slice that shows
the heart best. Edit the geometry of the 3_Coronal_Localizer
to position the slice package in the center of the heart (axial
view, Fig. 2a) and run the scan.
6. Load the Protocol-4_Sagittal_LAX_Ig. Pick the image with
best slice position (center slice of the heart in coronal view)
from the 3_Coronal_Localizer and use it as the reference to
plan the scan. Adjust the geometry of the 4_Sagittal_LAX_Ig
scan such that the slice is perpendicular to the reference image
and centered along the long axis (LAX) of the left ventricle of
the heart (Fig. 2b). This protocol is based on a self-gated
(Intragate) imaging method (see Note 4). It requires providing
an estimate of the current heart rate (HR) and respiration rate
Fig. 3 (a) Representative image of the mouse heart obtained from the 4_Sagit-
tal_LAX_Ig scan with overlaid slice contour of the 5_Coronal_LAX_Ig illustrating
how to plan the geometry. (b) Representative image obtained from 5_Coronal_-
LAX_Ig scan
(RR) of the animal just before starting the scan. If the initial
guess for nominal HR and RR is too far off then the images
may have severe artifacts. In this case the scan must be repeated
with different nominal HR and RR. After each scan the Intra-
gate method provides both HR and RR values that are derived
from the MR data (a so-called navigator) which can be found in
the reconstruction panel and can help to adjust the nominal
HR and RR to improve image quality. Set the nominal HR and
RR, then run the scan.
7. Load the Protocol-5_Coronal_LAX_Ig. Use the 4_Sagittal_-
LAX_Ig as reference image and browse through the CINE
images of the different heart phases to select the one depicting
diastole (maximum dilation of the left ventricle). Position the
slice from the 5_Coronal_LAX_Ig scans such that the slice is
perpendicular to the reference image and centered along the
long axis of the left ventricle of the heart (Fig. 3a). Set the HR
and RR, then run the scan. Check the scan result and pick the
diastolic image from the coronal view (Fig. 3b).
8. Load the Protocol-6_SAX_CINE_Ig: Use the image from
5_Coronal_LAX_Ig as reference and browse to the CINE
movie frame of diastole. Align the navigator slice (blue box in
Fig. 4a) (see Note 4) along the long axis of the left ventricle and
perpendicular to the reference image. Adjust the slice geometry
of the 6_SAX_CINE_Ig scan to be perpendicular to the heart
long axis and the reference image (yellow box in Fig. 4a); move
the slice to the middle of the heart. Set the HR and RR, then
run the scan, which serves as a test scan for the following series
of short axis (SAX) images. Check the obtained images for all
16 heart phases (Fig. 4b) and pick the diastolic phase (here is
phase 8; also see Fig. 5a).
Cardiac MRI 277
Fig. 4 (a) Representative slice for long axis (LAX) view obtained from the 5_Coronal_LAX_Ig scan. Overlaid is
the navigator slice outline (blue box) and the short axis image slice outline (yellow box) of the 6_SAX_CINE_Ig
scan. (b) 16 phases of cardiac short axis (SAX) views from a test scan of the 6_SAX_CINE_Ig scan
Fig. 5 (a) A representative slice for SAX view in diastole from 6_SAX_CINE_Ig scan with overlaid slice outline
(yellow box) of the second, improved LAX (5_Coronal_LAX_Ig scan). (b) A representative slice for improved
LAX view obtained with Protocol-5_Coronal_LAX_Ig. LV left ventricle, RV right ventricle
9. For an improved LAX view, duplicate the previous Protocol-

5_Coronal_LAX_Ig, use the diastolic movie frame of the
6_SAX_CINE_Ig as reference, and orientate the slice position
as illustrated in Fig. 5a (yellow box). Adjust the angle of the
slice to make it perpendicular to the heart’s short axis, set the
HR and RR, then run the scan. Check the scan result and pick
the diastolic CINE movie frame of the LAX images (Fig. 5b).
3.4 Heart Short Axis 1. Duplicate the Protocol-6_SAX_CINE_Ig scan. Use the
View Images improved LAX images (Fig. 5b) as the scan reference and pick
the diastolic CINE frame to fine adjust the geometry of the
Fig. 6 (a) Slice positioning of the navigator slice (blue box) and 9 individual SAX CINE image slices (yellow
boxes) for 6_SAX_CINE_Ig based on the improved LAX view of the mouse heart. Adjust the slice angles
according to the heart orientation. (b, c) Stacks of 9 SAX image slices that represent the main result of the MRI
examination used for heart function assessment. (b) Mouse heart SAX images obtained with the conventional
room temperature RF coil (RT). (c) SAX images of another mouse provided by the cryogenic RF coil (CP). Please
note that the scan time by using the cryogenic RF coil is reduced to 1/4 compared to the room temperature RF
coil (see Note 2). Numbers 1–9 in yellow indicate the slice positions shown in Fig.6a
image slice and navigator slice. Shift the image slice to the
heart’s apex (slice 1 in Fig. 6a). Set the HR and RR, then run
the scan. Duplicate the scan protocol and place the second slice
1.1 mm above the slice 1. Repeat this procedure until covering
the entire heart, in particular the left ventricle (create nine slices
in this case) (Fig. 6a). These images represent the main result of
the MRI examination, which will be then used for heart func-
tion assessment.
3.5 Heart Long Axis Load the Protocol-7_LAX_CINE_Ig. Pick the diastolic slice from
View Images (2–4- a short axis view (the scan result from Protocol-6_SAX_CINE_Ig)
Chambers) and place the slice position (similar with the slice position shown in
Fig. 5a); adjust the angle of the slice to make it perpendicular to the
heart short axis. Set the HR and RR, then run the scan. Fig. 7
shows the scan result of Protocol-7. Please note that depending on
the angle of the slice orientation, different views of the heart’s long
axis might present.
3.6 End 1. Carefully remove the mouse from the mouse holder.
of the Experiment 2. Move the animal into the preparation room and place in a
supine position on a pre-warmed mat to recover from the
anesthesia.
3.7 Data Analysis 1. On the scanner console export all images in DICOM (Digital
Imaging and Communications in Medicine) format.
2. The analysis software CMR42 requires the LAX and all SAX
images to have the same DICOM frame of reference UID (user
Cardiac MRI 279
Fig. 7 Scan result from Protocol-7_LAX_CINE_Ig
identification) (not the case for individual LAX/SAX scans).

Edit the DICOM header of these scans: copy the frame of
reference UID (DICOM Tag (0020,0052)) of the first SAX
slice to the other images and store the result, using a program
written in MATLAB or a suitable DICOM viewer/editor (see
Note 6).
3. Import the DICOM data set (LAX and SAX slice package) in
CMR42. Select all the SAX slices covering the heart (slices 1–9
as previously mentioned) for assessment of cardiac function.
4. For left ventricular function assessment, manually outline the
endo-cardial borders (red circle as shown in upper panel,
Fig. 8) and epi-cardial borders (green circle in upper panel,
Fig. 8) for each slice.
5. To assess the right ventricular function, manually outline the
endo- (yellow circle, Fig. 9) and epi-cardial borders (green
circle, Fig. 9). To assess the left ventricular function, manually
outline the endo- (red circle, Fig. 9) and epi-cardial borders
(blue circle, Fig. 9). CMR42 will then automatically calculate
the myocardial mass, myocardial wall thickness, stroke volume
(SV), ventricular end-systolic volume (ESV), end-diastolic vol-
ume (EDV), and ejection fraction (EF) based on the depicted
contour areas from either right or left ventricles. Note here that
when defining the contour for the left ventricular endo-cardial
border, the papillary muscles haves to be excluded (Fig. 9).
4 Notes
1. The temperature probes are rather fragile and you may want to
consider using protectively coated GaAs probes instead. In that
case be aware that a GaAs crystal based system shows a temper-
ature offset caused by the magnetic field (approx. 4.7 C at
9.4 T).
Fig. 8 Illustration of image analysis in CMR42. There are shown the SAX images of 3 slices for 10 cardiac
phases, including systole (yellow box) and diastole (green box). Endo- (red circle) and epi-cardial (green circle)
borders in each slice have to be outlined manually
Fig. 9 Illustration of outlining contour for endo- and epi-cardial borders for both
LV and RV. White dashed arrows indicate papillary muscles which has to be
excluded in contour
2. A cryogenically cooled RF coil is substantially more expensive

than a common RF coil that operates at room temperature. If
the MRI system used is not already equipped with cryogenically
cooled RF coil, but the planned study requires CMR with very
high temporal or spatial resolution, you might consider
Cardiac MRI 281
alternative means of increasing the speed (acceleration methods

such as parallel imaging and compressed sensing) and increas-
ing SNR (spatially adaptive non-local mean filter (SANLM,
VBM8 Toolbox, SPM, http://www.fil.ion.ucl.ac.uk/spm/
and http://dbm.neuro.uni-jena.de/vbm8/) or simply more
averaging.
3. The temperature of the water will be much higher than the
temperature of the rubber mat and depends on the length and
material of the tubing used. Hence the temperature of the bath
must be adapted to the local setup.
4. Intragate FLASH is a so-called self-gated CINE imaging
method. CINE indicates that the method permits capturing
the motion of the heart in a movie with each movie frame
representing a phase of the heart cycle. However, CMR is not
a real-time imaging technique like echocardiography (ultra-
sound). The data needed to create CINE MR images must be
collected over many heart cycles (phases). This requires a syn-
chronization of the MR data acquisition with the heart motion.
Clinical CMR commonly employs an electrocardiogram
(ECG) to track heart motion. In mice, and especially at ultra-
high magnetic fields of 9.4 Tesla and more, the use of an ECG
is rather challenging. The ECG recording is degraded by MRI
interferences (magnetic gradients, RF pulses) and by the
magneto-hydrodynamic effect (motion of the blood in the
magnetic field produces an electric current). This can result in
unreliable detection of the R-wave and hence poor synchroni-
zation of the MR data acquisition with the heart motion. A very
good alternative to ECG-gating are self-gated imaging meth-
ods. Such techniques avoid the challenging use of an ECG and
hence reduce the animal experiment duration markedly. Self-
gated techniques might be more susceptible to arrhythmia or
variable heart rates than ECG-gated acquisitions. The Intragate
FLASH, described in this protocol, is a self-gated method. It is
part of the cardiac MR pulse sequence package available for
Bruker MRI systems, and must be purchased separately.
5. For assessment of cardiac function with CMR, images recorded
with self-gating strategies may lead to pronounced flow arti-
facts and susceptibility effects at the myocardium-lung inter-
face. Flow artifacts may compromise the measurement of
chamber volumes or proper visualization of fine structures in
the heart. Minimizing TE helps to overcome these problems.
One acquisition technique that achieves this is ultra-short echo
time (UTE). Intragate UTE is a self-gated ultra-short echo
time imaging method (Fig. 10) tailored for CMR or rodents
[11]. It is available for Bruker MRI systems, and must be
purchased separately. NB: When using UTE sequences a careful
calibration, the so-called trajectory measurement, is crucial for
Fig. 10 (a, b) SAX images acquired with the cryogenically cooled RF coil (a) and with the room temperature RF
coil (b), using the conventional Ig-FLASH protocol. Flow artifacts are highlighted by red arrows. (c) In contrast,
the SAX images acquired with the Ig-UTE protocol show no flow artifacts (room temperature RF coil; images
were kindly provided by Bruker Biospin MRI GmbH, Ettlingen, Germany)
Cardiac MRI 283
obtaining good quality images. The detailed description of the

trajectory measurement procedure is beyond the scope of this
protocol. In case you need help with this please seek advice
from an MR expert or the MR system vendor.
6. When using ParaVision software version 6, some DICOM tags
from exported DICOM images has to be removed in order to
display all the images correctly in CMR42 (for windowing
reason). When using ParaVision software version 5, this is
not necessary. Remove DICOM tag as following: “Rescale
Intercept” (0028,1052), “Rescale Slope” (0028,1053), and
“Rescale Type” (0028,1054).
Acknowledgement
usebrink from Bruker

The authors wish to thank Thomas Basse-L€
Biospin MRI GmbH, Ettlingen, Germany, for providing the
Ig-UTE image data.
References
1. Camacho P, Fan H, Liu Z, He JQ (2016) Small French BA, Epstein FH (2012) T(2) -weighted
mammalian animal models of heart disease. Am MRI of post-infarct myocardial edema in mice.
J Cardiovasc Dis 6(3):70–80 Magn Reson Med 67(1):201–209. https://
2. Zaragoza C, Gomez-Guerrero C, Martin- doi.org/10.1002/mrm.22975
Ventura JL, Blanco-Colio L, Lavin B, 7. Musthafa HS, Dragneva G, Lottonen L,
Mallavia B, Tarin C, Mas S, Ortiz A, Egido J Merentie M, Petrov L, Heikura T,
(2011) Animal models of cardiovascular dis- Yla-Herttuala E, Yla-Herttuala S, Grohn O,
eases. J Biomed Biotechnol 2011:497841. Liimatainen T (2013) Longitudinal rotating
https://doi.org/10.1155/2011/497841 frame relaxation time measurements in
3. Stuckey DJ, McSweeney SJ, Thin MZ, Habib J, infarcted mouse myocardium in vivo. Magn
Price AN, Fiedler LR, Gsell W, Prasad SK, Reson Med 69(5):1389–1395. https://doi.
Schneider MD (2014) T(1) mapping detects org/10.1002/mrm.24382
pharmacological retardation of diffuse cardiac 8. Vandsburger MH, Epstein FH (2011)
fibrosis in mouse pressure-overload hypertro- Emerging MRI methods in translational car-
phy. Circ Cardiovasc Imaging 7(2):240–249. diovascular research. J Cardiovasc Transl Res
https://doi.org/10.1161/CIRCIMAGING. 4(4):477–492. https://doi.org/10.1007/
113.000993 s12265-011-9275-1
4. Coelho-Filho OR, Shah RV, Mitchell R, Neilan 9. Wagenhaus B, Pohlmann A, Dieringer MA,
TG, Moreno H Jr, Simonson B, Kwong R, Els A, Waiczies H, Waiczies S, Schulz-Menger J,
Rosenzweig A, Das S, Jerosch-Herold M Niendorf T (2012) Functional and morpho-
(2013) Quantification of cardiomyocyte hyper- logical cardiac magnetic resonance imaging of
trophy by cardiac magnetic resonance: implica- mice using a cryogenic quadrature radiofre-
tions for early cardiac remodeling. Circulation quency coil. PLoS One 7(8):e42383. https://
128(11):1225–1233. https://doi.org/10. doi.org/10.1371/journal.pone.0042383
1161/CIRCULATIONAHA.112.000438 10. Niendorf T, Pohlmann A, Reimann HM,
5. Abeykoon S, Sargent M, Wansapura JP (2012) Waiczies H, Peper E, Huelnhagen T,
Quantitative myocardial perfusion in mice Seeliger E, Schreiber A, Kettritz R, Strobel K,
based on the signal intensity of flow sensitized MC K, Waiczies S (2015) Advancing cardiovas-
CMR. J Cardiovasc Magn Reson 14:73. cular, neurovascular, and renal magnetic reso-
https://doi.org/10.1186/1532-429X-14-73 nance imaging in small rodents using cryogenic
6. Beyers RJ, Smith RS, Xu Y, Piras BA, radiofrequency coil technology. Front
Salerno M, Berr SS, Meyer CH, Kramer CM,
Pharmacol 6:255. https://doi.org/10.3389/ Reson 15:59. https://doi.org/10.1186/

fphar.2015.00255 1532-429X-15-59
11. Hoerr V, Nagelmann N, Nauerth A, Kuhlmann 12. Ring J, Hoerr V, Tuchscherr L, Kuhlmann MT,
MT, Stypmann J, Faber C (2013) Cardiac- Loffler B, Faber C (2014) MRI visualization of
respiratory self-gated cine ultra-short echo Staphyloccocus aureus-induced infective endo-
time (UTE) cardiovascular magnetic resonance carditis in mice. PLoS One 9(9):e107179.
for assessment of functional cardiac parameters https://doi.org/10.1371/journal.pone.
at high magnetic fields. J Cardiovasc Magn 0107179
Chapter 17
In Utero MRI of Mouse Embryos

Jiangyang Zhang, Dan Wu, and Daniel H. Turnbull
Abstract
Genetically engineered mouse models are used extensively as models of human development and develop-
mental diseases. Conventional histological approaches are static and two-dimensional, and do not provide a
full understanding of the dynamic, spatiotemporal changes in developing mouse embryos. Magnetic
resonance imaging (MRI) offers a noninvasive and longitudinal approach for three-dimensional in utero
imaging of normal and mutant mouse embryos. In this chapter, we describe MRI approaches that have been
developed for imaging the living embryonic mouse brain and vasculature. Details are provided on the
animal preparation and setup, MRI equipment, acquisition and reconstruction methods that have been
found to be most useful for in utero MRI, including examples of applications to fetal mouse neuroimaging.
Key words Diffusion MRI, Diffusion weighted gradient and spin echo (DW-GRASE), Field of
excitation (FOE), Fractional anisotropy (FA), High-field MRI, Mn-enhanced MRI (MEMRI), Phased
array coil, Three-dimensional (3D)
1 Introduction
As a model for mammalian embryogenesis and human develop-

mental diseases, the mouse has provided a wealth of resources and
research opportunities through the use of increasingly sophisticated
genetic engineering technologies over more than two decades
[1]. Despite these advances, the mouse embryo remains a difficult
model organism, compared to lower species such as drosophila,
c. elegans, and zebrafish. Specifically, embryogenesis involves com-
plex spatiotemporal changes in the brain, heart, vasculature, and
other organs that are difficult to observe directly in mouse
embryos, which are encased deep within the maternal uterus
[2]. High frequency ultrasound imaging has provided a noninvasive
in utero imaging approach that enables relatively high-throughput,
longitudinal three-dimensional (3D) analysis of dynamic anatomi-
cal changes in mouse embryos [3, 4], but provides limited ability to
manipulate image contrast to better assess changes at the organ/
tissue level. While magnetic resonance imaging (MRI) has been
used extensively for high-resolution (20–50 μm) ex vivo imaging
285
286 Jiangyang Zhang et al.
of fixed mouse embryos (reviewed in [2, 4]), in vivo MRI of mouse

embryonic development in the normal maternal uterus has been
less common. In this chapter we focus exclusively on new advances
in MRI as an (in vivo) in utero imaging method that offers more
control over image contrast than ultrasound.
To illustrate the potential of 3D in utero MRI for studies of
developing mouse embryos, we show examples of applications in
cardiovascular and brain imaging. Although these organ systems
have seen the most applications to date, there is no doubt that in
utero MRI can be developed and applied in future to multiple
organ systems. In the brain, we have developed both in utero
Manganese (Mn)-enhanced MRI (MEMRI) [5] and diffusion
MRI [6] approaches that provide alternate contrast for quantitative
analyses of many important tissue structures at the mesoscopic scale
(~100 μm). These initial studies have demonstrated the feasibility
of MEMRI and diffusion MRI to examine axonal microstructures
and neuronal differentiation in the normal and mutant embryonic
mouse brain, providing techniques for quantitative and longitudi-
nal analysis of living mouse embryos, in utero.
2 Materials
2.1 High-Field MRI 1. Small animal MRI system: a high-field (preferably 7 Telsa or
Systems higher) MRI system is required to perform in utero MRI of
mouse embryos. The in utero MRI experiments described in
this chapter were performed on horizontal 7 Tesla (T) and
11.7 T MRI systems (Bruker Biospin, Billerica, MA, USA),
each with an integrated shim and actively shielded three-axis
gradient (B-GA9S, Bruker Biospin, Billerica, MA, USA, inner
diameter ¼ 90 mm, maximum gradient strength ¼ 740
mT/m). The 11.7 T scanner has a motor driven automatic
positioning system (Autopac, Bruker Biospin) for animal
positioning.
2. RF coils: MRI experiments (7 and 11.7 T) have been per-
formed using a 72 mm (inner diameter) quadrature volume
coil for transmit (Bruker Biospin) in combination with a
receive-only planar surface coil, using a pin-diode to decouple
the transmit and receive coils and a low-noise preamplifier on
the receive coil. At 11.7 T, a whole-body 8-channel receive-
only phased array coil was also used (Bruker Biospin, Part
No. T20035 V3) in combination with the transmit
volume coil.
3. Mouse Holders: We use either custom-made holders or the
Autopac compatible animal holders provided by the manufac-
turer of the MRI system (Bruker Biospin).
In Utero MRI 287
4. Small animal monitoring and gating system: We used a system

manufactured by the Small Animal Instruments, Inc. (SAII,
Stony Brook, NY, USA) to monitor the physiology of the
pregnant mice. The respiratory motion and body temperature
was monitored via a pressure sensor and a rectal temperature
probe and displayed on a computer next to the scanner console.
Respiratory motion can also be monitored via a self-gated
(MRI) motion signal [8]. In either case, trigger signals can be
generated and sent to the MRI system to synchronize MRI
acquisitions with respiratory motion. In addition, an animal
warm air heating system provided by the manufacturer of the
MRI system was used to maintain physiological body
temperature.
5. Gas anesthesia machine: A veterinary isoflurane vaporizer (VMS
Matrix Medical, Orchard Park, NY) was used to anesthetize
pregnant mice during MRI.
2.2 Supplies 1. Isoflurane: (Aerane, Baxter, Deerfield IL).

2. MnCl2 solution: An 30 mM solution of manganese chloride
tetrahydrate (e.g., Sigma 221279) in isotonic saline should be
prepared ahead of time.
3 Methods
3.1 MnCl2-Contrast For MEMRI, MnCl2 solution is administered as a maternal intraper-

Agent itoneal injection 4–24 h before MRI, at a dose of 40–80 mg MnCl2
/ kg body weight [5]. For all other MRI protocols described in this
chapter, no exogenous contrast agents are required (see Note 1).
3.2 Animal Setup Anesthesia is induced in pregnant mice using 3–5% isoflurane
mixed with air (via the vaporizer), or with a 3:1 air:oxygen mixture
(see Note 2). During MRI, the amount of isoflurane should be
reduced to 1–1.5% and adjusted regularly to maintain the respira-
tory rate at 30–60 breaths per minute. A pregnant mouse may have
up to 12 embryos, each residing in their own amniotic sacs and
distributed along the two uterine horns (Fig. 1a). In order to
achieve uniform excitation while maintaining high sensitivity, it is
ideal to use a large whole body volume coil as the transmission coil
(Tx) in combination with a receive-only surface coil or a receive-
only phased array body coil (Fig. 1b). The phased array coil pro-
vides more abdominal coverage than the surface coil but the surface
coil can be positioned as close to the target embryo as possible and
provide higher sensitivity (see Note 3).
Fig. 1 Animal setup and the initial 2D multi-slice T2-weighted images. (a) The anatomy of a pregnant mouse
showing 12 embryos in the two uterine horns: (b) Placement of a pregnant mouse with a receive-only planar
surface coil (top) and a receive-only phased array coil (bottom). (c) axial T2-weighted images of mouse
embryos at embryonic day 17 (E17) in a pregnant mouse. The embryos are surrounded by the amniotic fluid,
which has strong T2 signals. With high in-plane resolution, large internal organs, e.g., the brain, liver, and
spinal cord, can be identified. (d) Three-dimensional rendering of amniotic sacs (gray) and embryonic mouse
brains (purple) in the abdomen based on the axial T2-weighted images
3.3 Three- We have successfully used 3D gradient echo sequences to generate

Dimensional (3D) T1- T1-weighted images (echo / repetition times, TE/TR ¼ 5/40 ms;
or T2*-Weighted MRI Flip angle ¼ 35 ) for MEMRI [5], and T2*-weighted images
of Mouse Embryos (TE/TR ¼ 20/50 ms; Flip angle ¼ 20 ) for in utero cardiovascular
imaging [7–9]. With a close-fitting surface coil, in utero (T1-
weighted) MEMRI with respiratory gating was used to acquire
100 μm isotropic resolution images of the mouse embryonic
brain between E12.5–17.5 (Fig. 2a) (see Note 4) including volu-
metric analysis of forebrain defects in Nkx2.1/ mutant embryos
[5]. Using self-gated acquisition of 3D T2*-weighted images com-
bined with image co-registration [8], we showed that motion
artifacts can be significantly suppressed enabling acquisition of
high quality vascular images from early stage (E10.5–14.5) mouse
embryos [9] (Fig. 2b). Similar approaches have enabled acquisition
and analyses of cardiovascular images in later stage (E17.5)
In Utero MRI 289
Fig. 2 In utero T1-weighted MEMRI and T2*-weighted vascular MRI. (a) Mn-enhancement and respiratory
gating work in combination to improve in utero MRI. Images at E14.5 with gating and without Mn (Left panel,
Gatingþ/Mn), with Mn and without gating (Middle panel, Gating/Mnþ) and with gating and Mn (Right
panel, Gatingþ/Mnþ). Scale bar is 1 mm. The images are modified with permission from [5]. (b) 3D maximum
intensity projections (MIPs) show the developing vasculature from E10.5 to E14.5. Labels: DA dorsal aorta, FV
facial vein, H heart, JV jugular vein, NV nasal vein, OA optic artery, VC vena cava, VS venous sinus. The images
are modified with permission from [9]
embryos, including the detection of novel vascular phenotypes in

Gli2/ mutant embryos [7].
In the following sections, a systematic approach is described for
identifying individual embryos and acquiring high-resolution T2-
weighted and diffusion MRI images of the embryonic mouse
brain [6].
3.4 Two- A fast survey of the complex anatomy can be obtained using 2D
Dimensional multi-slice T2-weighted MRI, which can provide satisfactory con-
(2D) Multi-Slice trasts to distinguish individual embryos in a relatively short time
T2-Weighted MRI (~5 min) (Fig. 1c). Because of the low through-plane resolution of
of Pregnant Mice 2D MRI, both axial and sagittal images of the mouse abdomen
need to be acquired to visualize the locations and orientations of
mouse embryos and guide subsequent imaging.
Axial and sagittal T2-weighted images can be acquired using

the Rapid Acquisition with Relaxation Enhancement (RARE)
sequence provided by the manufacturer of the MRI system with
the following parameters: echo time (TE) ¼ 50 ms, repetition time
(TR) ¼ 3000 ms, 2 signal averages, a RARE factor of 8, in-plane
resolution ¼ 0.16 mm 0.16 mm, approximately 50 slices with a
thickness of 1 mm covering the mid to lower abdomen. To remove
artifacts due to respiratory motion, respiratory triggering should be
enabled. Fig. 1d shows the distribution of embryos within the
amniotic sacs (gray) and embryonic mouse brains (purple) in the
uterus based on the 2D multi-slice T2-weighted MRI results.
3.5 Generation The next step is to select an embryo for localized imaging based on
of Tailored Radio- the 2D multi-slice T2-weighted images (see Subsection 3.4). For
Frequency (RF) Pulses localized imaging, it is necessary to define a so-called field of
for Localized Imaging excitation (FOE) that encloses the selected embryo or embryonic
of Mouse Embryos brain. To facilitate the process, we have designed a software tool to
display the T2-weighted images and define the FOE (Fig. 3). The
software is based on Matlab (Mathworks, mathworks.com) and can
run on the scanner console. Once the size and location of an FOE
are defined, the software generates a tailored 90 selective excita-
tion RF pulse based on a linear class of large tip-angle (LCLTA)
pulses [10] with spiral k-space trajectories that start and end at the
origin. Under the “incoherently refocused” condition, a 2D selec-
tive 90 RF pulse can be derived by inverse Fourier transform of the
desired excitation profile [10]. The pulse is designed to excite a
rectangular FOE in the x–y plane that covers the target region, with
a duration of 3 ms, an amplitude of 9–10 μT, and a 12-turn spiral-in
excitation k-space (maximum gradient strength ¼ 148 mT/m).
Details on the performance of the designed RF pulses can be
found in [11].
3.6 3D T2-Weighted To acquire high-resolution images of the embryonic mouse brain,

and Diffusion MRI we use a 3D diffusion-weighted gradient and spin echo
of the Embryonic (DW-GRASE) sequence [12, 13] with an echo train length of
Mouse Brain 20 for fast imaging and two navigator echoes appended after the
imaging echoes to correct phase errors due to motion and instru-
ment instability. Once the selective excitation RF pulse is generated
for a particular FOE, it can be inserted in the sequence with a slab-
selective refocusing RF pulse [14] to restrict the imaging slab in the
third direction. Figure 4 shows a diagram of the sequence. In our
previous experiments [6], 3D Diffusion MRI data were acquired
using the DW-GRASE sequence with the following parameters:
TE/TR ¼ 21/500 ms; two signal averages; spectral
width ¼ 120 kHz; four b0 images and 30 diffusion directions [15];
b-value ¼ 1000 s/mm2; FOV ¼ 12.8 mm 12.8 mm 8 mm; and
spatial resolution ¼ 0.2 mm 0.2 mm 0.2 mm in 72 min (2 min
per diffusion-weighted image) or 0.16 mm 0.16 mm 0.16 mm
In Utero MRI 291
Fig. 3 Definition of the field of excitation (FOE, represented by the red box) based on the 2D multi-slice
T2-weighted images using a software tool. Using the software, users can display the T2-weighted images,
define the size and spatial offsets of an FOE, and calculate the selective excitation RF pulse to excite the FOE
with the desired parameters. The lower panel displays the amplitude and phase profiles of the generated RF
pulse
in 113 min. High-resolution 3D T2-weighted images were acquired

using the same setup but without diffusion weighting: TE/TR ¼ 24/
1000 ms and resolution ¼ 0.13 0.13 0.13 mm in 10 min (see
Note 5).
3.7 Image The 3D k-space data are first apodized with a tapered cosine win-
Reconstruction dow, zero-padded to twice the original size, and reconstructed in
Matlab. The twin-navigator echoes are Fourier transformed along
the readout direction, which are then used to correct the phases of
odd- and even-numbered echoes from each repetition and phase
errors caused by intra-scan motion [12]. The 30 direction
Fig. 4 A diagram of the 3D DW-GRASE sequence with a spatially selective excitation pulse. The diagram
shows the timing of the 2D selective excitation pulse together with the spiral gradient in the x–y plane, the
diffusion encoding gradients (represented by turquoise trapezoids), the GRASE readout module, and the twin-
navigator echoes. Each GRASE readout module acquires four gradient echoes and one spin echo, and the
readout is repeated four times to achieve an acceleration factor of 20 compared to the conventional spin echo
sequence. The images are modified with permission from [6]
diffusion-weighted images (DWIs) are aligned to the mean DWI

using 3D rigid transformation to correct any inter-scan motion.
The effects of both navigator echoes and rigid transformation based
motion corrections are shown in Fig. 5. The amount of transla-
tional motion in each diffusion direction can be estimated based on
the image registration results. Images with motion above 0.6 mm
(3 voxels) may be excluded from the following analysis (see Note 6).
Diffusion tensor fitting can be performed in DtiStudio (www.
mristudio.org). Using the log-linear fitting method implemented
in DTIStudio (http://www.mristudio.org), the diffusion tensor
was calculated at each pixel, along with the apparent diffusion
coefficient (ADC), fractional anisotropy (FA), and primary eigen-
vector [16]. Figure 6a shows the reconstructed T2-weighted, FA,
and directionally encoded colormap images of an E17.5 mouse
brain. White tracts can be reconstructed using the same software
with a fractional anisotropy (FA) threshold of 0.15 and maximum
angle of 60 (Fig. 6c).
4 Notes
1. For in utero MRI, we have found that MnCl2 can cause embry-
onic toxicity and death, especially for embryos staged earlier
than embryonic day (E)13.5 [5]. For embryos staged E13.5 or
In Utero MRI 293
Fig. 5 Methods for motion correction employed in in utero imaging of the embryonic mouse brain. (a) Due to
intra-scan motion, the boundary of the fourth ventricle in the embryonic mouse brain (indicated by the yellow
arrow) is not clear. Real-time navigator echoes can be used to correct intra-scan motion, and the boundary of
the 4th ventricle becomes clear. (b) Motions between scans cause images acquired sequentially to be
mis-adlinged, as indicated by the cross-hair in the images. This mis-alignment can be corrected using linear
rigid image registration. (c) The inter-scan motion can be estimated using the linear rigid image registration.
The plot shows he amount of overall brain displacement of 5 E17.5 mouse brains during 60-min scans. The
images are modified with permission from [6]
older, we recommend an intraperitoneal dose of 40 mg

MnCl2/kg (maternal) body weight, injected 24 h before MRI
to achieve adequate MEMRI contrast while avoiding acute
Mn-toxicity to the embryos.
2. All animals used in the experiments described in this chapter
were maintained under protocols approved by the Institutional
Animal Care and Use Committees at New York University
School of Medicine and Johns Hopkins University School of
Medicine.
3. The phased array coil used in our study (Part No. T20035V3)
was designed for imaging a rat body, which is significantly
Fig. 6 In utero MRI of the embryonic mouse brain. (a) Coronal T2-weighted, FA, and directionally encoded
colormap (DEC) images of an E17.5 mouse brain. (b) Surface rendering of the ventricules in the E17.5 mouse
brain based on the T2-weighted images. (c) rendering of the gray matter structures and early white matter
tracts in the E17.5 mouse brain based on the diffusion MRI data. Abbreviations: CP cortical plate, cp cerebral
peduncle, IZ intermediate zone, fi fimbria, Hi hippocampus, LV lateral ventricle, opt optical tract, st stria
terminalis, Th thalamus, 3V and 4V the third and forth ventricles
larger than the pregnant mouse. For optimal image quality, it

will be critical to design a series of phased array coils that
accommodate pregnant mice at different stages of pregnancy
and increase sensitivity. These technical improvements are
ongoing, but a description of the design, fabrication, and
testing of MRI coils is beyond the scope of this chapter.
4. By convention, mouse embryos are staged by embryonic day
(E), where E0.5 is defined to be noon of the day that a vaginal
plug is detected after overnight mating.
5. The DW-GRASE sequence with selective excitation can be
easily modified to image the entire mouse embryo. Because it
requires a relatively long echo time to accommodate the
GRASE readout model, the sequence cannot be used for
T1-weighted MRI, which requires a short echo time. However,
the selective excitation pulse can be used in standard gradient
echo and spin echo sequences to achieve localized T1-weighted
MRI required for MEMRI.
6. The twin-navigator echo approach can only be used to correct
intra-scan motions along the readout direction. For imaging
the mouse embryos, maternal respiratory motion is the main
In Utero MRI 295
concern, and the readout direction should be defined along the

direction of maternal respiratory motion to fully utilize the
navigator-based motion correction scheme. In addition, the
motion correction scheme introduced in [8] (see Subheading
3.3) can also be easily incorporated into the DW-GRASE
sequence to further reduce the effects of subject motion.
Acknowledgements
We thank all the people, current and past, in the Zhang, Wu and
Turnbull labs who have contributed to developing the protocols
described in this chapter. This research was supported, in part, by
grants from the National Institutes of Health: R01NS038461 and
R01HL078665 (DHT); R01NS070909 and R01HD974593
(JZ); R21NS098018 (DW).
References
1. International Mouse Knockout Consortium, mouse embryonic vasculature with MRI. Magn
Collins FS, Rossant J, Wurst W (2007) A Reson Med 67(1):251–257. https://doi.org/
mouse for all reasons. Cell 128(1):9–13. 10.1002/mrm.22991
https://doi.org/10.1016/j.cell.2006.12.018 8. Nieman BJ, Szulc KU, Turnbull DH (2009)
2. Nieman BJ, Wong MD, Henkelman RM Three-dimensional, in vivo MRI with self-
(2011) Genes into geometry: imaging for gating and image coregistration in the mouse.
mouse development in 3D. Curr Opin Genet Magn Reson Med 61(5):1148–1157. https://
Dev 21(5):638–646. https://doi.org/10. doi.org/10.1002/mrm.21945
1016/j.gde.2011.08.009 9. Parasoglou P, Berrios-Otero CA, Nieman BJ,
3. Aristizabal O, Mamou J, Ketterling JA, Turn- Turnbull DH (2013) High-resolution MRI of
bull DH (2013) High-throughput, high-fre- early-stage mouse embryos. NMR Biomed 26
quency 3-D ultrasound for in utero analysis of (2):224–231. https://doi.org/10.1002/nbm.
embryonic mouse brain development. Ultra- 2843
sound Med Biol 39(12):2321–2332. https:// 10. Pauly J, Nishimura D, Macovski A (1989) A
doi.org/10.1016/j.ultrasmedbio.2013.06. linear class of large-tip-angle selective excita-
015 tion pulses. J Magn Reson 82(3):571–587.
4. Nieman BJ, Turnbull DH (2010) Ultrasound https://doi.org/10.1016/0022-2364(89)
and magnetic resonance microimaging of 90219-9
mouse development. Methods Enzymol 11. Wu D, Reisinger D, Xu J, Fatemi SA, van Zijl
476:379–400. https://doi.org/10.1016/ PC, Mori S, Zhang J (2014) Localized diffu-
S0076-6879(10)76021-3 sion magnetic resonance micro-imaging of the
5. Deans AE, Wadghiri YZ, Berrios-Otero CA, live mouse brain. NeuroImage 91:12–20.
Turnbull DH (2008) Mn enhancement and https://doi.org/10.1016/j.neuroimage.
respiratory gating for in utero MRI of the 2014.01.014
embryonic mouse central nervous system. 12. Aggarwal M, Mori S, Shimogori T,
Magn Reson Med 59(6):1320–1328. https:// Blackshaw S, Zhang J (2010) Three-
doi.org/10.1002/mrm.21609 dimensional diffusion tensor microimaging
6. Wu D, Lei J, Rosenzweig JM, Burd I, Zhang J for anatomical characterization of the mouse
(2014) In utero localized diffusion MRI of the brain. Magn Reson Med 64(1):249–261.
embryonic mouse brain microstructure and https://doi.org/10.1002/mrm.22426
injury. J Magn Reson Imaging 42:717. 13. Wu D, Xu J, McMahon MT, van Zijl PC,
https://doi.org/10.1002/jmri.24828 Mori S, Northington FJ, Zhang J (2013) In
7. Berrios-Otero CA, Nieman BJ, Parasoglou P, vivo high-resolution diffusion tensor imaging
Turnbull DH (2012) In utero phenotyping of of the mouse brain. NeuroImage 83:18–26.
https://doi.org/10.1016/j.neuroimage. https://doi.org/10.1007/978-1-61737-992-
2013.06.012 5_6
14. Mao J, Mareci TH, Andrew ER (1988) 16. Jiang H, van Zijl PC, Kim J, Pearlson GD,
Experimental-study of optimal selective Mori S (2006) DtiStudio: resource program
180-degrees radiofrequency pulses. J Magn for diffusion tensor computation and fiber
Reson 79(1):1–10. https://doi.org/10. bundle tracking. Comput Methods Prog
1016/0022-2364(88)90317-4 Biomed 81(2):106–116. https://doi.org/10.
15. Jones DK, Leemans A (2011) Diffusion tensor 1016/j.cmpb.2005.08.004
imaging. Methods Mol Biol 711:127–144.
Chapter 18
Oxygenation Imaging by Nuclear Magnetic Resonance

Methods
Heling Zhou, Nuria Arias-Ramos, Pilar López-Larrubia, Ralph P. Mason,
Sebastián Cerdán, and Jesús Pacheco-Torres
Abstract
Oxygen monitoring is a topic of exhaustive research due to its central role in many biological processes,
from energy metabolism to gene regulation. The ability to monitor in vivo the physiological distribution
and the dynamics of oxygen from subcellular to macroscopic levels is a prerequisite to better understand the
mechanisms associated with both normal and disease states (cancer, neurodegeneration, stroke, etc.). This
chapter focuses on magnetic resonance imaging (MRI) based techniques to assess oxygenation in vivo. The
first methodology uses injected fluorinated agents to provide quantitative pO2 measurements with high
precision and suitable spatial and temporal resolution for many applications. The second method exploits
changes in endogenous contrasts, i.e., deoxyhemoglobin and oxygen molecules through measurements of
T2* and T1, in response to an intervention to qualitatively evaluate hypoxia and its potential modulation.
Key words MRI, Oximetry, pO2, BOLD, Hypoxia, Perfluorocarbons, Quantification
1 Introduction
Tissue oxygenation comprises a vast field of study with a direct

impact on the understanding and clinical outcome of multiple
disorders including, but not limited to, neurodegenerative diseases
[1, 2], traumatic brain injury [3], heart attack [4], stroke [5], and
cancer [6, 7]. The importance of oxygenation has been of particu-
larly interest in oncology since the classic studies of Gray et al. more
than 60 years ago demonstrated that hypoxia can influence the
efficacy of radiotherapy on tumors [8]. Since then, multiple
attempts have tried to modulate or exploit hypoxia to increase
tumor responsiveness to treatments [9–11], but translation to the
clinic has shown marginal efficacy [12]. The meta-analysis by Over-
gaard et al. [9, 13] of more than 10,000 patients indicated a clinical
benefit for manipulating tumor hypoxia. However, the overall con-
clusion was that there was a pressing need to identify those tumors
(viz. patients) who would actually benefit.
297
298 Heling Zhou et al.
The Cancer Imaging Program of the National Cancer Institute

sponsored a workshop on hypoxia imaging technologies in 2004
[12], with multiple reviews addressing this issue since then [6, 7,
14]. The ideal properties of a pO2 measurement technique would
include noninvasiveness, appropriate spatial and temporal resolu-
tion, quantitative measurement, no oxygen consumption, adequate
range of pO2 values measured and, if an exogenous agent is
required, exhaustive knowledge of its toxicology and pharmaco-
dynamic properties. Unfortunately, there is currently no in vivo
method that fulfills satisfactorily all these requirements [6].
The literature in this field is broad, spanning several decades of
research and imposing, necessarily, a selection of the topics consid-
ered, so that many valuable contributions could not be mentioned
due to space limitations. Interested readers are referred to previous
review articles on additional methods to measure tissue oxygen levels
in situ, such as oxygen electrodes [15–17], positron emission tomog-
raphy (PET) [18–20], single photon emission computed tomogra-
phy (SPECT) [21, 22], immunohistochemistry [23], near-infrared
spectroscopy [24, 25], and phosphorescence [24, 26], and more
recently multispectral optoacoustic tomography [27]. Additional
Magnetic Resonance (MR) based methodologies to those presented
here have been proposed to assess indirectly oxygenation through
the alterations in perfusion and water diffusion [28, 29].
In this chapter, we aim to provide a general overview of evalua-
tion of tumor hypoxia by MR methods in vivo. We focus on two
methods: one providing quantitative measurements of pO2 by
MRI, based on the quantification of T1 of perfluorocarbons
(PFCs) by19F MRI; and a new emerging technique termed
Dynamic Oxygen Challenge Evaluated by NMR T1 and T2*
(DOCENT) based on the changes associated with increased oxy-
gen availability in tissues and blood, respectively.
PFCs are organic molecules in which all hydrogen atoms have
been replaced by fluorine atoms. The PFC spin-lattice relaxation
rate (R1 ¼ 1/T1) is extremely dependent on pO2, obeying a linear
relationship (R1 ¼ a + b pO2) at a constant magnetic field and
temperature [30, 31]. The application of that phenomenon to
oximetry was pioneered by Thomas et al. [32] and has been devel-
oped extensively by several research labs including Mason et al.
[30, 33, 34], Sotak et al. [35], and recently Gallez et al.
[36]. Once calibration curves have been determined for a particular
PFC in vitro, in vivo measurements of R1 can be transformed easily
into pO2 maps [7, 20]. This technique has been applied to many
organs/disease sites with extensive validation [37]. If the19F atoms
present in the PFCs are not chemically equivalent, then the mole-
cule will present several resonance frequencies and a system of
equations may allow the simultaneous determination of tempera-
ture and pO2 [33, 34]. Nevertheless, single-resonance PFCs have
MR Oximetry 299
been favored, because they simplify the imaging and the processing
of the pO2 maps and tend to provide higher signal-to-noise ratio
[38]. The PFC administration protocol is an important consider-
ation. In this chapter, we will describe the direct administration of
hexafluorobenzene (HFB) at the site of measurement. Neverthe-
less, alternative methods are also available based on the particular
needs of the experiment [6].
We may note that19F MRI remains relatively esoteric on human
clinical systems and thus an analogous proton MRI approach was
proposed. Hexamethyldisiloxane (HMDSO) may serve as a
novel1H NMR oximetry probe analogous to PFC [39]. It shares
some important properties with PFCs making a highly suitable
oximetry probe [40]. Its development would avoid the limitation
of special hardware and sequences needed for19F MRI allowing,
therefore, a straightforward implementation in clinical scanner.
This would make possible the combination of oximetry informa-
tion with that derived from other routinely used1H MR-based
methods, such as dynamic contrast enhancement, diffusion mea-
surements, and MR spectroscopy. We do not describe this method-
ology here, but it may be implemented using a similar protocol,
with the addition of water suppression. Interested readers can go to
[39–41].
There are alternative methods to measure oxygenation using
endogenous reporters, based on the magnetic nature of the oxygen
molecule (O2) or its main blood transporter, hemoglobin. Blood
Oxygen Level Dependent (BOLD) contrast relies on the fact that
deoxyhemoglobin (dHb) is paramagnetic, whereas its oxygenated
counterpart, oxyhemoglobin (HbO2), is diamagnetic. Briefly,
deoxygenated red blood cells generate local magnetic field gradi-
ents causing signal dephasing and thus signal loss in T2*-weighted
images [42, 43]. This approach was originally developed to investi-
gate cerebral activity (fMRI), being extended more recently to the
assessment of tumor oxygenation in response to hyperoxic gas
challenge [21, 44]. The magnitude of this effect depends on multi-
ple factors (see Note 1), preventing a simple and direct relationship
between T2* changes and pO2 [45]. These circumstances
prompted the introduction of the acronym “FLOOD” (flow and
oxygenation dependent) to refer to the BOLD experiment, as
applied to the study of oxygenation in tumors [46]. Nontheless
several reports have indicated a strong correlation between changes
in R2* (or T2*-weighted signal) and pO2 [47–50].
More recently, T1 changes in response to increased oxygen
content in the gas breathed have been demonstrated in both
tumor [50–52] and normal tissues [53] providing the so-called
Tissue Oxygen Level Dependent (TOLD) approach. This tech-
nique generally does not measure absolute pO2 directly, except in
rare circumstances such as vitreous humor or CSF [45, 54], since
many factors can influence R1. However, it has been applid to assess
changes in pO2 (ΔpO2) in response to interventions

[51, 52]. While BOLD depends on the concentration of deoxyhe-
moglobin, TOLD primarily depends on tissue pO2, although there
is evidence that deoxyhemoglobin may also influence T1 for highly
vascularized hypoxic regions [55]. The combination of BOLD and
TOLD response to oxygen challenge provides particularly robust
insight into tumor hypoxia and its potential modulation. This is the
basis of the simple test DOCENT that will be described here.
Notably, BOLD indicates changes in vascular oxygenation, which
is followed by diffusion of O2 into the tissue generating a TOLD
effect, thereby confirming improved tissue oxygenation. This
approach presents several advantages; it can be undertaken readily
in any MRI system, does not need any extrinsic contrast agent and
thus can be used repetitively.
2 Materials
2.1 Chemicals 1. Eye ointment.

and Materials 2. PFC: Several PFCs that have been demonstrated [38], but we
favor hexafluorobenzene.
3. Hamilton syringe with a fine sharp 32G needle.
4. Isoflurane.
5. Medical air and medical oxygen or carbogen (95% O2 and 5%
CO2).
2.2 MRI System For pO2 mapping with PFC, the MRI system must provide dual
nuclei capability or a broad frequency response.
2.2.1 Magnetic
Resonance Imaging
System
2.2.2 MRI Coil 1. The expected changes for T1- and T2*-weighted images are
small, so it is critical to achieve excellent signal-to-noise ratio
(SNR) and field homogeneity. We used a1H selective birdcage
resonator of 38 mm and a gradient system with 360 mT/m of
maximum intensity.
2. For pO2 mapping with PFC a tunable coil to1H or19F is needed
(see Note 2). We use a homebuilt 2 or 3.5 cm single-turn
solenoid volume coil.
2.2.3 Sequences 1. Anatomical images: Used for localization of the tumor and to
acquire representative images to overly the different acquired
maps (see Note 3).
2. DOCENT: Spoiled Gradient Echo images are needed. Sug-
gested parameters: (1) T1 weighted, repetition time
MR Oximetry 301
(TR) ¼ 30 ms, echo time (TE) ¼ 5 ms, flip angle (FA) ¼ 45 ;
and (2) T2* weighted, TR ¼ 150 ms, TE ¼ 20 ms and FA ¼ 20
(see Note 4).
3. pO2 mapping: T1 map is acquired applying the FREDOM
sequence, which uses Pulse Burst Saturation Recovery
(PBSR) echo planar imaging (EPI), by arraying 14 delay
times (τ) [38]. The FREDOM parameters are: TR ¼ 50 ms,
TE ¼ 21 ms, τ ranges from ¼ 0.2 to 90 s, number of excitations
(NEX) ¼ 1–12 (depending on τ), field of view
(FOV) ¼ 40 40 mm with 32 32 acquisition matrix, slice
thickness 10 mm, giving a total acquisition time of 6½ min.
T1 map can also be performed by using a standard inversion
recovery sequence, but acquisition time will be significantly
increased, or Modified Look Locker sequence [36].
2.2.4 Physiological At minimum, it is necessary to monitor the animal temperature and

Monitoring and Control breathing rate. An MRI compatible temperature control unit is
System needed. Breathing rate can be monitored using a device sensitive
to pressure positioned on the chest of the animal. Other physiolog-
ical parameters that are recommended to assess are heart rate,
oxygen saturation, and pulse distension. They can be monitored
throughout the session using an MRI compatible sensor with foot
clip (MouseOx, Starr Life Sciences, Oakmont, US or Pulse Oxime-
try, Small Animal Instruments, Inc., US).
2.2.5 Heating System It is important to keep the animal temperature in the physiological
range (see Notes 5 and 6). The two common methodologies to
achieve this are blowing warm air over the subject or placing the
subject on a bed of circulating water connected to a water bath
controlled by a temperature regulatory system (see Note 7). If the
heating system only warms a small surface of the animal (i.e.,
waterbed), wrapping the animal to minimize heat loss would be
very convenient.
2.2.6 Anesthesia 1. Injectable anesthetics: ketamine-xylazine, propofol, remifenta-

nil, urethane, etc.
2. If using inhaled anesthesia: e.g., isoflurane and a vaporizer
system are also required (see Subheading 3.1 for further
details).
2.3 Software for Data Processing methods require software capable of handling MR
Post-processing images. For DOCENT analysis, DCE-like software is sufficient. In
the case of pO2 mapping, software for fitting T1 curves is needed.
There are many good free software options for such applications as
ImageJ (NIH, USA). Other choices include MATLAB (The Math-
Works Inc., Natick, MA, USA), IDL (Exelis Visual Information
Solutions, Boulder, CO, USA), or standard graphing packages such
as Sigmaplot (Systat Software, San Jose, CA, USA).
3 Methods
3.1 Animal Handling All animal work should be carried out only upon review and
and Anesthesia approval of the methods by your institution’s Animal Care and
Use Committee [56, 57]. For those new to MRI, prior to initiating
any studies, training and advice should be sought from experts in
the field.
Anesthetized subject must be maintained in as natural as possi-
ble conditions because tissue oxygenation will be dependent on
many physiological parameters such as inspired O2, blood flow,
vascular reactivity, and heart rate [58]. Thus, proper monitoring
on the subject is crucial to obtain consistent results.
Anesthesia plays a central role in study design. There are several
anesthetics that can be used in these experiments, but it is impor-
tant to bear in mind that all of them have certain impact in the
oxygenation and in blood flow [59, 60]. The final selection will
depend on multiple factors such as the species (rat [61] vs. mouse
[62]), the duration of the experiment [63], etc. As a general rule,
injectable anesthetics provide a stable imaging condition for up to
2 h, whereas inhaled anesthesia provides stable anesthesia level over
a prolonged period [64] and can be adjusted during the course of
the experiment, but also requires additional equipment. With
injectable anesthetics, ketamine-xylazine tends to result in lower
tissue pO2 values [65], whereas propofol and remifentanil have
been reported to have limited effects on perfusion and oxygenation
[66]. Urethane has been reported to render a stronger BOLD
signal [67], but it can be used only in terminal experiments because
it is carcinogenic and compulsory euthanization must be performed
at the end of the experiment.
3.2 DOCENT 1. Place the animal in an anesthetic induction chamber with 3–5%
isoflurane in medical air (1 L/min) and wait until the animal is
sedated. If an injectable anesthetic is used, inject it and wait
until absence of withdrawal reflexes (toe and ear pinch). Addi-
tional doses can be injected depending on the specific protocol
(see Note 8).
2. Transfer the animal to the MRI bed. Check correct positioning
of the animal so the heating system can maintain physiological
temperature.
3. Check that the correct stage of anesthesia has been maintained.
If not, increase the level of anesthetic or inject additional doses.
Maintain anesthesia with isoflurane at 1.5–2%.
4. Insert the rectal temperature probe using lubricating jelly and
tape in place.
5. Place the physiological monitoring device (i.e., MRI compati-
ble sensor with foot clip) or the breathing sensor.
MR Oximetry 303
6. In order to prevent corneal drying, employ ophthalmic gel on

each eye.
7. If a surface or circular coil is used, put it as closed as possible to
the area where DOCENT is to be assessed.
8. Place the animal inside the RF coil. Try to position the area of
interest at the magnet isocenter.
9. Acquire anatomical images in the three orthogonal planes.
10. Using anatomical images acquired in step 9, plan TOLD and
BOLD images (and T1 and T2* maps if required) so they
include the area of interest (i.e., tumor) and normal tissue
(i.e., normal brain or muscle). Slice positioning and geometry
must be maintained through all the subsequent experiments.
See Subheading 2.2.3 and see Note 4 for sequence details.
11. Acquire baseline images during baseline conditions (air breath-
ing). Begin by acquiring T1 and T2* maps, if required. Then
acquire a series of five T1-weighted and five T2*-weighted
images in an interleaved fashion to verify baseline stability or
reveal fluctuations.
12. Proceed to the challenge. In our case, we change the breathing
from air to hyperoxygenated gas (i.e., pure oxygen or carbo-
gen, 1 L/min).
13. Acquire a series of 20–30 interleaved T1- and T2*-weighted
images as previously described. Finally, acquire T1 and T2*
maps if required.
14. Acquire an anatomical image with the same geometry as the
T1- and T2*-weighted images. We recommend using higher
(at least double) in-plane spatial resolution. While this is not
necessary, it will provide anatomical information helping in the
identification of the areas being activated.
A summary of the imaging protocol is presented in Fig. 1.
Fig. 1 Scheme of gas breathing sequence and images acquisition in each experimental condition.
3.3 PFC pO2 Mapping 1. Inject PFC (see Note 9) directly into the tumor with a fine
sharp needle (32G). In order to minimize tissue damage, we
recommend the use of a Hamilton syringe (typically, 100 μL
capacity). To adequately sample the whole tumor’s pO2, inject
the PFC in a fan pattern in a single plane, as described in
[38]. Ideally, the PFC should be saturated with N2 (by gentle
bubbling in the bottle for a few minutes) before injection into
tissue to minimize perturbation of measured pO2 levels.
2. Transfer the animal to the MRI bed and place the physiological
monitoring device as described in steps 2–8 of the DOCENT
protocol.
3. Place the animal in a dual1H/19F coil.
4. Tune the coil to the1H frequency and perform shimming.
5. Acquire anatomical images in the three orthogonal planes.
6. Using anatomical images acquire in step 5 to plan your slice
selection. Slice positioning and geometry must be maintained
through all the subsequent experiments.
7. It is recommended to acquire an anatomical image with the
same geometry than the T119F mapping images. Whether this
is not necessary, it will provide anatomical information helping
in the identification of the areas where we are measuring pO2.
8. Tune the coil to19F frequency and perform further shimming
on the19F signal if needed.
9. Acquire T119F maps to quantify the spin-lattice relaxation rate
of the injected PFC. Obtain a series (3–5) of pO2 measure-
ments for each condition studied. Usually, we acquire five
baseline points (i.e., air breathing), five during the oxygen
challenge (i.e., breathing 100% O2 or carbogen) and five points
during return to baseline conditions or after performing any
other intervention (i.e., administration of some drug).
3.4 Data Analysis For the DOCENT experiment, data analysis can be performed on a
region of interest (ROI) or on a voxel-by-voxel basis. Briefly, signal
intensity (SI) in the baseline (air breathing) T1 and T2* images are
averaged to obtain mean baseline images. Normalized image SI
values are then calculated by following the expression:
SIi, m, n
ΔSI ð%Þi, m, n ¼ P ∗100 100 ð1Þ
5
i¼1 SIi, m, n =5
where i is the image number, m and n are the image matrix, and SI
is the voxel intensity. Similar approach can be used if a ROI analysis
is applied.
If R1 and R2* maps are acquired, R1 values for a given voxel are
obtained by fitting the signal intensities corresponding to different
TRs to a monoexponential function, using a Levenberg-Marquardt
MR Oximetry 305
algorithm. R2* values for a given voxel were obtained by fitting an

exponential model of signal decay curve [68]. Common voxels are
compared under oxygen challenge and air breathing and differences
calculated using the following equations:
ΔR1 ¼ ðR1OxygenChallenge R1Air Þ ð2Þ

ΔR∗
2 ¼ R ∗
2 Oxygen Challenge R ∗
2 Air ð3Þ
Resulting maps can be overlaid on T2 anatomical images

acquired with the same geometry. Figure 2 shows an example of a
DOCENT experiment performed on a rat brain bearing a C6
glioma (see Note 10).
For pO2 mapping using PFC, R1 relaxation rate can be esti-
mated on a voxel-by-voxel basis using a monoexponential function:
τ

=
SI ¼ S 0 1 e T 1 þ k ð4Þ
where SI is signal intensity at τ, S0 represents the original magneti-
zation, and k is a constant.
The pO2 (torr) value for each voxel can be measured using the
following relationship:
R1 R1a
pO2 ¼ ð5Þ
R1p =K
where R1 is the relaxation rate for that particular voxel, R1a is the
anoxic relaxation rate, R1p is the relaxation rate due to the para-
magnetic contribution of oxygen and K is the Henry’s constant.
These parameters must be known or measured for each PFC (each
resonance), temperature and magnetic field. A detailed list of these
parameters can be found in [38]. Figure 3 shows an example of the
results obtained in a dynamic oximetry experiment performed with
HFB in a mouse bearing an orthotopic MDA-MB-231/luc human
breast tumor xenograft implanted in the left upper mammary fat
pad. Figure 4 shows an example of the pO2 maps (using HFB) and
DOCENT results obtained in a nude rat bearing a human lung
tumor line A549 implanted subcutaneously in the thigh.
4 Notes
1. It is important to bear in mind that T2* changes depend mainly

on the oxy/deoxyhemoglobin ratios. The amount of either
form of hemoglobin can vary by changing blood volume,
blood flow, hemoglobin saturation, or a combination of any
of these parameters. Even if we assume that total hemoglobin is
constant (something far from true in some organs, like the
306
Heling Zhou et al.
Fig. 2 DOCENT oximetry. Changes in T1- and T2*-weighted image SI in response to transition from air to oxygen breathing in both healthy brain tissue (C,
contralateral) and high grade gliomas (T, tumor) induced in Wistar rats by stereotaxic injection of C6 cells in the right caudate nucleus. (A, B) T2*-weighted gradient
echo cross-sectional images of the tumor and normal tissue acquired while breathing (A) air (baseline) and (B) oxygen (image selected to show maximum change).
Maps are overlaid on anatomical T2-weighted images. (C) Mean BOLD (normalized SI) response in three adjacent image slices (1–3) for tumors (T SI1, T SI2, T SI3)
and contralateral hemisphere (C SI1, C SI2, C SI3), and mean value over whole area of interest (T Mean for tumor and C Mean for healthy brain tissue). (D, E) T1-
weighted gradient echo images, acquired while breathing (D) air and (E) oxygen. (F) Mean variations across areas of interest of normalized SI change versus time
(TOLD response). Heat scale bar shows percentage change.
Fig. 3 Dynamic oximetry using FREDOM. FREDOM of a SCID mouse bearing an orthotopic MDA-MB-231/luc human breast tumor xenograft implanted in the left
upper mammary fat pad. HFB (50 μL) was injected directly into the tumor in a fan pattern to ensure distribution to various locations and provide pO2 maps
representative of whole tumor. MR was performed using a 4.7 T small animal scanner. Three pO2 measurements were obtained for baseline air, and four with
oxygen breathing challenge. An indole-based vascular disrupting agent (VDA), OXi8007 was administered IP in situ (350 mg/kg) and 18 more pO2 measurements
were obtained over 2 h. In total 25 FREDOM maps were acquired over 3 h. (A) T1 curve fitting for the whole tumor at three representative points: during air
MR Oximetry
breathing, oxygen breathing, and 2 h after administration of VDA while continuing breathing oxygen. (B) Dynamic changes of the mean pO2 values in response to
oxygen breathing challenge and VDA administration. (C) pO2 maps of the tumor during air, oxygen breathing and 2 h after administration of VDA while continuing
breathing oxygen. (D) Histogram showing pO2 distributions at the three stages (OXi8007 kindly provided by Dr. Kevin Pinney, Baylor University, and a
307
comprehensive study is reported in [72]).

Fig. 4 19F and1H MRI oximetry of subcutaneous A549 human tumor xenograft in a nude rat. FREDOM was used
to acquire pO2 maps and their corresponding histograms, while the rat breathed air (A, E) and oxygen (B, F).
TOLD effect is reflected in changes in signal intensity in T1-weighted images while the rat breathed air (C) and
oxygen (D). TOLD approach reflects tissue oxygenation while BOLD reflects blood oxygenation through mean
changes in signal intensity (G). If FREDOM and DOCENT are porfermed in the same animal, it is crucial to perform
DOCENT before PFC administration to avoid interferences in T- and T-weighted images
brain), there is a nonlinear relationship between hemoglobin

saturation and pO2 due to the shape of the oxyhemoglobin
dissociation curve. To make things even more complicated,
there are other factors that will influence this relationship
such as vessel orientation [69], red cell geometry [70], and
magnetic field [71].
2. The gyromagnetic ratio of1H is 42.58 MHz/T and of19F is
40.05 MHz/T; so for a 7 (or 4.7) T study, one needs frequen-
cies of 300 (or 200) MHz for1H and 282.2 (or 188.05) MHz
for19F. R1 is also field dependent, so it needs calibration para-
meters for corresponding B0.
3. For the latter purpose, it could be useful to acquire these
anatomical images with a matrix dimension that are an integer
multiple of the acquired matrix for maps (i.e., if the map is
64 64, acquire a 128 128 anatomical image with the same
geometry).
4. Some studies also incorporate R1 and R2* measurements
before and after the oxygen challenge. In that case, sequences
are required for quantification of T1 and T2*. As an example for
T1 quantification, we use Spin Echo Multiple Slice (SEMS)
sequence with TR arrayed as nine values from 100 to
3500 ms and TE 20 ms. For T2* mapping, a Gradient Echo
MR Oximetry 309
Multiple Slice (GEMS) sequence with TR 195 ms and TE

arrayed as 7 values from 7 to 49 ms with constant echo spacing
of 6 ms was used.
5. It is important because DOCENT signal is partially based on
the changes in the vascular bed and in the blood flow origi-
nated by the oxygen challenge. These vascular responses are
influenced by temperature, being decreased or even eliminated
if the subject is not at physiological temperature.
6. Relaxivity of PFCs is dependent on the oxygenation and also on
the temperature. Calibrations curves are measured at physio-
logical temperature. Thus, in order to transform R1 values to
pO2, it is important to be sure that the animal is at proper
temperature during image acquisition.
7. Circulating liquids can introduce imaging artifacts if the
acquired volume includes the waterbed. Several solutions are
available, as repositioning the waterbed away from the coil,
rotating the phase-encoding plane away from the water to
minimize flow artifacts or applying a spatial saturation band
to suppress the signal from the water bed.
8. If the experiment is carried out under non-inhaled anesthesia, it
is important to take into account that some animal strains are
more susceptible of anesthetics mixture. In this case, we rec-
ommend injecting the animal without previous exposure to
other anesthetic.
9. We favor HFB, but many effective studies have exploited a
perfluorinated crown ether (15C5), as PFC reporter. Both
agents exhibit a single19F resonance and are highly sensitive
to pO2 changes; however 15C5 is considerably more sensitive
to temperature and thus a miss calibration can lead much
greater errors in estimate of pO2.
10. In considering T1 and R1 maps, it is important to note that the
reciprocal of the mean T1 will likely be somewhat different
from the mean of the reciprocal T1s, if voxel-by-voxel analysis
is performed.
Acknowledgements
Method development and application supported in part by CPRIT

RP140399, RP120670-03, P30 CA142543, and P41 EB015908.
References
1. Lassmann H (2016) Demyelination and neu- of hypoxia. Ann Neurol 79(4):520–521.
rodegeneration in multiple sclerosis: The role https://doi.org/10.1002/ana.24632
2. Peers C, Dallas ML, Boycott HE, Scragg JL, importance in tumor biology, noninvasive mea-
Pearson HA, Boyle JP (2009) Hypoxia and surement by imaging, and value of its measure-
neurodegeneration. Ann N Y Acad Sci ment in the management of cancer therapy. Int
1177:169–177. https://doi.org/10.1111/j. J Radiat Biol 82(10):699–757. https://doi.
1749-6632.2009.05026.x org/10.1080/09553000601002324
3. Figueroa XA, Wright JK (2016) Hyperbaric 13. Overgaard J, Horsman MR (1996) Modifica-
oxygen: B-level evidence in mild traumatic tion of hypoxia-induced radioresistance in
brain injury clinical trials. Neurology 87 tumors by the use of oxygen and sensitizers.
(13):1400–1406. https://doi.org/10.1212/ Semin Radiat Oncol 6:10–21
WNL.0000000000003146 14. Matsuo M, Matsumoto S, Mitchell JB, Krishna
4. Kones R (2011) Oxygen therapy for acute MC, Camphausen K (2014) Magnetic reso-
myocardial infarction-then and now. A century nance imaging of the tumor microenvironment
of uncertainty. Am J Med 124 in radiotherapy: perfusion, hypoxia, and
(11):1000–1005. https://doi.org/10.1016/j. metabolism. Semin Radiat Oncol 24
amjmed.2011.04.034 (3):210–217. https://doi.org/10.1016/j.
5. Rink C, Khanna S (2011) Significance of brain semradonc.2014.02.002
tissue oxygenation and the arachidonic acid 15. Nordsmark M, Bentzen SM, Rudat V,
cascade in stroke. Antioxid Redox Signal 14 Brizel D, Lartigau E, Stadler P, Becker A,
(10):1889–1903. https://doi.org/10.1089/ Adam M, Molls M, Dunst J, Terris DJ, Over-
ars.2010.3474 gaard J (2005) Prognostic value of tumor oxy-
6. Pacheco-Torres J, Lopez-Larrubia P, genation in 397 head and neck tumors after
Ballesteros P, Cerdan S (2011) Imaging primary radiation therapy. An international
tumor hypoxia by magnetic resonance meth- multi-center study. Radiother Oncol 77
ods. NMR Biomed 24(1):1–16. https://doi. (1):18–24
org/10.1002/nbm.1558 16. Vaupel P, Höckel M, Mayer A (2007) Detec-
7. Mason RP, Zhao D, Pacheco-Torres J, Cui W, tion and characterization of tumor hypoxia
Kodibagkar VD, Gulaka PK, Hao G, Thorpe P, using pO2 histography. Antioxid Redox Signal
Hahn EW, Peschke P (2010) Multimodality 9(8):1221–1236. https://doi.org/10.1089/
imaging of hypoxia in preclinical settings. Q J ars.2007.1628
Nucl Med Mol Imaging 54(3):259–280 17. Griffiths JR, Robinson SP (1999) The OxyLite:
8. Gray LH, Conger AD, Ebert M, Hornsey S, a fibre-optic oxygen sensor. Br J Radiol 72
Scott OC (1953) The concentration of oxygen (859):627–630
dissolved in tissues at the time of irradiation as a 18. Peeters SG, Zegers CM, Yaromina A, Van
factor in radiotherapy. Br J Radiol 26 Elmpt W, Dubois L, Lambin P (2015) Current
(312):638–648 preclinical and clinical applications of hypoxia
9. Overgaard J (2007) Hypoxic radiosensitiza- PET imaging using 2-nitroimidazoles. Q J
tion: adored and ignored. J Clin Oncol 25 Nucl Med Mol Imaging 59(1):39–57
(26):4066–4074. https://doi.org/10.1200/ 19. Peeters SG, Zegers CM, Lieuwes NG, van
JCO.2007.12.7878 Elmpt W, Eriksson J, van Dongen GA,
10. Tredan O, Grantab R, Dumontet C (2008) Dubois L, Lambin P (2015) A comparative
Can tumor hypoxia be turned into a chemo- study of the hypoxia PET tracers [(1)(8)F]
therapeutic advantage? Bull Cancer 95 HX4, [(1)(8)F]FAZA, and [(1)(8)F]FMISO
(5):528–534. https://doi.org/10.1684/bdc. in a preclinical tumor model. Int J Radiat
2008.0637 Oncol Biol Phys 91(2):351–359. https://doi.
11. Hay MP, Hicks KO, Wang J (2014) Hypoxia- org/10.1016/j.ijrobp.2014.09.045
directed drug strategies to target the tumor 20. Krohn KA, Link JM, Mason RP (2008) Molec-
microenvironment. Adv Exp Med Biol ular imaging of hypoxia. J Nucl Med 49(Suppl
772:111–145. https://doi.org/10.1007/ 2):129S–148S. https://doi.org/10.2967/
978-1-4614-5915-6_6 jnumed.107.045914
12. Tatum JL, Kelloff GJ, Gillies RJ, Arbeit JM, 21. Padhani AR, Krohn KA, Lewis JS, Alber M
Brown JM, Chao KS, Chapman JD, Eckelman (2007) Imaging oxygenation of human
WC, Fyles AW, Giaccia AJ, Hill RP, Koch CJ, tumours. Eur Radiol 17(4):861–872
Krishna MC, Krohn KA, Lewis JS, Mason RP, 22. Foo SS, Abbott DF, Lawrentschuk N, Scott
Melillo G, Padhani AR, Powis G, Rajendran AM (2004) Functional imaging of intratu-
JG, Reba R, Robinson SP, Semenza GL, Swartz moral hypoxia. Mol Imaging Biol 6
HM, Vaupel P, Yang D, Croft B, Hoffman J, (5):291–305
Liu G, Stone H, Sullivan D (2006) Hypoxia:
MR Oximetry 311
23. Ljungkvist AS, Bussink J, Kaanders JH, van der Blood Substit Immobil Biotechnol 22
Kogel AJ (2007) Dynamics of tumor hypoxia (4):1141–1153
measured with bioreductive hypoxic cell mar- 35. Sotak CH, Hees PS, Huang HN, Hung MH,
kers. Radiat Res 167(2):127–145 Krespan CG, Raynolds S (1993) A new per-
24. Vikram DS, Zweier JL, Kuppusamy P (2007) fluorocarbon for use in fluorine-19 magnetic
Methods for noninvasive imaging of tissue hyp- resonance imaging and spectroscopy. Magn
oxia. Antioxid Redox Signal 9(10):1745–1756 Reson Med 29(2):188–195
25. Liu H, Gu Y, Kim JG, Mason RP (2004) Near 36. Jordan BF, Cron GO, Gallez B (2009) Rapid
infrared spectroscopy and imaging of tumor monitoring of oxygenation by 19F magnetic
vascular oxygenation. Methods Enzymol resonance imaging: Simultaneous comparison
386:349–378 with fluorescence quenching. Magn Reson
26. Roussakis E, Li Z, Nichols AJ, Evans CL Med 61(3):634–638. https://doi.org/10.
(2015) Oxygen-sensing methods in biomedi- 1002/mrm.21594
cine from the macroscale to the microscale. 37. Liu S, Shah SJ, Wilmes LJ, Feiner J, Kodibag-
Angew Chem 54(29):8340–8362. https:// kar VD, Wendland MF, Mason RP, Hylton N,
doi.org/10.1002/anie.201410646 Hopf HW, Rollins MD (2011) Quantitative
27. Mason RP (2017) Oxygen breathing chal- tissue oxygen measurement in multiple organs
lenge—the simplest theranostics. Theranostics using 19F MRI in a rat model. Magn Reson
7(16):3873–3875. https://doi.org/10.7150/ Med 66(6):1722–1730. https://doi.org/10.
thno.20655 1002/mrm.22968
28. Charnley N, Donaldson S, Price P (2009) 38. Zhao D, Jiang L, Mason RP (2004) Measuring
Imaging angiogenesis. Methods Mol Biol changes in tumor oxygenation. Methods Enzy-
467:25–51 mol 386:378–418
29. Jordan BF, Runquist M, Raghunand N, 39. Kodibagkar VD, Cui W, Merritt ME, Mason
Baker A, Williams R, Kirkpatrick L, Powis G, RP (2006) Novel 1H NMR approach to quan-
Gillies RJ (2005) Dynamic contrast-enhanced titative tissue oximetry using hexamethyldisi-
and diffusion MRI show rapid and dramatic loxane. Magn Reson Med 55(4):743–748
changes in tumor microenvironment in 40. Kodibagkar VD, Wang X, Pacheco-Torres J,
response to inhibition of HIF-1alpha using Gulaka P, Mason RP (2008) Proton imaging
PX-478. Neoplasia 7(5):475–485 of siloxanes to map tissue oxygenation levels
30. Shukla HP, Mason RP, Woessner DE, Antich (PISTOL): a tool for quantitative tissue oxim-
PP (1995) A comparison of three commercial etry. NMR Biomed 21(8):899–907
perfluorocarbon emulsions as high-field 19F 41. Gulaka PK, Rastogi U, McKay MA, Wang X,
NMR probes of oxygen tension and tempera- Mason RP, Kodibagkar VD (2011)
ture. J Magn Reson B 106(2):131–141. Hexamethyldisiloxane-based nanoprobes for
https://doi.org/10.1006/jmrb.1995.1023 (1) H MRI oximetry. NMR Biomed 24
31. Ali Hamza MH, Serratrice G, Stébé M-J, Del- (10):1226–1234. https://doi.org/10.1002/
puech J-J (1981) Fluorocarbons as oxygen car- nbm.1678
riers. II. An NMR study of partially or totally 42. Greve JM (2011) The BOLD effect. Methods
fluorinated alkanes and alkenes. J Magn Reson Mol Biol 771:153–169. https://doi.org/10.
(1969) 42(2):227–241. https://doi.org/10. 1007/978-1-61779-219-9_8
1016/0022-2364(81)90214-6 43. Nield LE, Qi X, Yoo SJ, Valsangiacomo ER,
32. Thomas SR, Millard RW, Pratt RG, Shiferaw Y, Hornberger LK, Wright GA (2002)
Samaratunga RC (1994) Quantitative pO2 MRI-based blood oxygen saturation measure-
imaging in vivo with perfluorocarbon F-19 ments in infants and children with congenital
NMR: tracking oxygen from the airway heart disease. Pediatr Radiol 32(7):518–522.
through the blood to organ tissues. Artif Cells https://doi.org/10.1007/s00247-001-0652-
Blood Substit Immobil Biotechnol 22 9
(4):1029–1042 44. Ndubuizu O, La Manna JC (2007) Brain tissue
33. Mason RP, Shukla H, Antich PP (1993) In vivo oxygen concentration measurements. Antioxid
oxygen tension and temperature: simultaneous Redox Signal 9(8):1207–1219
determination using 19F NMR spectroscopy of 45. Dunn JF (2011) MR oximetry. Methods Mol
perfluorocarbon. Magn Reson Med Biol 771:227–240. https://doi.org/10.1007/
29:296–302 978-1-61779-219-9_12
34. Mason RP (1994) Non-invasive physiology: 46. Howe FA, Robinson SP, McIntyre DJ,
19F NMR of perfluorocarbons. Artif Cells Stubbs M, Griffiths JR (2001) Issues in flow
and oxygenation dependent contrast
(FLOOD) imaging of tumours. NMR Biomed models. Cancer Res 76(4):787–795. https://
14(7–8):497–506 doi.org/10.1158/0008-5472.CAN-15-2062
47. Baudelet C, Gallez B (2005) Current issues in 56. Tennant DA, Duran RV, Gottlieb E (2010)
the utility of blood oxygen level dependent Targeting metabolic transformation for cancer
MRI for the assessment of modulations in therapy. Nat Rev Cancer 10(4):267–277.
tumor oxygenation. Curr Med Imaging Rev 1 https://doi.org/10.1038/nrc2817
(3):229–243. https://doi.org/10.2174/ 57. European Convention for the Protection of
157340505774574754 vertebrate animals used for experimental and
48. Al-Hallaq HA, River JN, Zamora M, other scientific purposes (2006) Appendix
Oikawa H, Karczmar GS (1998) Correlation A. Guidelines for accommodation and care of
of magnetic resonance and oxygen microelec- animals (Article 5 of the Convention)
trode measurements of carbogen-induced 58. Sicard KM, Duong TQ (2005) Effects of hyp-
changes in tumor oxygenation. Int J Radiat oxia, hyperoxia, and hypercapnia on baseline
Oncol Biol Phys 41(1):151–159 and stimulus-evoked BOLD, CBF, and
49. Zhao D, Jiang L, Hahn EW, Mason RP (2009) CMRO2 in spontaneously breathing animals.
Comparison of1H blood oxygen level- NeuroImage 25(3):850–858. https://doi.
dependent (BOLD) and19F MRI to investigate org/10.1016/j.neuroimage.2004.12.010
tumor oxygenation. Magn Reson Med 62 59. Masamoto K, Kanno I (2012) Anesthesia and
(2):357–364 the quantitative evaluation of neurovascular
50. Hallac RR, Zhou H, Pidikiti R, Song K, coupling. J Cereb Blood Flow Metab 32
Stojadinovic S, Zhao D, Solberg T, Peschke P, (7):1233–1247. https://doi.org/10.1038/
Mason RP (2014) Correlations of noninvasive jcbfm.2012.50
BOLD and TOLD MRI with pO2 and rele- 60. Duong TQ (2007) Cerebral blood flow and
vance to tumor radiation response. Magn BOLD fMRI responses to hypoxia in awake
Reson Med 71(5):1863–1873. https://doi. and anesthetized rats. Brain Res 1135
org/10.1002/mrm.24846 (1):186–194. https://doi.org/10.1016/j.
51. Matsumoto K, Bernardo M, Subramanian S, brainres.2006.11.097
Choyke P, Mitchell JB, Krishna MC, Lizak MJ 61. Hendrich KS, Kochanek PM, Melick JA, Schid-
(2006) MR assessment of changes of tumor in ing JK, Statler KD, Williams DS, Marion DW,
response to hyperbaric oxygen treatment. Ho C (2001) Cerebral perfusion during anes-
Magn Reson Med 56(2):240–246. https:// thesia with fentanyl, isoflurane, or pentobarbi-
doi.org/10.1002/mrm.20961 tal in normal rats studied by arterial spin-
52. Beeman SC, Shui YB, Perez-Torres CJ, Engel- labeled MRI. Magn Reson Med 46
bach JA, Ackerman JJ, Garbow JR (2016) (1):202–206
O2-sensitive MRI distinguishes brain tumor 62. Schroeter A, Schlegel F, Seuwen A,
versus radiation necrosis in murine models. Grandjean J, Rudin M (2014) Specificity of
Magn Reson Med 75(6):2442–2447. https:// stimulus-evoked fMRI responses in the
doi.org/10.1002/mrm.25821 mouse: the influence of systemic physiological
53. O’Connor JP, Naish JH, Jackson A, Waterton changes associated with innocuous stimulation
JC, Watson Y, Cheung S, Buckley DL, under four different anesthetics. NeuroImage
McGrath DM, Buonaccorsi GA, Mills SJ, 94:372–384. https://doi.org/10.1016/j.
Roberts C, Jayson GC, Parker GJ (2009) Com- neuroimage.2014.01.046
parison of normal tissue R1 and R*2 modula- 63. Sonnay S, Just N, Duarte JM, Gruetter R
tion by oxygen and carbogen. Magn Reson (2015) Imaging of prolonged BOLD response
Med 61(1):75–83 in the somatosensory cortex of the rat. NMR
54. Berkowitz BA, McDonald C, Ito Y, Tofts PS, Biomed 28(3):414–421. https://doi.org/10.
Latif Z, Gross J (2001) Measuring the human 1002/nbm.3263
retinal oxygenation response to a hyperoxic 64. Liu ZM, Schmidt KF, Sicard KM, Duong TQ
challenge using MRI: eliminating blinking arti- (2004) Imaging oxygen consumption in fore-
facts and demonstrating proof of concept. paw somatosensory stimulation in rats under
Magn Reson Med 46(2):412–416 isoflurane anesthesia. Magn Reson Med 52
55. O’Connor JP, Boult JK, Jamin Y, Babur M, (2):277–285. https://doi.org/10.1002/
Finegan KG, Williams KJ, Little RA, mrm.20148
Jackson A, Parker GJ, Reynolds AR, Waterton 65. Lei H, Grinberg O, Nwaigwe CI, Hou HG,
JC, Robinson SP (2016) Oxygen-enhanced Williams H, Swartz HM, Dunn JF (2001) The
MRI accurately identifies, quantifies, and effects of ketamine-xylazine anesthesia on cere-
maps tumor hypoxia in preclinical cancer bral blood flow and oxygenation observed
MR Oximetry 313
using nuclear magnetic resonance perfusion 69. Zhang N, Yacoub E, Zhu XH, Ugurbil K,
imaging and electron paramagnetic resonance Chen W (2009) Linearity of blood-
oximetry. Brain Res 913(2):174–179 oxygenation-level dependent signal at micro-
66. Mikkelsen ML, Ambrus R, Miles JE, Poulsen vasculature. NeuroImage 48(2):313–318.
HH, Moltke FB, Eriksen T (2016) Effect of https://doi.org/10.1016/j.neuroimage.
propofol and remifentanil on cerebral perfu- 2009.06.071
sion and oxygenation in pigs: a systematic 70. Dunn JF, Swartz HM (1997) Blood oxygena-
review. Acta Vet Scand 58(1):42. https://doi. tion. Heterogeneity of hypoxic tissues moni-
org/10.1186/s13028-016-0223-6 tored using bold MR imaging. Adv Exp Med
67. Paasonen J, Salo RA, Shatillo A, Forsberg MM, Biol 428:645–650
Narvainen J, Huttunen JK, Grohn O (2016) 71. Silvennoinen MJ, Clingman CS, Golay X,
Comparison of seven different anesthesia pro- Kauppinen RA, van Zijl PC (2003) Compari-
tocols for nicotine pharmacologic magnetic son of the dependence of blood R2 and R2* on
resonance imaging in rat. Eur Neuropsycho- oxygen saturation at 1.5 and 4.7 Tesla. Magn
pharmacol 26(3):518–531. https://doi.org/ Reson Med 49(1):47–60. https://doi.org/10.
10.1016/j.euroneuro.2015.12.034 1002/mrm.10355
68. Zhao D, Pacheco-Torres J, Hallac RR, 72. Zhou H, Hallac RR, Lopez R, Denney R, Mac-
White D, Peschke P, Cerdan S, Mason RP Donough MT, Li L, Liu L, Graves EE, Trawick
(2015) Dynamic oxygen challenge evaluated ML, Pinney KG, Mason RP (2015) Evaluation
by NMR T1 and T2*—insights into tumor of tumor ischemia in response to an indole-
oxygenation. NMR Biomed 28(8):937–947. based vascular disrupting agent using BLI and
https://doi.org/10.1002/nbm.3325 (19)F MRI. Am J Nucl Med Mol Imaging 5
(2):143–153
Chapter 19
Molecular Magnetic Resonance Imaging (mMRI)

Maxime Gauberti, Antoine P. Fournier, Denis Vivien,
and Sara Martinez de Lizarrondo
Abstract
Molecular magnetic resonance imaging (mMRI) enables the detection of a protein of interest in vivo, in a
noninvasive manner. The general concept of mMRI is to target a contrast agent to a protein of interest, and
to perform a contrast-sensitive MRI sequence. Typically, contrast agents are made of a “contrastophore”
(the part of the construct responsible for the contrast on the images) and a targeting moiety (“pharmaco-
phore”). Recently, the development of a new family of contrastophore carrying a high payload of iron oxide
(micro-sized particles of iron oxide, MPIO) has led to a dramatic increase in the sensitivity of mMRI. Here,
we describe the production of targeted MPIO using commercially available reagents and the MRI protocols
to allow their detection in vivo.
Key words Molecular imaging, Inflammation, Vascular-cell adhesion molecule, Intercellular adhesion
molecule, Selectins, Leucocytes, USPIO, Microparticles, Endothelial cells, Endothelium, Platelets
1 Introduction
Magnetic resonance imaging (MRI) appears particularly well suited

for molecular imaging experiments. It offers high spatial resolution
(up to ~50 μm isotropic resolution for high-field MRI), very good
soft tissue contrast, and a virtually unlimited depth penetration for
preclinical imaging. Moreover, it does not involve ionizing radia-
tion, making it particularly relevant for longitudinal follow-up
involving multiple acquisitions. However, it displays limited sensi-
tivity to exogenous contrast agents. Indeed, whereas positron emis-
sion tomography (PET) can detect βþ emitting atoms at picomolar
concentration, MRI presents a sensitivity in the micromolar range.
Since very few targets present such a high concentration in living
organisms, the development of amplification techniques was man-
datory to achieve reliable molecular MRI (mMRI).
To this aim, larger particles (in the 10–100 nanometer range)
carrying large amount of contrast agent have been developed, such
as ultrasmall particles of iron oxide (USPIO) [1]. This method
315
316 Maxime Gauberti et al.
allows the binding of numerous contrast-producing atoms to a

single molecular target and therefore to reach sensitivity in the
nanomolar range. Still, the contrast agent needs to reach its target
after its administration [2]. Thus, the use of larger particles that are
unable to cross the endothelial barrier limited the use of such agents
in pathological conditions with severely impaired endothelium,
such as cancer, multiple sclerosis, or stroke [3–6]. Unfortunately,
in these conditions, the distinction between the “enhanced perme-
ability and retention effect” (EPR effect) [7] and the actual binding
of the contrast agent to its target remains challenging to perform.
Regarding targets expressed by the endothelium, mMRI enhanced
by USPIO or similar contrast agents displays a sensitivity too low
for reliable imaging, despite significant efforts made in that direc-
tion by the mMRI community. These drawbacks may explain why
the use of USPIO-based targeted contrast agents remains restricted
to in vitro or preclinical proof of concept studies without any
clinical applications to date.
In this context, a new family of contrast carrying particles has
been developed, known as micro-sized particles of iron oxide
(MPIO), which uses even larger particles (in the micrometer
range) [8]. These MPIOs convey a large payload of iron oxide
(0.1–1.6 pg iron/MPIO particle), which is an order of magnitude
larger than USPIO-based contrast agents. Thanks to this high
payload of iron oxide, MPIO results in strong hypointense con-
trast effects on T2*-weighted images that extend up to 50 times
the physical diameter of the particle (i.e., 50 μm for a single
MPIO) [9]. This phenomenon, known as “blooming effect,”
provides high sensitivity for in vivo MPIO detection by MRI.
Notably, when using a spatial resolution close to 50 μm, a single
particle can be detected [10]. Since the first description of the
possibility to image vascular-cell adhesion molecule-1 (VCAM-1)
using targeted MPIOs in acute brain inflammation, several studies
demonstrated the wide applicability of this mMRI method for
other targets (intercellular adhesion molecule-1, P-Selectin, integ-
rins, platelet receptors) [4, 11, 12] and in other organs (brain,
heart, kidney, solid tumors) [13, 14]. Although limited to endo-
thelial targets because of their large size, targeted MPIOs may
have large preclinical and potentially clinical applications (Fig. 1)
[13, 14].
In the present chapter, we will first explain how to produce
targeted MPIOs using commercially available reagents. Then, we
will describe some MRI sequences to reveal MPIOs in vivo in the
brain, heart, and kidney.
Fig. 1 Schematic representation and in vivo mMRI of VCAM-1 (vascular cell adhesion molecule-1) in the brain, kidney, and heart using MPIOs (microparticles of
Molecular MRI
iron oxide) labeled with a rat monoclonal antibody against VCAM-1 (MPIO-αVCAM-1) in a mouse model of sepsis (intraperitoneal LPS injection, 1 mg/kg). MPIOs
bind to activated endothelial cells (VCAM-1 positive vessels) as presented in the immunofluorescence images of the upper panels. Lower panels: mMRI allows
detection of endothelial activation (red images), whereas no binding is detectable in healthy organs (green images)
317
2 Materials
Prepare all solutions using ultrapure water (prepared by purifying

deionized water, to attain a resistivity of 18 MΩ cm at 25 C) and
analytical grade reagents. Prepare and store all reagents at room
temperature (unless indicated otherwise). We do not add sodium
azide to reagent.
2.1 Contrastophores 1. One micrometer sized MPIOs. Dynabeads® Myone Tosylacti-

and Pharmacophores vated (Thermo Fisher) (see Note 1). These MPIOs present
active surfaces (tosylactivated) which are able to bind primary
amines and sulfhydryl groups of pharmacophores, such as
monoclonal antibodies [15].
2. Antibodies for MPIO targeting. All antibodies are potentially
suitable, as long as they bind an extracellular protein and that
the epitope is also extracellular (i.e., not intramembranous or
intracellular). To date, positive results were obtained using
mouse VCAM-1 targeted monoclonal antibodies (Rat anti-
mouse VCAM-1, clone A(429); BD-Bioscience) [10, 14, 16,
17], rat VCAM-1 targeted monoclonal antibodies (Mouse
anti-rat VCAM-1, MR106, BD-Bioscience), mouse P-selectin
targeted polyclonal antibodies (Goat anti-mouse P-Selectin,
AF737, R&D Systems) [18] or monoclonal antibodies (Rat
anti-mouse P-Selectin, Clone RB40.3, BD-Bioscience) and
mouse ICAM-1 (intercellular adhesion molecule-1) targeted
monoclonal antibodies (Rat anti-mouse ICAM-1, YN1/1.7.4,
Biolegend) [11, 12]. For untargeted MPIOs, use an isotype
matched immunoglobulin (see Note 2).
2.2 Buffers 1. Borate buffer. Prepare 1 L of a 0.1 M borate

for Targeted MPIOs (M.W. 61.83 g mol1) buffer by weighting 6.183 g of sodium
Production borate in a glass beaker and add water to a volume of 900 mL.
Adjust pH to 9.5 using 10 M NaOH. Make up to 1 L with
water. Store at 4 C.
2. 10% Tween®-PBS buffer. First prepare 200 mL of phosphate
buffered saline (PBS) at pH 7.4 (to this aim, we use PBS
tablets, Sigma-Aldrich). Then mix 1 mL of Tween®-20
(Sigma-Aldrich) to 9 mL of PBS. Store at 4 C.
3. Blocking buffer. Prepare 100 mL of blocking buffer by weight-
ing 0.5 g of bovine serum albumin in 100 mL of PBS. Then,
add 500 μL of the 10% Tween®-PBS buffer. Store at 4 C for
immediate use or make 10 mL aliquots and store at 20 C for
longer storage duration.
4. Washing buffer. Prepare 100 mL of blocking buffer by weight-
ing 0.1 g of bovine serum albumin in 100 mL of PBS. Then,
add 500 μL of the 10% Tween®-PBS buffer. Store at 4 C for
Molecular MRI 319
immediate use or make 20 mL aliquots and store at 20 C for

longer storage duration.
5. Ammonium sulfate buffer. Prepare 100 mL of a 3 M ammo-
nium sulfate (M.W. 132.1 g mol1) by weighting 39.6 g of
ammonium sulfate in 80 mL of borate buffer (do not worry if
the dissolution is not perfect after 10 min of stirring). Adjust
pH to 9.5 using solid pellets of NaOH. When the pH reaches
9.5, all ammonium sulfate should be dissolved. Complete to
100 mL using borate buffer. Store at 4 C for immediate use or
make 5 mL aliquots and store at 20 C for longer storage
duration.
2.3 Experimental 1. Stereotaxic Frame (Standard Rat and mouse Stereotaxic).

Surgery 2. Anesthesia Inducer: Isoflurane Vaporizer and Oxygen/Nitrous
Oxide gas blender.
3. Residual Anesthesia Cartridge trap (HALOSORB Cartridges;
Minerve, Esternay, France).
4. Mice Temperature Control Unit (Temperature Control Unit
HB 101/02 RS, Panlab Harvard Apparatus).
5. Surgical Operating Microscope (Leica M80 StereopMicro-
scope coupled to 16/15 Leica Lenses).
6. Micro dissecting scissors.
7. Curved fine tips Forceps.
8. Mouse tail vein Catheter.
2.4 Magnetic 1. Bruker Pharmascan 7T MRI.

Resonance Imaging 2. MRI holder system (Autopac mice bed).
(MRI)
3. Mouse Brain Surface Coil (RF SUC 500 IH M.BR QSN RO
QD, Bruker).
4. Anesthesia and breathing monitoring (Small Animal Monitor-
ing and Gating system).
5. Operation and Acquisition Manual (Paravision 6.0.1).
2.5 Perfusion- 1. Peristaltic tubing Pump (Cole-Palmer Instrument Company).

Immunohistology 2. Cryomicrotome (Leica CM 3050S Research Cryostat).
3. Microtome blades (MX35 Ultra Microtome Blades,
ThermoScientific).
4. Poly-Lysine slides.
5. Coverslides.
6. Fluorescence microscope (Leica DM 6000B System).
7. Microscopy Automation and Image Analysis Software
(Metamorph).
2.6 Other Materials 1. Vortex.

2. Separating magnet (PureProteome™ Magnetic Stand, 15 mL
from Millipore).
3. Sonicator (Labsonic U sonicator, B. Braun).
4. Rotating Hybridization oven/shaker incubator.
5. Tuber Roller.
3 Methods
Carry out all procedures at room temperature unless otherwise

specified.
3.1 Production 1. Resuspend Dynabeads® Myone Tosylactivated (Thermo

of Targeted Fisher) which may have sedimented in the vial by vortexing
and Control MPIOs during 5 min at room temperature.
2. Add 100 μL of the MPIO solution (Dynabeads® Myone Tosy-
lactivated, Thermo Fisher) in a 15 mL Falcon tube (see Note
3).
3. Wash the MPIOs by adding 5 mL of borate buffer to the
15 mL Falcon tube containing the MPIOs and vortex for
30 s. Place the Falcon tube on a separating magnet (such as
the PureProteome™ Magnetic Stand, 15 mL from Millipore)
for at least 1 min, allowing the beads to separate from the
borate buffer (Fig. 2). Remove and discard the supernatant.
Repeat this washing step three times. At the end, you should
obtain a small amount of purified MPIO accumulated on one
side of the Falcon Tube.
Fig. 2 Magnetic separation of MPIOs using a portable magnet. Left: before

magnetic separation. Right: after magnetic separation
Molecular MRI 321
4. Add 200 μg of your pharmacophore (typically a monoclonal

antibody) to the purified MPIOs, then add 1500 μL of the
ammonium sulfate buffer and 2600 μL of borate buffer.
Depending on the concentration of your pharmacophore solu-
tion (see Note 4), the total volume should approximate
4500 μL (see Note 5).
5. Incubate for 48 h under constant agitation at 37 C to allow
binding of your pharmacophore to the MPIOs (see Note 6).
6. Place the 15 mL Falcon tube containing the beads on a separ-
ating magnet for at least 1 min. Remove and discard the
supernatant.
7. Add 5 mL of the blocking buffer and incubate for 24 h under
constant agitation at 37 C. This step allows blockade of the
remaining active site on the MPIO surface by adding a saturat-
ing concentration of bovine serum albumin.
8. At the end of the incubation, sonicate the beads for 60 s under
constant chilling. This step is necessary to break any conjugated
MPIO aggregates which may have formed during the complex-
ation process. The sonication intensity should be adapted to
induce beads agitation without leading to protein
denaturation.
9. Wash the conjugated MPIOs: Place the Falcon tube on a
separating magnet for at least 1 min, allowing MPIOs separa-
tion from the blocking buffer. Remove and discard the super-
natant. Add 5 mL of washing buffer to the 15 mL Falcon tube
containing the conjugated MPIOs and vortex for 30 s. Repeat
this washing step three times.
10. Finally, suspend the conjugated MPIOs in 8 mL of washing
buffer and store at 4 C under constant agitation (see Note 7).
3.2 Injection 1. Animal handling should be performed according to institu-

of MPIOs In Vivo tional guidelines and after ethical approval of the protocol.
2. To allow successively (1) pre-contrast imaging, (2) intravenous
injection of the contrast agent while inside the MRI magnet,
and (3) post-contrast imaging, we implant a catheter in the
mouse tail vein (see Note 8). Most mouse strains are suitable for
the described protocol.
3. Start by inducing anesthesia using standard protocols. We use
5% isoflurane administered in a mixture of 20%/80% O2/N2O
for 2 min. Anesthesia is thereafter maintained using 1.5–2%
isoflurane administered through a face mask throughout the
procedure [19]. HALOSORB cartridges were used to trap
residual anesthesia particles and protect the operator against
residual anesthetic. Besides, mice’s temperature during all
phases of the surgical procedure was monitored.
4. After securing the mouse and its tail, locate the tail veins (one
main vein is present on each side of the tail). Under an
operating microscope, incise the skin of the tail laterally using
a scalpel. Dissect the tail vein using micro-scissors and curved
fine tips forceps.
5. Once isolated from the adjacent tissue, the superior part of the
vein should be cut using the micro-scissors allowing insertion
of a tail vein catheter in the venous lumen (SAI Infusion
technologies, Mouse Tail Vein Catheter, MTV-01) (see Note
9) [20].
6. Success of the procedure can be ascertained by two methods:
either intravenous injection of a small amount of sterile saline
(100 μL), which should not induce any swelling of the tail, or
visualization of a blood reflow inside the catheter (that is not
constantly observed despite adequate placement of the cathe-
ter) (see Note 10).
7. Once in place, the tail vein catheter should be secured on the
tail skin using medical plaster.
8. If pre-contrast imaging is required, it should be done at
this step.
9. Slowly inject 200 μL per mouse of the MPIOs solution
through the tail vein catheter. Then, rinse the catheter using
100 μL saline.
10. A delay of 5 min between injection and imaging should be
respected to allow MPIOs binding to their target and their
clearance from the blood by the reticuloendothelial system.
The mouse is then ready for post-contrast imaging.
3.3 In Vivo mMRI 1. Place the animal in the MRI holder system in order to avoid, as
much as possible, any movement of the imaged parts. Place a
pressure captor on the animal body to monitor breath rate and
anesthesia.
2. Both surface and volume coils are suitable for mMRI applica-
tions. However, surface coils allow a better signal-to-noise ratio
at the expense of a reduced field of view. We use a Bruker
Pharmascan 7 T magnet equipped with a mouse brain
surface coil.
3. If possible, maintain anesthesia throughout MRI acquisition
using 100% O2. This reduces the deoxyhemoglobin of the veins
into oxyhemoglobin. Since oxyhemoglobin is devoid of mag-
netic susceptibility effects, this allows to specifically observe the
susceptibility effects of the injected MPIOs [21, 22].
4. Brain imaging (Fig. 3): after standard shimming and localizer
scan, to visualize MPIOs, we use a coronal 3D-GEFC (gradient
echo with flow compensation) sequence with the following
Molecular MRI 323
Fig. 3 Representative mMRI of VCAM-1 in the brain in four experimental models showing the wide diversity of
labeling pattern according to the inflammatory stimulus
parameters: repetition time ¼ 200 ms, echo time ¼ 13.2 ms,

flip angle ¼ 22 , bandwidth ¼ 20,000 Hz, acquisition
matrix ¼ 256 * 256 * 24, field of view ¼ 17.92 * 17.92 *
1.68 mm, spatial resolution ¼ 70 μm (isotropic), acquisition
time ¼ 20 min 30 s. This sequence only allows partial covering
of the brain but is sensitive enough to detect single MPIO
[10, 16]. To increase spatial coverage, the field of view and
the acquisition matrix should be increased in the same propor-
tion to preserve the spatial resolution.
5. Kidney imaging: after standard shimming and localizer scan, to
visualize MPIOs, we use an axial 3D-FLASH (Fast low angle
shot) with fat suppression sequence with the following para-
meters: respiratory gating (minimal repetition time ¼ 200 ms),
echo time ¼ 8.313 ms, flip angle ¼ 18 , band-
width ¼ 25,000 Hz, acquisition matrix ¼ 256 * 192 * 12,
field of view ¼ 25.6 * 19.2 * 1.8 mm, spatial resolu-
tion ¼ 100 * 100 * 150 μm, minimal acquisition time ¼ 7 min
41 s. This sequence only allows partial covering of the kidneys
(both kidneys are visible on the same axial plane). To increase
spatial coverage, the field of view and the acquisition matrix
should be increased in the same proportion to preserve the
spatial resolution.
6. Cardiac imaging: After standard shimming and localizer scan,
to visualize MPIOs, we use an axial 3D-GEFC (gradient echo
with flow compensation) sequence with the following para-
meters: respiratory and cardiac gating (minimal repetition
time ¼ 50 ms, acquisition is only triggered during the end
diastole), echo time ¼ 10.16 ms, flip angle ¼ 20 , band-

width ¼ 27,778 Hz, acquisition matrix ¼ 256 * 192 * 12,
field of view ¼ 20.48 * 20.48 * 1.68 mm, spatial resolu-
tion ¼ 80 * 107 * 140 μm, minimal acquisition time ¼ 1 min
55 s. This sequence only allows partial covering of the heart
[14]. To increase spatial coverage, the field of view and the
acquisition matrix should be increased in the same proportion
to preserve the spatial resolution.
3.4 Visualization 1. Perfuse transcardially deeply anesthetized mice with cold hepa-
of MPIOs Using rinized saline (15 mL) followed by 150 mL of fixative (PBS
Microscopy 0.1 M. pH 7.4 containing 2% paraformaldehyde and 0.2%
picric acid or alternatively 4% paraformaldehyde without picric
acid) [23] using a peristaltic tubing pump.
2. (Facultative) Brain, kidney, or heart can be post-fixed overnight
in the same fixative solution (paraformaldehyde with or with-
out picric acid) at 4 C.
3. The collected organs should be cryoprotected using 20%
sucrose in PBS buffer during 24 h at 4 C before freezing in
Tissue-Tek (Miles Scientific, Naperville, IL, USA).
4. Using a cryomicrotome, cut the organs into 8–10 μm slices and
collect them on poly-lysine slides. If required, the slides can be
stored at 80 C until further use.
5. To allow direct white-light detection of MPIOs, washed sec-
tions can be directly coverslipped with antifade medium and
observed using a microscope. In that case, MPIOs appear as
~1 μm black and white spheres (usually a white dot surrounded
by a black ring, Fig. 4) [16].
6. To allow detection of MPIO using fluorescence microscopy,
co-incubate the sections on the slides with a fluorescently
labeled anti-pharmacophore antibody (usually, a secondary
anti-IgG antibody) for at least 90 min at ambient temperature
(see Note 11). In that case, MPIOs appear as ~1 μm fluorescent
spheres (Fig. 1) [24].
Fig. 4 Representative microphotograph of MPIO using white light microscopy.

MPIOs appear are small white dots surrounded by black rings
Molecular MRI 325
4 Notes
1. Alternatively, larger MPIOs could be used such as Dynabeads®

M-280 Tosylactivated (Thermo Fisher) or Dynabeads® M-450
Tosylactivated (Thermo Fisher) that have a diameter of 2.8 μm
and 4.5 μm, respectively. These MPIOs are easier to detect by
MRI given their larger iron content, but may plug into the
smaller vessels.
2. For instance, for MPIO-αVCAM-1 (monoclonal rat anti-
mouse VCAM-1, clone A(429)), we used purified Rat IgG2a,
κ Isotype Control (BD-biosciences).
3. Larger production batches can be produced simply by upscal-
ing the quantities described.
4. It is important to keep your pharmacophore solution free from
any molecules that may bind the activated surface of the
MPIOs (such as ethanolamine, tris-buffer, sulfhydryl groups,
etc.). PBS buffers works well. If necessary, perform a buffer
exchange using dialysis cassettes against a 0.1 M borate buffer
(pH 9.5).
5. If your pharmacophore solution is very diluted, it is mandatory
to reduce the added volume of borate buffer in order to keep
the final volume around 4500 μL.
6. We use a rotating hybridization oven/shaker incubator
(SI 20H, Stuart Scientific) to control the temperature, but
other methods may also be suitable.
7. We store MPIOs batches for a maximum of 30 days after
production at 4 C under rotation. Longer storage duration
may alter MPIOs binding to their target and lead to microor-
ganism development.
8. Direct intravenous injection is also a suitable way to administer
MPIOs.
9. We use cotton swabs to remove the blood from the wounded
area in the operator field of view, while inserting the mouse tail
vein catheter a few centimeters into the tail vein. Be careful to
remove any air gas bubbles from the catheter before its
insertion.
10. In case of failure, the catheter should be completely removed
and replaced in a more proximal location. Alternatively, the
second tail vein could be used.
11. For instance, for MPIO-αVCAM-1, we use Fab’2 fragments of
Donkey anti-rat linked to FITC (fluorescein isothiocyanate)
from Jackson Immunoresearch Laboratories.
References
1. Laurent S, Forge D, Port M, Roch A, Robic C, iron oxide particles. Magn Reson Med 53
Vander Elst L, Muller RN (2008) Magnetic (2):329–338. https://doi.org/10.1002/
iron oxide nanoparticles: synthesis, stabiliza- mrm.20342
tion, vectorization, physicochemical character- 10. Montagne A, Gauberti M, Macrez R,
izations, and biological applications. Chem Rev Jullienne A, Briens A, Raynaud JS, Louin G,
108(6):2064–2110. https://doi.org/10. Buisson A, Haelewyn B, Docagne F, Defer G,
1021/cr068445e Vivien D, Maubert E (2012) Ultra-sensitive
2. Mahmoudi M, Sheibani S, Milani AS, molecular MRI of cerebrovascular cell activa-
Rezaee F, Gauberti M, Dinarvand R, Vali H tion enables early detection of chronic central
(2015) Crucial role of the protein corona for nervous system disorders. NeuroImage 63
the specific targeting of nanoparticles. Nano- (2):760–770. https://doi.org/10.1016/j.
medicine (Lond) 10(2):215–226. https://doi. neuroimage.2012.07.018
org/10.2217/nnm.14.69 11. Deddens LH, van Tilborg GA, van der
3. Le Behot A, Gauberti M, Martinez De Toorn A, van der Marel K, Paulis LE, van
Lizarrondo S, Montagne A, Lemarchand E, Bloois L, Storm G, Strijkers GJ, Mulder WJ,
Repesse Y, Guillou S, Denis CV, Maubert E, de Vries HE, Dijkhuizen RM (2013) MRI of
Orset C, Vivien D (2014) GpIbalpha-VWF ICAM-1 upregulation after stroke: the impor-
blockade restores vessel patency by dissolving tance of choosing the appropriate target-
platelet aggregates formed under very high specific particulate contrast agent. Mol Imag-
shear rate in mice. Blood 123 ing Biol 15(4):411–422. https://doi.org/10.
(21):3354–3363. https://doi.org/10.1182/ 1007/s11307-013-0617-z
blood-2013-12-543074 12. Blezer EL, Deddens LH, Kooij G, Drexhage J,
4. Gauberti M, Montagne A, Quenault A, Vivien van der Pol SM, Reijerkerk A, Dijkhuizen RM,
D (2014) Molecular magnetic resonance imag- de Vries HE (2015) In vivo MR imaging of
ing of brain-immune interactions. Front Cell intercellular adhesion molecule-1 expression
Neurosci 8:389. https://doi.org/10.3389/ in an animal model of multiple sclerosis. Con-
fncel.2014.00389 trast Media Mol Imaging 10(2):111–121.
5. Vivien D, Gauberti M, Montagne A, Defer G, https://doi.org/10.1002/cmmi.1602
Touze E (2011) Impact of tissue plasminogen 13. Melemenidis S, Jefferson A, Ruparelia N, Akh-
activator on the neurovascular unit: from clini- tar AM, Xie J, Allen D, Hamilton A, Larkin JR,
cal data to experimental evidence. J Cereb Perez-Balderas F, Smart SC, Muschel RJ,
Blood Flow Metab 31(11):2119–2134. Chen X, Sibson NR, Choudhury RP (2015)
https://doi.org/10.1038/jcbfm.2011.127 Molecular magnetic resonance imaging of
6. Serres S, Soto MS, Hamilton A, McAteer MA, angiogenesis in vivo using polyvalent cyclic
Carbonell WS, Robson MD, Ansorge O, RGD-iron oxide microparticle conjugates.
Khrapitchev A, Bristow C, Balathasan L, Theranostics 5(5):515–529. https://doi.org/
Weissensteiner T, Anthony DC, Choudhury 10.7150/thno.10319
RP, Muschel RJ, Sibson NR (2012) Molecular 14. Belliere J, Martinez de Lizarrondo S, Choudh-
MRI enables early and sensitive detection of ury RP, Quenault A, Le Behot A, Delage C,
brain metastases. Proc Natl Acad Sci U S A Chauveau D, Schanstra JP, Bascands JL,
109(17):6674–6679. https://doi.org/10. Vivien D, Gauberti M (2015) Unmasking
1073/pnas.1117412109 Silent Endothelial Activation in the Cardiovas-
7. Iyer AK, Khaled G, Fang J, Maeda H (2006) cular System Using Molecular Magnetic Reso-
Exploiting the enhanced permeability and nance Imaging. Theranostics 5
retention effect for tumor targeting. Drug Dis- (11):1187–1202. https://doi.org/10.7150/
cov Today 11(17–18):812–818. https://doi. thno.11835
org/10.1016/j.drudis.2006.07.005 15. Jefferson A, Ruparelia N, Choudhury RP
8. McAteer MA, Sibson NR, von Zur Muhlen C, (2013) Exogenous microparticles of iron
Schneider JE, Lowe AS, Warrick N, Channon oxide bind to activated endothelial cells but,
KM, Anthony DC, Choudhury RP (2007) In unlike monocytes, do not trigger an endothe-
vivo magnetic resonance imaging of acute brain lial response. Theranostics 3(6):428–436.
inflammation using microparticles of iron https://doi.org/10.7150/thno.5895
oxide. Nat Med 13(10):1253–1258. https:// 16. Gauberti M, Montagne A, Marcos-Contreras
doi.org/10.1038/nm1631 OA, Le Behot A, Maubert E, Vivien D (2013)
9. Shapiro EM, Skrtic S, Koretsky AP (2005) Ultra-sensitive molecular MRI of vascular cell
Sizing it up: cellular MRI using micron-sized adhesion molecule-1 reveals a dynamic
Molecular MRI 327
inflammatory penumbra after strokes. Stroke JJ, Rothwell NJ, Rubio M, Waeber C, Young
44(7):1988–1996. https://doi.org/10.1161/ AR, Touze E, Vivien D (2016) Efficacy of
strokeaha.111.000544 Alteplase in a Mouse Model of Acute Ischemic
17. Marcos-Contreras OA, Martinez de Stroke: A Retrospective Pooled Analysis.
Lizarrondo S, Bardou I, Orset C, Pruvost M, Stroke 47(5):1312–1318. https://doi.org/
Anfray A, Frigout Y, Hommet Y, Lebouvier L, 10.1161/strokeaha.116.012238
Montaner J, Vivien D, Gauberti M (2016) 21. Gaberel T, Gakuba C, Hebert M, Montagne A,
Hyperfibrinolysis increases blood-brain barrier Agin V, Rubio M, Emery E, Vivien D, Gauberti
permeability by a plasmin- and bradykinin- M (2013) Intracerebral hematomas disappear
dependent mechanism. Blood 128 on T2*-weighted images during normobaric
(20):2423–2434. https://doi.org/10.1182/ oxygen therapy. Stroke 44(12):3482–3489.
blood-2016-03-705384 https://doi.org/10.1161/strokeaha.113.
18. Quenault A, Martinez de Lizarrondo S, 002045
Etard O, Gauberti M, Orset C, Haelewyn B, 22. Gaberel T, Gakuba C, Goulay R, Martinez De
Segal HC, Rothwell PM, Vivien D, Touze E, Lizarrondo S, Hanouz JL, Emery E, Touze E,
Ali C (2017) Molecular magnetic resonance Vivien D, Gauberti M (2014) Impaired glym-
imaging discloses endothelial activation after phatic perfusion after strokes revealed by
transient ischaemic attack. Brain 140 contrast-enhanced MRI: a new target for fibri-
(Pt 1):146–157. https://doi.org/10.1093/ nolysis? Stroke 45(10):3092–3096. https://
brain/aww260 doi.org/10.1161/strokeaha.114.006617
19. Gakuba C, Gauberti M, Mazighi M, Defer G, 23. Gauberti M, Martinez de Lizarrondo S,
Hanouz JL, Vivien D (2011) Preclinical evi- Orset C, Vivien D (2014) Lack of secondary
dence toward the use of ketamine for recombi- microthrombosis after thrombin-induced
nant tissue-type plasminogen activator- stroke in mice and non-human primates. J
mediated thrombolysis under anesthesia or Thromb Haemost 12(3):409–414. https://
sedation. Stroke 42(10):2947–2949. https:// doi.org/10.1111/jth.12487
doi.org/10.1161/strokeaha.111.620468 24. Briens A, Gauberti M, Parcq J, Montaner J,
20. Orset C, Haelewyn B, Allan SM, Ansar S, Vivien D, Martinez de Lizarrondo S (2016)
Campos F, Cho TH, Durand A, El Amki M, Nano-zymography Using Laser-Scanning
Fatar M, Garcia-Yebenes I, Gauberti M, Confocal Microscopy Unmasks Proteolytic
Grudzenski S, Lizasoain I, Lo E, Macrez R, Activity of Cell-Derived Microparticles. Ther-
Margaill I, Maysami S, Meairs S, anostics 6(5):610–626. https://doi.org/10.
Nighoghossian N, Orbe J, Paramo JA, Parienti 7150/thno.13757
Part V
MRI and MRS in Animal Models of Disease

Chapter 20
Magnetic Resonance Spectroscopy Studies of Mouse

Models of Cancer
Menglin Cheng and Kristine Glunde
Abstract
Magnetic resonance spectroscopy (MRS) or spectroscopic imaging (MRSI) enables the detection of
metabolites, amino acids, and lipids, among other biomolecules, in tumors of live mouse models of cancer.
Tumor-bearing mice are anesthetized by breathing isoflurane in a magnetic resonance (MR) scanner
dedicated to small animal MR. Here we describe the overall setup and steps for measuring 1H and 31P
MRS and 1H MRSI of orthotopic breast tumor models in mice with surface coils. This protocol can be
adapted to the use of volume coils to measure 1H and 31P MRS(I) of tumor models that grow inside the
body. We address issues of animal handling, setting up the measurement, measurement options, and data
analysis.
Key words Cancer, Magnetic resonance spectroscopic imaging, Animal setup, Shimming, Chemical
shift imaging, Metabolite, Amino acid, Lipid
1 Introduction
A commonly found hallmark of cancer are distinct metabolic altera-

tions in glucose (Glc) and lactate (Lac) [1, 2], glutamine (Gln) [3],
and choline phospholipid [4, 5] metabolism in cancer cells [6]. Typ-
ically, cancer cells utilize significantly less oxygen than their respec-
tive normal cell counterparts. This radically alters their energy
production and modifies their mitochondrial function from mainly
energy production to the creation of biosynthetic intermediates
that support cancer cell growth [7]. Consequently, cancer cells
have an increased consumption of Glc [1, 2], Gln [3, 8], and
choline (Cho) [5], and an increased production of Lac [2]. More-
over, phosphocholine (PC) [5] and glycerophosphocholine (GPC)
[9] levels are altered owing to upregulated choline transporters,
choline kinase α, and phosphatidylcholine-specific phospholipases
D and C [5]. Proton magnetic resonance spectroscopy (MRS) and
magnetic resonance spectroscopic imaging (MRSI) are able to
detect these metabolites in vivo and are starting to be used in the
331
332 Menglin Cheng and Kristine Glunde
clinic in addition to conventional magnetic resonance imaging

(MRI) for cancer diagnosis and treatment monitoring
[10, 11]. Typically, in vivo 1H MRS detects sum signals of a few
overlapping metabolites, including total choline (tCho) (PC plus
free choline (Cho) plus GPC), total creatine (tCr) (creatine (Cr)
plus phosphocreatine (PCr)), separate CH2 and CH3 peaks from
mixed fatty acid signals in lipids (Lip), as well as individual meta-
bolites, such as N-acetylaspartate (NAA) in the brain and Lac
[12]. As there is a lot of interest in monitoring metabolites for
cancer diagnosis, treatment planning [11, 13], and treatment
response [14] in vivo, 1H MRS and MRSI are constantly being
improved to increase their sensitivity in vivo, to detect endogenous
metabolites in the absence of contrast agents or tracers with
improved spatial resolution.
Another possibility of detecting the elevated concentrations of
water-soluble phospholipid metabolites, such as the phosphomo-
noesters (PMEs) PC and phosphoethanolamine (PE) and the phos-
phodiesters (PDEs) GPC and glycerophosphoethanolamine (GPE)
in vivo is the use of 31P MRS [15, 16]. This is particularly attractive
because the total choline (tCho) signal detected by 1H MRS can in
many cases not be spectrally resolved into its individual metabolites
free choline (Cho), PC and GPC, due to the low spectral resolution
in vivo [11, 17, 18]. With 31P MRS, it is possible to detect individ-
ual PME and PDE signals, and, at higher field, often times even
partially resolve these signals into PC and PE, as well as GPC and
GPE [18]. Several preclinical and clinical MR groups are working
toward accurate measurement and quantification of these phospho-
lipid metabolites with 31P MRS in vivo, which is particularly impor-
tant in the clinic as consistent changes in phospholipid metabolite
levels can aid in cancer diagnosis, prognosis, and treatment
response monitoring [19].
While clinical oncological 1H and 31P MRS is being further
developed in some of the advanced clinical centers, there is also a lot
of work going on in animal models of cancer, in particular as novel
high-field applications are being tested in animals first. Small animal
MRS(I) has its own difficulties of working with mice, which require
anesthesia for the duration of the MR scan, and which are obviously
much smaller than human subjects or patients. In this chapter, we
will provide an overview of how to anesthetize and set up tumor-
bearing mice in MR scanners, how to set up MRS(I) measurements,
how to use different measurement options, and initial guidance for
data analysis.
2 Materials
1. For xenograft models, cells can be obtained from the American

Type Culture Collection (ATCC). For breast tumor xenograft
models, highly aggressive human MDA-MB-231 breast cancer
MRS of Mouse Models of Cancer 333
2.1 Mice Growing cells and weakly aggressive human MCF-7 cells can be pur-
Orthotopic Tumor chased. Other cell lines and cancer types are possible and have
Xenografts been widely published.
2. Cell culture media containing fetal bovine serum and antibio-
tics. MDA-MB-231 cells: Maintain cells in RPMI 1640 (Invi-
trogen Corp.) supplemented with 10% fetal bovine serum,
100 units/ml penicillin, and 100 μg/ml streptomycin (Invitro-
gen Corp.). MCF-7 cells: Culture cells in EMEM medium
(Mediatech, Inc.) supplemented with 10% fetal bovine serum
and the same antibiotics as used for MDA-MB-231 cells.
3. All cells require a humidified atmosphere of 5% CO2 in air, at
37 C, which is achieved in a cell culture incubator.
4. A sterile hood is required for mammalian cell culture, which has
to occur under sterile conditions.
5. A mixture of ketamine (20–50 mg/kg) and acepromazine
(0.5–2.5 mg/kg) is injected intraperitoneally to anesthetize
mice for cell inoculation into the mammary fat pad.
6. Hank’s balanced salt solution (HBSS).
7. Matrigel™ (Corning Life Sciences or BD Biosciences).
8. 0.72 mg/pellet 17β-estradiol 60-day release pellets to achieve
growth of estrogen-sensitive xenografts in mice.
9. Female homozygous athymic nude mice or severe combined
immunodeficient (SCID) mice can be purchased from differ-
ent, country-specific vendors and are obtained at 4–6 weeks of
age (see Note 1).
10. A state-of-the-art animal facility is required to house and take
care of mice.
11. An animal protocol that is in compliance with the performance
site’s country and institutional regulations and policies needs
to be in place prior to performing animal experiments. In our
case, all surgical procedures and animal handling are approved
by the Johns Hopkins University Institutional Animal Care and
Use Committee, and conform to the National Institutes of
Health (NIH) Guide for the Care and Use of Laboratory
Animals.
12. Surgical instruments including surgical scissors, scalpels,
razors, trocar, and syringes and needles are required. All surgi-
cal instruments need to be sterilized in an autoclave prior to use
and should only be used once after sterilization. Cell inocula-
tions are performed with a 26 Gauge needle. Gloves are neces-
sary for animal handling. Dedicated space in an animal facility is
needed to perform tumor cell inoculations.
13. Cotton padding, laboratory tape, and thin cardboard are nec-
essary for the mouse setup on the animal holder.
2.2 Major Equipment 1. A small animal MR scanner of relatively high field strength is
required. This could be a Bruker Biospec small animal MRI
scanner (Bruker Co., Billerica, MA) operating at 4.7 T, 7.0 T,
9.4 T, or 11.7 T. Higher fields will provide better sensitivity,
but shimming could potentially be more difficult at higher field
strengths.
2. Anesthesia equipment is needed to initially put to sleep the
mouse with isoflurane.
3. An animal holder that carries the surface solenoid or the vol-
ume coil to be used for the MR measurement is required. It is
also advisable to attach a warming pad for the mouse to stay
warm (see Note 2). Such animal holders with adequate coils are
either commercially available from Bruker, or they can be
home-built. Coils can also be obtained from specialized com-
panies such as MRcoils BV, Drunen, The Netherlands. The
holder is typically constructed on a half-shell of a plastic cone
of cylindrical shape, which fits the bore of the MR scanner to be
used. This plastic holder carries the coil, all electronics, tuning
screw sticks, cables, mouse holder, warming pad, and a nose
cone for continuous supply of isoflurane gas anesthesia includ-
ing the gas connection tube to the anesthesia equipment.
4. It is helpful to use animal monitoring equipment to continu-
ously monitor the breathing rate of the mouse with a move-
ment sensor to assess the depth of anesthesia throughout the
experiment. Additionally, it is also good to monitor the body
temperature of the animal with an anal thermometer.
2.3 Data Analysis 1. A personal computer that is on the internet or intranet is

required for data analysis.
2. Software to handle MRS(I) data analysis is needed. Commer-
cial software supplied by the vendor of the Small Animal MR
scanner that is used can also be used for data analysis. However,
we recommend using the latest version of JMRUI software
[19] for the use of its flexible fitting algorithms. Other software
tools are available as well.
3 Methods
3.1 Generate Breast 1. For MCF-7 tumor xenografts, subcutaneously implant one
Cancer Xenograft 0.72 mg/pellet 17β-estradiol 60-day release pellet per mouse
Models in Mice in the scruff region 24 h prior to tumor cell inoculation.
2. Prepare sterile single cell solutions of breast cancer cells by
suspending 2 107 cells per mL in a 50:50 (v/v) solution of
Hank’s balanced salt solution and Matrigel under a
sterile hood.
3. Anesthetize mice with a mixture of ketamine (20–50 mg/kg)

and acepromazine (0.5–2.5 mg/kg) injected intraperitoneally.
4. Inject sterile solution of 2 106 MDA-MB-231 or MCF-7
cells per mouse in 50 μL of HBSS/Matrigel using a 26 Gauge
needle into the fourth right thoracic mammary fat pad of
anesthetized, 4–6 week old, female athymic nude or SCID
mice with a body weight of about 23–28 g (see Note 3).
5. Keep mice comfortably warm under a heating lamp or on a
heating pad until they have recovered from anesthesia. Regu-
larly monitor mice until they have recovered from anesthesia by
checking them after 6, 12, 24, and 48 h following inoculation.
6. Monitor and measure primary tumor volumes once to twice
per week. Once tumors are palpable, measure tumor volumes
once per week in the beginning, and in later stages of tumor
growth twice per week. Calculate tumor sizes from caliper
measurements using the formula for vol-
ume ¼ (4/3) π (radius)3 ¼ π/6 (a b c) where a,
b, and c are the three measured orthogonal diameters.
7. Depending on the cell line inoculated, tumor xenografts typi-
cally develop to full size and are ready for MRS measurement at
8–12 weeks following orthotopic inoculation, at which stage
primary tumor sizes reach 500–800 mm3 (see Note 4).
3.2 Mouse Setup 1. Prepare mouse holder close to magnet by attaching all cables,
on Animal Holder heating pad, and gas anesthesia line. Log into scanner software
in MR Magnet and open up measurement protocol to be used in software.
2. Measure weight and tumor size of the mouse to be MR
scanned. Tumor sizes of about 500 mm3 and larger will provide
good signal-to-noise ratio (SNR) for MRS measurements (see
Note 4).
3. Anesthetize the tumor-bearing mouse to be scanned by insert-
ing her into the isoflurane apparatus, which infuses into a
closed chamber inhalatable isoflurane anesthetic. Use 3.5–4%
of inhalatable isoflurane anesthetic in this chamber for about
2 min until the mouse is deeply asleep.
4. Quickly transfer the mouse to the animal holder to which all
electronics, circulating water bath lines for heating pad, and gas
line for inhalatable isoflurane anesthetic are already attached.
Position the mouse with the tumor hanging into the surface
coil of about 12 mm in diameter underneath the mouse (see
Note 5). Adjust the nose cone such that the mouse’s nose and
face are well covered with the nose cone mask for breathing
isoflurane. Use laboratory tape to hold the mouse in its posi-
tion. Be careful to not squeeze the mouse too hard. Attach the
movement sensor of the breathing monitor at the mouse’s
abdomen and gently fix it with tape. Apply specialized eye
ointment to the eyes of the mouse so that they do not dry out
during the relatively long MRS scanning time. During the MRI
and MRS measurements, which will last for up to 2 h per
mouse, wrap mice in cotton/plastic paper for warmth, and
place a pad circulated with warm water on top of the mouse
to maintain normal animal body temperature.
5. After transfer to the animal holder, continuously apply an
average of 1.5% of inhalatable isoflurane anesthetic by nose
cone mask for maintenance of anesthesia while performing
the MRI and MRS(I) scans on a small animal scanner. Contin-
uously monitor breathing throughout the MRI and MRS
(I) measurement with the attached movement sensor. Check
anesthetic depth via the toe pinch, and adjust the isoflurane
accordingly.
6. Carefully and slowly, with all lines (electronics, water bath,
anesthesia) attached, push the animal holder with the attached
mouse into the magnet, just far enough to be at the correct
depth, which should have been pre-marked from previous
measurements with the same animal holder and coil.
3.3 Noninvasive In 1. Set up the overall MR Scan for this mouse in the Paravision
Vivo 1H MRSI Studies Software (Bruker Co., Billerica, MA) that controls the MR
Scanner using the commands in Paravision Software. Initially,
adjust the spectrometer frequency (auto SF), set the transmit-
ter attenuator for optimal flip angle (auto RF gain), and set the
optimal receiver gain (auto RG), then wobble by adjusting the
wobble signal to be as far at the bottom and in the center as
possible by using the two wobble screws and going back and
forth between them as often as needed for optimal results (see
Note 6).
2. Start the positioning scan and optimally place the animal holder
in the center of the magnet for optimal signal strength by
pushing it further inside the magnet or pulling it out millimeter
by millimeter observing the signal strength on the monitor
while scanning with the GSP command. GSP reads out the
MR signal from one acquisition without adding up signal from
multiple repeats of the pulse program used. Go step-wise from
a large field of view of initially 50 cm for initial positioning,
then to 10 cm, and eventually to 3.2 cm for fine-tuned optimi-
zation of the position of the mouse inside the magnet.
3. Use a three-dimensional (3D) tripilot (also called triplanar)
scan for setting up the geometry of your MRI/MRSI scans.
First optimize the Tx0 and Tx1 pulses with 6 db units in
between them, optimize the gain, and acquire the tripilot
(triplanar) scan with the GOP command, which starts the
actual acquisition.
4. Acquire high-resolution 3D T1-weighted MRI scan, which is

needed to serve as high-resolution anatomic reference image
datasets for the 3D MRSI data. To this end, it is possible to use
a 3D rapid acquisition with relaxation enhancement (RARE)
spin echo fast imaging sequence to acquire 3D T1-weighted
images. RARE acquires multiple spin echoes using the Carr-
Purcell-Meiboom-Gill sequence with slice-selective radiofre-
quency pulses. 3D RARE can be performed with the following
parameters: echo time (TE) ¼ 7.2 ms, repetition time
(TR) ¼ 500 ms, RARE factor of 4, flip angle of 90 degree,
field of view (FOV) of 1.0 cm by 1.0 cm by 1.0 cm, 64 phase
encode steps (64 by 64 by 64 voxels), number of averages of
4. The total acquisition time with these parameters is 13 min.
The reconstruction of MRI reference images can be performed
using ParaVision software.
5. Setup and shim of 1H MRSI with water suppression: Load
3D CSI (three-dimensional chemical shift imaging) scan and
import geometry from 3D RARE scan. For shimming, set CSI
to 2D, no water suppression, echo time 15 ms, echo position
1%. Auto SF and then optimize pulses Tx0 and Tx1 to achieve
maximal signal. Go to “Edit GS” and set readout to “calculate
area of raw data.” During GSP-scanning, use the shim control
tool for manual shimming of first and second order B0 gradient
fields by iteratively assessing the water line width and area of the
raw data obtained from the entire tumor. During this process,
in other words, optimize the water signal line width and area
for shims x, y, z, x, y, z, z2, etc. in cycles with the goal of
minimizing the water signal line width and maximizing the
area of the raw signal. Next, once shimming is completed, set
number of averages to 1 (NA ¼ 1), water suppression to
VAPOR, TE to 82 ms, and echo position to 50%, TR to
1000 ms, sweep width to 4000 Hz, and spectral resolution to
3.9 Hz/point. Now change values of Tx4 and Tx5, with the
same power, to suppress water while observing the water signal
using GSP scanning. Optimize Tx4 and Tx5 such that the
water signal gets minimized at the lowest possible Tx4/Tx5
power (see Note 7). Now, switch back to 3D CSI, adjust the
field of view (FOV) to 1-cm by 1-cm by 1-cm, enclosing the
tumor with the same geometry as used for the corresponding
3D RARE acquisition, with an in-plane resolution of 1.25-mm
by 1.25-mm by 1.25-mm. Use standard rectangular k-space
sampling with a matrix size of 8 by 8 by 8 (zero filled to 64 by
64 by 64), optimize the gain to about 120% (see Note 8), set
the number of averages (NA) to 4, and start the measurement
with GOP. This water-suppressed 3D CSI measurement will
result in a total measurement time of 33 min. Make sure body
temperature and breathing rate of the mouse are stable while

the measurement is running.
6. Setup and shim of 1H MRSI without water suppression. This
scan is used as quantitative reference scan to relate metabolite
signals from the 3D CSI acquisition with water suppression to
the water signal in 3D CSI without water suppression, the
latter of which reflects cell density [20–22]. Copy all para-
meters from 3D CSI with water suppression into a new scan
(clone scan), switch off water suppression, and set echo time to
15 ms and echo position to 1%. Optimize gain to 120% (see
Note 7) and set number of averages to 2. Start measurement
with GOP command. This water-unsuppressed 3D CSI mea-
surement will result in a total measurement time of 17 min.
7. Quantitative analysis of in vivo 1H MRS(I) data can be achieved
with commercial, freely available, or home-built software. In
our case, an in-house IDL program was used to reconstruct
water-unsuppressed and water-suppressed 1H MR spectro-
scopic images. This IDL-based software Fourier transforms
over both spatial and spectral axes of MRSI raw data and
automatically detects all metabolite peaks by applying a sliding
window across the whole spectral range with a window width of
0.3 ppm and a step size of 0.3 ppm [21, 22]. It applies a SNR
threshold to the water-suppressed 3D MRSI data to detect
metabolite peaks. The SNR threshold was set to three in our
studies [21, 22]. The concentration maps of all detected
metabolite signals from the water-suppressed MRSI data
(TE ¼ 82 ms) as well as the lipid CH3 signal at 0.9 ppm from
the water-unsuppressed MRSI data (TE ¼ 15 ms) are quanti-
fied by normalizing their signals to the water signal in the
corresponding water-unsuppressed MRSI data [20]. Figure 1
shows examples of such 3D tCho and CH3 concentration maps
(see Note 9).
3.4 Noninvasive In 1. In vivo 31P MRS requires a double tuned coil that is tuned for
Vivo 31P MRS Studies the 1H and 31P frequencies, which was in our case a solenoid
coil with an inner diameter of 12 mm (MRcoils BV, Drunen,
The Netherlands) [15, 16]. Set up the mouse as described in
Subheading 3.2.
2. Set up the overall MR Scan for a new mouse in the Paravision
Software at the MR Scanner and perform steps 1–5 in the same
way as described under Subheading 3.3 for 1H MSRI measure-
ments including shimming.
3. Acquire non-localized pulse-acquire 31P MRS with an adiabatic
excitation (BIR4 45 , 200 μs, 120 ppm band width), TR of 1 s,
and 2000 averages. Use a saturation slab, which is an adiabatic
full passage pulse driven at half the amplitude to achieve
A B mmol/kg C mmol/kg
6.0 0.8
5.25 0.7
4.5 0.5
3.75 0.5
3.0 0.4
2.25 0.3
1.5 1.2
0.75 0.1
0.0 0.0
D E
0 6 mmol/kg 0 1 mmol/kg
Fig. 1 (a) Example of a water-suppressed 1H CSI MRSI spectrum from a representative MDA-MB-231-HRE-
tdTomato tumor (approximately 360 mm3) obtained with a spatial resolution of 0.9 0.9 0.8 mm showing
tCho at 3.2 ppm and lipid CH3 at 0.9 ppm. (b) Example of a two-dimensional tCho concentration map obtained
from the signal at 3.2 ppm in the water-suppressed MRSI data (TE ¼ 82 ms). (c) Example of a
two-dimensional lipid CH3 concentration map obtained from the signal at 0.9 ppm in the water-unsuppressed
MRSI data (TE ¼ 15 ms). Example of (d) a three-dimensional tCho concentration volume (green) and (e) a
three-dimensional lipid CH3 concentration volume (cyan) with the tumor boundary shown as a yellow grid.
Adapted from [21]
excitation with fully dispersed phase [23], covering the mouse

body to eliminate signals from muscles in the body
[15, 16]. The combination of the drop off of 31P radiofre-
quency (RF) field strength perpendicular to the solenoid coil
in which the tumor is placed, and the saturation slab positioned
on the mouse body ensures that only signal from tumor tissue
is acquired [15, 16]. The combination of a short repetition
time and 45 flip angle approximates pulse-acquire acquisition
with Ernst-angle excitation for PE and PC [15, 16].
4. Quantitative analysis of in vivo 31P MRS data can be achieved
with Lorentzian line fitting using JMRUI software [19] and
the AMARES algorithm [24] as described recently
[15, 16]. Figure 2 shows a typical fitting result of in vivo
pulse-acquire 31P MRS (top) and specialized BINEPT 31P
Fig. 2 Example of in vivo pulse-acquire (PA, top) and BINEPT (bottom) 31P MR spectra of a representative
MCF-7 (left) and MDA-MB-231 (right) orthotopic human breast tumor xenografts. Lorentzian lines as fitted by
JMRUI are shown below each MR spectrum. All phosphorylated metabolites are visible in the PA spectrum,
whereas the BINEPT spectrum only contains signals from phospholipid metabolites with H-P-coupling such as
PE, PC, GPE, and GPC. Note the broad, uneven baseline in the 0–5 ppm region of the PA spectra, where signals
from mobile membrane phospholipids are resonating. Adapted from [16]
MRS (bottom) from orthotopic MCF-7 (left) and MDA-MB-

231 (right) breast tumor xenografts [16]. For quantitative
analysis using such fitting, set the resonance of phosphocrea-
tine to 0 ppm. In the fitting, constrain the line widths of
phosphomonoesters and phosphodiesters to the line width of
PCr, and fix the frequency difference between PC and phos-
phoethanolamine and between GPC and GPE to 100 Hz. One
commonly performed option of quantifying metabolite levels is
as ratios with respect to β-nucleotide triphosphate (NTP). If
this is done, it is necessary to correct metabolite levels for
differences in T1 relaxation, using metabolite T1 values that
were measured in vivo by progressive saturation series in
tumors that are similar to the tumors being analyzed
[15, 16]. To compare MR spectra of different mice with differ-
ent coil loads and gain settings, metabolite levels can alterna-
tively be quantified as ratios to the noise, which is measured
from the standard deviation of the last 200 points in the time
domain signal. These metabolites should also be corrected for
differences in T1 relaxation (see Note 9).
4 Notes
1. Choice of mouse model: Appropriate mouse and tumor models

should be chosen depending on the research question. In
general, for MRS studies, larger tumors will give better signal.
If a xenograft model is chosen, the use of orthotopic models,
even though they might be harder to inoculate and image, is
preferred because they more closely resemble human disease as
compared to subcutaneous inoculation in the flank [25]. The
choice of an appropriate host mouse strain, i.e., athymic nude
mice or SCID mice or variations thereof, will depend on the
xenograft model of choice, as not all immune-modulated
mouse hosts grow a given xenograft equally well [15, 16, 21,
22, 26–30]. Another variation of xenograft models are patient-
derived xenograft (PDX) models, in which tumor tissue is
directly transferred into mice after surgical removal from
human patients [31, 32]. PDX models recapitulate the
biological diversity of a given cancer more accurately than
classical xenograft models that are generated from inoculated
cell lines [31, 32]. PDX models are particularly helpful for drug
response studies [31, 32]. Genetically engineered mouse mod-
els (GEMM) of cancer, such as for example for different types
of breast cancer [33, 34], can be purchased from the Jackson
Laboratory (Maine, US). For breast cancer, it is possible to use
STAT1/ mice as model for ERþ breast cancer, MMTV-
Neu mice as model for ErbB2/HER2þ breast cancer, and C3
(1)SV40 T-antigen mice as model for triple negative breast
cancer, altogether representing around 75% of all human breast
cancers [33, 34]. The advantage of using GEMM is that they
are highly molecularly heterogeneous and therefore more real-
istically represent human disease [33, 34]. In terms of practical
considerations, some GEMM develop tumors very slowly and
are therefore difficult to use in MRS studies, which require
relatively large tumor sizes of around 500 mm3 to obtain
robust in vivo MRS data.
2. Keeping a mouse warm during a relatively long MRS scan is
important. Mice that cool out too much do not recover well
after an MRS scan and might get sick. This is particularly
important when using nude mice. Different warming pad solu-
tions are possible. In our case, we are using a recirculating water
bath whose lines are attached to a warming pad through which
the water is flowing slowly. The water bath is set to 42 C,
which results in a warming pad temperature of 37 C as the
water is cooling down as it is running through the lines into the
warming pad. In our measurements, we are placing the heating
pad on top of the mouse, loosely wrapping it around her top
parts, as on her bottom, there are the coil and the electronics.
The warming pad is also attached to the animal holder with

laboratory tape. Make sure that all the water lines are tightly
secured so that no water is spilling into the magnet.
3. The use of Matrigel for inoculating tumor cells in mice is
optional and depends on the tumors cells and inoculation site
used [28, 29]. The advantage of using Matrigel is that a more
even spread of cancer cells in the inoculation site is typically
achieved. However, many protocols available for the genera-
tion of tumor xenograft models do not require the use of
Matrigel. Many xenograft models also require the use of pieces
of tissue from tumors grown in mice or human patients, i.e.,
PDX models, rather than single cell suspensions. In terms of
injection site, it is preferable to inoculate in the orthotopic site,
i.e., into the mammary fat pad for breast cancer. Deep-seated
tissues, such as for example the prostate, may require survival
surgery for the generation of orthotopic xenograft models
[25]. For MRS studies, it is also helpful to scan a tumor that
grows in an area that is not prone to motion. In terms of
orthotopic breast tumor models, this means that xenograft
inoculation in the fourth rather than one of the upper mam-
mary fat pads helps with being further away from body sites
that move more vigorously during breathing.
4. Tumor size considerations: Large tumor sizes of 500 mm3
are generally better for achieving a good SNR in MRS studies
than smaller tumors. However, depending on the scientific
question under study, it may be necessary to start scanning
the tumors while they are still small. Also, depending on the
tumor model used, it may be counterproductive to grow the
tumors too large if they develop necrosis, which may then
reduce or alter their MRS signal [16]. The size of a tumor to
be MRS scanned should also fill the coil to be used as much as
possible to achieve a high filling factor. When placing a tumor
in the coil, the geometry and position of the tumor relative to
the mouse should be recorded for later registration with ex vivo
histology or other ex vivo imaging techniques to be performed
later-on from the same tumor [21].
5. A mouse becomes quite limp and difficult to stabilize once she
is anesthetized. Therefore, it might be good to prepare a card-
board holder around the mouse that has a cutout for the tumor
to stabilize the body of the mouse above the coil into which the
tumor is hanging. Simply use an almost paper-thin, easily fold-
able, rectangular-shaped piece of cardboard, cut into it a circu-
lar hole of about 0.5–1.0 cm diameter where the tumor of the
mouse is located, and wrap the mouse in it prior to putting her
onto the animal holder. The filling factor of the mouse tumor
in the coil should be as large as possible to avoid artifacts,
problems with shimming, and signal loss.
6. When setting up an MRS scan, which includes wobbling and

perfectly positioning the animal holder inside the MR scanner,
it is necessary for the person performing these tasks at the
scanner to be able to observe the monitor that shows the MR
scanning results in the adjacent computer room. In some cases,
depending on the setup of the MR scanner and adjacent com-
puter room, it might be helpful to place a strategically placed
mirror to make it possible to see the monitor of the computer
running the scanner while being at the scanner.
7. When adjusting the pulses Tx4 and Tx5 for VAPOR water
suppression, it is best to start with the weakest possible pulses,
incrementally increase them, and carefully observe the water
signal after every increase of Tx4 and Tx5. With increasing
water suppression pulse strength, the water signal will at some
point dramatically decrease following an incremental step size
increase in pulses, which likely are the desired pulses to be used.
Going beyond this “flip” in water signal decrease in terms of
suppression pulse strength would likely result in too strong of a
water suppression, which would in turn affect metabolites close
to the water signal, whose quantification could be rendered
inaccurate by a too strong water suppression.
8. The rationale for using a gain of 120% is that the shimming was
done on the total tumor signal, while the gain for the actual
measurement is for the signal per voxel.
9. Quality assessment of MRS data: To assess the quality of the
obtained in vivo MR data, it is advisable to measure the line
width of a signal of interest in the spectra from each tumor
(such as for example PCr in the in vivo 31P MRS data). It is also
possible to assess the necrotic fraction of each tumor by using
the dark areas in the corresponding 3D RARE T1-weighted
images of each tumor [16, 35, 36]. The tumor–air boundary
should be segmented in 3D by using a threshold that is set to
the highest 10% in the histogram. Segment necrotic dark spots
inside the tumor by manually adjusting a threshold to approxi-
mately the lowest 10% in the histogram. Quantify necrotic
regions as necrotic fraction per tumor by counting the number
of necrotic voxels, which are the dark spots inside the tumor
divided by the total number of voxels inside the tumor.
Overly broad line widths of the designated signal of choice in
the MR spectra may be used as a criterion to exclude tumors
with poor spectral resolution and large necrotic regions from
further analysis as fitting of MR spectra becomes unreliable in
cases where the line width is broad. The cutoff value in line
width to be chosen should be a tradeoff between including as
much data as possible and excluding data with poor quality
[15, 16, 21, 22].
Acknowledgements
This work was supported by NIH R01 CA134695, R01

CA154725, and P50 CA103175.
References
1. Gatenby RA, Gillies RJ (2004) Why do cancers Nelson MT, Yee D, Garwood M, Bolan PJ
have high aerobic glycolysis? Nat Rev Cancer 4 (2009) Metabolite quantification and high-
(11):891–899. https://doi.org/10.1038/ field MRS in breast cancer. NMR Biomed 22
nrc1478 (1):65–76. https://doi.org/10.1002/nbm.
2. Warburg O (1956) On the origin of cancer 1217
cells. Science 123(3191):309–314 12. Howe FA, Barton SJ, Cudlip SA, Stubbs M,
3. Kaelin WG Jr, Thompson CB (2010) Q&A: Saunders DE, Murphy M, Wilkins P, Opstad
Cancer: clues from cell metabolism. Nature KS, Doyle VL, McLean MA, Bell BA, Griffiths
465(7298):562–564. https://doi.org/10. JR (2003) Metabolic profiles of human brain
1038/465562a tumors using quantitative in vivo 1H magnetic
4. Negendank W (1992) Studies of human resonance spectroscopy. Magn Reson Med 49
tumors by MRS: a review. NMR Biomed 5 (2):223–232. https://doi.org/10.1002/mrm.
(5):303–324 10367
5. Glunde K, Bhujwalla ZM, Ronen SM (2011) 13. Serkova NJ, Brown MS (2012) Quantitative
Choline metabolism in malignant transforma- analysis in magnetic resonance spectroscopy:
tion. Nat Rev Cancer 11(12):835–848. from metabolic profiling to in vivo biomarkers.
https://doi.org/10.1038/nrc3162 Bioanalysis 4(3):321–341. https://doi.org/
10.4155/bio.11.320
6. Hanahan D, Weinberg RA (2011) Hallmarks
of cancer: the next generation. Cell 144 14. Nelson MT, Everson LI, Garwood M,
(5):646–674. https://doi.org/10.1016/j.cell. Emory T, Bolan PJ (2008) MR spectroscopy
2011.02.013 in the diagnosis and treatment of breast cancer.
Semin Breast Dis 11(2):100–105. https://doi.
7. DeBerardinis RJ, Lum JJ, Hatzivassiliou G, org/10.1053/j.sembd.2008.03.004
Thompson CB (2008) The biology of cancer:
metabolic reprogramming fuels cell growth 15. Wijnen JP, Jiang L, Greenwood TR, Cheng M,
and proliferation. Cell Metab 7(1):11–20. Dopkens M, Cao MD, Bhujwalla ZM,
https://doi.org/10.1016/j.cmet.2007.10. Krishnamachary B, Klomp DW, Glunde K
002 (2014) Silencing of the glycerophosphocholine
phosphodiesterase GDPD5 alters the phos-
8. Wise DR, Thompson CB (2010) Glutamine pholipid metabolite profile in a breast cancer
addiction: a new therapeutic target in cancer. model in vivo as monitored by (31) P MRS.
Trends Biochem Sci 35(8):427–433. https:// NMR Biomed 27(6):692–699. https://doi.
doi.org/10.1016/j.tibs.2010.05.003 org/10.1002/nbm.3106
9. Moestue SA, Giskeodegard GF, Cao MD, 16. Wijnen JP, Jiang L, Greenwood TR, van der
Bathen TF, Gribbestad IS (2012) Glyceropho- Kemp WJ, Klomp DW, Glunde K (2014)
sphocholine (GPC) is a poorly understood bio- 1H/31P polarization transfer at 9.4 Tesla
marker in breast cancer. Proc Natl Acad Sci U S for improved specificity of detecting phospho-
A 109(38):E2506; author reply E2507. monoesters and phosphodiesters in breast
https://doi.org/10.1073/pnas.1208226109 tumor models. PLoS One 9(7):e102256.
10. Danishad KK, Sharma U, Sah RG, Seenu V, https://doi.org/10.1371/journal.pone.
Parshad R, Jagannathan NR (2010) Assess- 0102256
ment of therapeutic response of locally 17. Meisamy S, Bolan PJ, Baker EH, Pollema MG,
advanced breast cancer (LABC) patients under- Le CT, Kelcz F, Lechner MC, Luikens BA,
going neoadjuvant chemotherapy (NACT) Carlson RA, Brandt KR, Amrami KK, Nelson
monitored using sequential magnetic reso- MT, Everson LI, Emory TH, Tuttle TM,
nance spectroscopic imaging (MRSI). NMR Yee D, Garwood M (2005) Adding in vivo
Biomed 23(3):233–241. https://doi.org/10. quantitative 1H MR spectroscopy to improve
1002/nbm.1436 diagnostic accuracy of breast MR imaging: pre-
11. Haddadin IS, McIntosh A, Meisamy S, liminary results of observer performance study
Corum C, Styczynski Snyder AL, Powell NJ, at 4.0 T. Radiology 236(2):465–475. https://
doi.org/10.1148/radiol.2362040836. mammary fat pad of mice to study tumor

236/2/465 [pii] growth. J Vis Exp 96. https://doi.org/10.
18. Klomp DW, van de Bank BL, Raaijmakers A, 3791/51967
Korteweg MA, Possanzini C, Boer VO, van de 28. Rizwan A, Bulte C, Kalaichelvan A, Cheng M,
Berg CA, van de Bosch MA, Luijten PR (2011) Krishnamachary B, Bhujwalla ZM, Jiang L,
31P MRSI and 1H MRS at 7 T: initial results in Glunde K (2015) Metastatic breast cancer
human breast cancer. NMR Biomed 24 cells in lymph nodes increase nodal collagen
(10):1337–1342. https://doi.org/10.1002/ density. Sci Rep 5:10002. https://doi.org/10.
nbm.1696 1038/srep10002
19. Naressi A, Couturier C, Devos JM, Janssen M, 29. Rizwan A, Cheng M, Bhujwalla ZM,
Mangeat C, de Beer R, Graveron-Demilly D Krishnamachary B, Jiang L, Glunde K (2015)
(2001) Java-based graphical user interface for Breast cancer cell adhesome and degradome
the MRUI quantitation package. Magma interact to drive metastasis. NPJ Breast Cancer
(New York, NY) 12(2–3):141–152 1:15017
20. Bolan PJ, Meisamy S, Baker EH, Lin J, Emory T, 30. Krishnamachary B, Glunde K, Wildes F,
Nelson M, Everson LI, Yee D, Garwood M Mori N, Takagi T, Raman V, Bhujwalla ZM
(2003) In vivo quantification of choline com- (2009) Noninvasive detection of lentiviral-
pounds in the breast with 1H MR spectroscopy. mediated choline kinase targeting in a human
Magn Reson Med 50(6):1134–1143. https:// breast cancer xenograft. Cancer Res 69
doi.org/10.1002/mrm.10654 (8):3464–3471. https://doi.org/10.1158/
21. Jiang L, Greenwood TR, Artemov D, 0008-5472.CAN-08-4120
Raman V, Winnard PT Jr, Heeren RM, Bhuj- 31. Tentler JJ, Tan AC, Weekes CD, Jimeno A,
walla ZM, Glunde K (2012) Localized hypoxia Leong S, Pitts TM, Arcaroli JJ, Messersmith
results in spatially heterogeneous metabolic WA, Eckhardt SG (2012) Patient-derived
signatures in breast tumor models. Neoplasia tumour xenografts as models for oncology
14(8):732–741 drug development. Nat Rev Clin Oncol 9
22. Jiang L, Greenwood TR, van Hove ER, (6):338–350. https://doi.org/10.1038/
Chughtai K, Raman V, Winnard PT Jr, Heeren nrclinonc.2012.61
RM, Artemov D, Glunde K (2013) Combined 32. Cho SY, Kang W, Han JY, Min S, Kang J,
MR, fluorescence and histology imaging strat- Lee A, Kwon JY, Lee C, Park H (2016) An
egy in a human breast tumor xenograft model. integrative approach to precision cancer medi-
NMR Biomed 26(3):285–298. https://doi. cine using patient-derived xenografts. Mol
org/10.1002/nbm.2846 Cells 39(2):77–86. 10.14348/molcells.2016.
23. Choi IY, Tkac I, Gruetter R (2000) Single- 2350
shot, three-dimensional “non-echo” localiza- 33. Borowsky AD (2011) Choosing a mouse
tion method for in vivo NMR spectroscopy. model: experimental biology in context--the
Magn Reson Med 44(3):387–394. https:// utility and limitations of mouse models of
doi.org/10.1002/1522-2594(200009) breast cancer. Cold Spring Harb Perspect Biol
44:3<387::AID-MRM8>3.0.CO;2-3. [pii] 3(9):a009670. https://doi.org/10.1101/
24. Vanhamme L, van den Boogaart A, van Huffel cshperspect.a009670
S (1997) Improved method for accurate and 34. Usary J, Zhao W, Darr D, Roberts PJ, Liu M,
efficient quantification of MRS data with use of Balletta L, Karginova O, Jordan J, Combest A,
prior knowledge. J Magn Reson 129(1):35–43 Bridges A, Prat A, Cheang MCU, Herschko-
25. Penet MF, Pathak AP, Raman V, Ballesteros P, witz JI, Rosen JM, Zamboni W, Sharpless NE,
Artemov D, Bhujwalla ZM (2009) Noninva- Perou CM (2013) Predicting drug responsive-
sive multiparametric imaging of metastasis- ness in human cancers using genetically engi-
permissive microenvironments in a human neered mice. Clin Cancer Res 19
prostate cancer xenograft. Cancer Res 69 (17):4889–4899. https://doi.org/10.1158/
(22):8822–8829. https://doi.org/10.1158/ 1078-0432.Ccr-13-0522
0008-5472.CAN-09-1782 35. Osbakken MD, Kreider JW, Taczanowsky P
26. Rashid OM, Takabe K (2015) Animal models (1986) Nuclear magnetic resonance imaging
for exploring the pharmacokinetics of breast characterization of a rat mammary tumor.
cancer therapies. Expert Opin Drug Metab Magn Reson Med 3(1):1–9
Toxicol 11(2):221–230. https://doi.org/10. 36. Dodd NJ, Moore JV, Poppitt DG, Wood B
1517/17425255.2015.983073 (1989) In vivo magnetic resonance imaging of
27. Kocaturk B, Versteeg HH (2015) Orthotopic the effects of photodynamic therapy. Br J Can-
injection of breast cancer cells into the cer 60(2):164–167
Chapter 21
MRI in the Study of Animal Models of Neurodegenerative

Diseases
Nyoman D. Kurniawan
Abstract
Magnetic Resonance Imaging (MRI) is an important tool to study various animal models of degenerative
diseases. This chapter describes routine protocols of T1-, T2-, and T2*-weighted and diffusion-weighted
MRI for rodent brain and spinal cord. These protocols can be used to measure atrophy, axonal and myelin
injury and changes in white matter connectivity.
Key words Magnetic Resonance Imaging, Diffusion-weighted imaging, Diffusion tensor imaging,
Animal model of neurodegenerative diseases, Brain, Spinal cord, In vivo MRI, Ex vivo MRI, High-field
MRI
1 Introduction
Magnetic Resonance Imaging (MRI) has been used extensively to

study various models of degenerative diseases in the central nervous
system (CNS) such as animal models of stroke [1, 2], traumatic
brain injury [3–5], spinal cord injury [6–10], epilepsy, motor neu-
ron [11, 12], multiple sclerosis [13–16], Alzheimer [17–19] and
Huntington [20–24] diseases. MRI provides the advantage to
perform noninvasive longitudinal study to monitor disease progres-
sion or recovery using the same animal, and the results from MRI
can be validated using immunohistochemistry and microscopy
[8, 25, 26].
MR neuroimaging of animal model requires higher spatial
resolution compared to human, as the size of their brain and spinal
cord structures are much smaller. For example, the size of an adult
mouse and rat brains are approximately 14 10 8 mm3 and
25 15 12 mm3, respectively, compared to human brain, which
is approximately 200 150 150mm3. The size of the mouse and
rat spinal cords is even smaller, which is about 2–4 mm in diameter.
347
348 Nyoman D. Kurniawan
MRI for such small sample size are often necessary to be performed
using high-field systems (>7 Tesla), as the high magnetic field
provides better signal sensitivity and these scanners are equipped
with strong gradients (>300 mT/m) to enable the acquisition of
high-resolution images [27].
1.1 In Vivo MRI In vivo mouse and rat brain/spinal cord MRI data are often acquired
with two-dimensional (2D) multi-slice sequences with in-plane reso-
lutions in the range of 0.08–0.15 mm and 0.5–1 mm slice thickness.
Recent cryoprobe hardware development for mouse and rat brain
MRI have tremendously improved the image quality and signal-to-
noise ratio (SNR), especially for functional MRI, and allowed the
acquisition of 3D datasets at a higher (0.15–0.20 mm) isotropic re-
solution [28, 29]. The improvement of conventional MRI sequences,
such as a modified 3D-gradient spin-echo (GRASE) sequence, have
enabled zoomed imaging and better reduction of physiological
motion artifacts using gating and navigator sequence [30].
In vivo MRI requires animal anesthesia, maintenance of animal
physiology, and monitoring of respiration and body temperature
[31–34]. The length of in vivo imaging is normally kept under
2–3 h using air–isoflurane gas mixture or ketamine/xylazine injec-
tion [33]. Longer acquisition time can result in deterioration of the
stability of animal physiology, especially increased difficulty to
maintain a stable breathing rate, which may already be compro-
mised in the disease model. For spinal cord imaging, respiratory
gating is especially required to minimize motion artifact. Respira-
tory gating, however, can significantly lengthen the acquisition
time, but its use is often critical for sequences which are very
sensitive to motion such as diffusion-weighted imaging (DWI)
[35]. If required, a MRI-compatible ventilator can be used to
synchronize MRI acquisition and controlled mechanical
respiration [36].
1.2 Ex Vivo MRI Ex vivo imaging using fixed samples allows the acquisition of much
higher resolution images compared to in vivo, as the length of scan
time is much less limited. Depending on sample size and the
scanner field-strength, ex vivo images may be obtained at 30-μm
isotropic resolution in 1–2 h acquisition, or as high as 10 μm 3D
isotropic resolution over a 16 h acquisition using a small 5 mm
solenoid coil [37, 38]. Ex vivo imaging also allows the enhance-
ment of the tissue white matter (WM) and gray matter (GM)
contrast post fixation using a low concentration of gadolinium
MRI contrast agent. MRI contrast agents improve image quality
by reducing tissue T1 relaxation time to enable the use of a short
recycling time (TR) and increased scan averages to increase signal-
to-noise ratio; and also to enhance the T1 and T2 WM/GM con-
trast [39, 40].
MRI in Neurodegenerative Diseases 349
Tissue fixation, however, may produce undesirable microscopic

change to the tissues. For example, protein crosslinking by alde-
hyde polymer fixation (e.g., 4% paraformaldehyde) can result in
some degrees of tissue shrinkage, hardening, reduction in the tissue
relaxation times, and changes in extracellular/intracellular micro-
scopic diffusion [41–43]. All of these will need to be considered
during the interpretation of quantitative MRI in comparison to
in vivo imaging.
1.3 The Scope of MRI Animal models of neurodegeneration can be imaged using a wide
Protocols array of MR modalities. This chapter limits the description into
and Analyses widely used protocols: high-resolution T1-weighted imaging to
measure volumetric changes; T1, T2, and T2* mapping to measure
changes in tissue relaxation, edema, and magnetic susceptibility
[44–48]; and diffusion-weighted imaging (DWI), which is impor-
tant to measure axonal injury/demyelination and changes in brain
white matter connectivity [49, 50]. Other MRI modalities such as
spectroscopy [51], quantitative susceptibility-weighted imaging
(qSWI) [15, 52], and magnetization transfer imaging (MTI)
[53, 54] can provide quantitative measurements of tissue metabo-
lites, iron, and myelin content, but they are beyond the scope of this
chapter.
This chapter also describes standard approaches to analyze MRI
data using voxel based analysis (VBA) [23, 55], or using manual
segmentation of regions-of-interest (ROIs) [56]. Whole brain VBA
is typically performed on a medium (n ¼ 10–20) and large data sets
(n > 30). VBA involves automatic registration of MRI data into
mouse/rat brain atlases in order to allow for voxel-wise statistical
analyses between groups. In comparison, ROI analyses involve
segmentations of anatomical regions by drawing on the MR images
slice by slice. Subsequent statistical analyses can be done using an
analysis of variance and t-test, with the appropriate corrections for
multiple comparisons. Manual segmentation can be a tedious pro-
cess and thus may be suitable only for small sample studies (n < 10)
with limited ROIs [57, 58]. ROI analyses may also be influenced by
operator bias, and thus the operators are often blinded to the group
assignments.
2 Materials
These protocols were written largely based on our experience using

the Bruker MR imaging platform at 16.4 T (Bruker Biospin, Karls-
ruhe, Germany). However, a wide range of other MRI scanners can
be used, including newly developed benchtop MRI scanners specif-
ically tailored for small animal imaging with the field strengths in
the range of 1–7 T (MR Solutions and Bruker Biospin).
2.1 In Vivo Mouse Bruker Micro 2.5 gradient coil, 20 mm volume head radiofre-
Brain Imaging quency (RF) coil, an animal bed and a probe base. This setup is
suitable for mouse imaging from approximately P28 (post-natal
2.1.1 MRI Equipment
28 day/weaning age) to adult, with maximum weight of approxi-
mately 35 g.
2.1.2 Auxiliary 1. MRI-compatible small animal monitoring system

Equipment (SA Instruments or Biopac). The respiration and rectal body
temperature monitoring are especially critical to monitor and
maintain the animal physiology (see Note 1).
2. An animal anesthesia machine with isoflurane module, an
induction chamber, and a scavenger unit (e.g., a Nederman
arm) to absorb excess isoflurane.
3. Animal warming pads on the preparation bench for maintain-
ing the animal temperature pre- and post-imaging (a generic
type from a pet shop is often sufficient).
4. To maintain animal body temperature, the gradient water cir-
culation temperature is increased to 36–37 C. Alternatively, an
animal bed with warm water circulation or heated air circula-
tion can be used. The animal body temperature should be
continuously monitored to avoid over/under heating.
5. 30G needles/1 mL syringes and a fine Tygon tube
(0.2–0.5 mm internal diameter) to create a catheter for con-
trast agent injection, if required.
2.2 Ex Vivo Mouse 1. A perfusion pump (e.g., Harvard Peristaltic Pump P-70) or a
Brain Imaging syringe pump (e.g., Harvard Pump 11) for cardiac perfusion
(see Notes 2–5).
2.2.1 Sample
Preparation 2. A small vacuum pump for degassing sample prior to MRI.
3. Chemicals: Lethabarb, pentobarbital or isoflurane (to induce
deep anesthesia prior to cardiac perfusion), phosphate buffer
saline (PBS), and heparin (optional, to avoid blood clotting).
4. MRI contrast agent, e.g., Gd-DTPA (diethylenetriaminepen-
taacetic acid) or Magnevist® (Bayer).
5. Perfluoro ether solution Fomblin Y06/06 (Solvay Solexis,
Italy) medium.
2.2.2 MRI Equipment 1. Bruker Micro 2.5 or Micro 5 gradients and microimaging
probes.
2. Imaging coils.
(a) 15 mm volume RF coil for juvenile and adult mouse brain
imaging.
(b) 10 mm quadrature birdcage volume RF coil for embry-
onic to neonate mouse brain imaging.
Fig. 1 Sample set up for ex vivo brain imaging. (a) The brain sample is held by two plastic scaffoldings placed
gently above and below the brain. The scaffolding should only touch the brain lightly so they do not distort the
brain shape. The fomblin solution should only be about 2–3 mm above the sample. Do not use excess fomblin
as it produces an upward force that may distort the tissue. A foam is used to keep the sample centered and the
tube is closed with a paraffin membrane. (b) T1/T2* gradient echo image acquired at 30 μm 3D isotropic
resolution at 16.4 T
(c) 5 mm solenoid RF coil for brain cortical or hippocampal

slab imaging. The slab size is limited to ~3 mm cube to fit
in the center of the coil.
3. 5, 10, or 15 mm tubes, cut short to fit in the respective coil
sizes. Plastic supports are used to keep the brain upright inside
the tube while immersed in the Fomblin solution (Fig. 1).
2.3 In Vivo Mouse 1. A single transmit/receive linear surface coil (1.5 3.0 cm) and
Spinal Cord Imaging micro 2.5 imaging gradient.
Alternatively, a separate whole body transmit only and a receive
2.3.1 MRI Equipment
only surface coil can be used. This setup is favorable as the
volume coil will provide more homogenous excitation of the
sample.
2. Animal respiratory monitoring and gating (e.g., SA Instrument
or Biopac) is essential.
2.4 Ex Vivo Mouse 1. Bruker Micro 2.5 gradient coil and microimaging probe base.
Spinal Cord Imaging (a) 10, 15, or 20 mm volume coils can be used for a single or
2.4.1 MRI Equipment tandem sample imaging.
(b) 5 mm solenoid coil for 3 mm spinal cord slab imaging.
2.5 Processing 1. MRI data processing is usually performed in a Linux (e.g.,

Platform Neurodebian, Ubuntu, Centos) or OSX platform.
2. Depending on the data size to be analyzed, the workstation

may have at least 1 TB of hard drive, 16 GB of RAM (64 GB or
above is recommended), a quad-core Intel Xeon processor and
1 GB of dedicated video RAM.
3. The processing software (see the list in Subheading 3.4) are
typically distributed with open-source licenses, so they can be
downloaded freely from the internet. They often need to be
compiled to run efficiently in the workstation. Some software is
distributed as Matlab scripts, so they will require a Matlab
to run.
3 Methods
3.1 In Vivo Mouse 1. Anesthesia induction. Mouse is placed in a small anesthesia

Brain Imaging induction chamber, and anesthesia is initiated with 2–4% iso-
flurane in oxygen mixture at the flow rate of 2.5 L/min. The
3.1.1 Preparation for In
anesthesia may be induced slowly using 2% isoflurane or quickly
Vivo Mouse Brain MRI
using 4% isoflurane. A quick anesthesia is often favorable to
reduce stress on the animal.
Once the animal breathing rate is down to approximately
60 breaths per minute (bpm), the isoflurane level may be
reduced to 3% and the oxygen flowrate to 1.5 L/min.
2. Animal restrain and maintenance. Animal can then be placed
into the animal bed or may be placed on a warming pad on the
bench for other preparation, if required. Throughout the
experiments, the animal respiratory should be maintained
around 40–60 bpm (see Note 6) by adjusting the isoflurane
level and the oxygen flowrate (~1–1.5% isoflurane at 1–1.5 L/
min). The body temperature is maintained at 36–37 C (see
Note 7).
3. Contrast agent administration (optional). This may be used to
determine leakage of blood–brain barrier (BBB). The animal
may be injected through the tail vein (see Note 8) using a
diluted concentration of clinical contrast agent. Magnevist®
stock solution (0.5 M) can be diluted to 1:10–1:20 v/v
(25–50 mM) using phosphate-buffer-saline (PBS). The maxi-
mum tail vein injected volume for 20–25 g mouse is
150–200 μL.
4. Animal positioning. The anatomical region of interest must be
placed in the center of the coil, the gradient and the magnet.
This is important to obtain maximum coil sensitivity and
reducing burden to the imaging gradients.
5. Scan adjustments. Once the animal is placed inside the scanner,
the scan adjustment routine can be started. This consists of tuning
and matching of the RF coil, standard shimming (x, y, z), basic
Fig. 2 Animal positioning for mouse brain imaging at 16.4 T. (a) Animal head is placed inside a 20 mm head
coil and restrained with a toothbar. Animal is lightly strapped with surgical tape, and a respiratory breathing
pad (blue tubing) is placed underneath the abdomen just below the diaphragm. Animal is maintained under
anesthesia using an oxygen/isoflurane mixture (clear tubing). (b, c, d) are coronal, sagittal, and axial MRI
slices
scan frequency adjustment, and RF reference power measure-

ment. Higher-order and/or B0-map shimming can be subse-
quently performed to improve the local field homogeneity.
6. Slice positioning. 2D MR images are ideally acquired with the
read/phase/slice (r, p, s) axes placed in parallel to the left-right,
dorso-ventral and rostro-caudal axes of the brain (Fig. 2, see
Notes 9 and 10).
3.1.2 Diffusion-Weighted One preferred method to obtain 2D DWI data at high-field is using
Imaging the Stejskal-Tanner pulse-field gradient spin-echo sequence (DWI
SE). This sequence produces only small distortion artifacts but its
acquisition is slow and it is susceptible to motion artifacts
[59]. DWI spin-echo is thus limited for obtaining data with a
small number of diffusion encoding directions (6–12 directions).
To obtain a high angular resolution diffusion imaging
(HARDI) dataset with 30–60 diffusion encoding directions, the
Stejskal-Tanner DWI sequence can be acquired using echo planar
imaging (EPI) read-out sequence (DWI EPI) to achieve fast acqui-
sition [60]. DWI EPI, however, is more susceptible to distortion
artifact than DWI spin echo, which can be improved using high-
order shimming (see Note 11). Additionally, the echo time
(TE) may be reduced using EPI segmentation, but this may result
in increased susceptibility to motion artifact due to imperfect align-
ment of the lines in the segmented k-space data. Typically, 2D DWI
EPI may be acquired without segmentation or with 2–4 EPI seg-
ments at ultra-high field. Respiratory gating can be used to reduce
motion artifacts, which may often be present at the caudal part of
the brain. The read axis may also be placed in the dorso-ventral
plane to reduce motion artefacts.
2D DWI SE Acquisition 1. The field of view (FOV) required to cover the brain axially is
Parameters (See Note 12) typically 1.5 1.5 cm, with a matrix size of 128 128 to
produce a voxel size of 117 117 μm. The slice thickness is

approximately 1 mm, with 16 contiguous coronal slices to
cover the brain [61].
2. The repetition time (TR) ¼ 1.5 s, echo time (TE) ¼ 20 ms, and
diffusion pulse gradient/mixing time (δ/Δ) ¼ 2/12 ms, a
single scan average (NEX ¼ 1), acquisition bandwidth ¼ 50 kHz
and dummy scans (DS) ¼ 4.
3. To acquire accurate apparent diffusion coefficient (ADC), a
multi-shell diffusion gradient strength can be applied in three
orthogonal directions using b-values of 0, 400, 800, and
1200 s/mm2 [61]. The total acquisition time is approximately
17 min without and 35 min with respiratory gating.
4. To obtain DTI parameters fractional anisotropy (FA), axial,
radial, and mean diffusivities (AD, RD, MD), this above proto-
col should be modified to acquire the data with at least a single-
shell diffusion encoding gradient and 6 orthogonal directions
(12 directions is recommended) using a b-value of
1500–2000 s/mm2 [50].
2D DWI EPI Acquisition Higher-order diffusion models, such as diffusion kurtosis imaging
Parameters (See Note 12) (DKI), constrained spherical deconvolution (CSD), and Neurite
Orientation Dispersion and Density Imaging (NODDI), requires
HARDI or multi-shell HARDI acquisitions (see Note 13). Such
dataset can only be acquired using DWI EPI sequences.
1. 2D DWI EPI may be acquired using 24 contiguous slices at
0.6 mm thickness with FOV ¼ 1.60 0.96 cm and matrix
size ¼ 128 64, to produce 125 150 μm2 in-plane resolu-
tion. A combination of partial Fourier transform (FT) and
zero-fill acceleration factors of 1.35 can be used to reduce TE
and speed up the scan time. If using a cryoprobe, a thinner slice
thickness can be used to obtain a pseudo 3D dataset at
(200 μm)3 isotropic resolution.
2. TR ¼ 6 s, TE ¼ 14 ms, and sampling bandwidth 300–500 kHz.
Diffusion gradients δ/Δ ¼ 2.4/6.4 ms, b-value of
3000 s/mm2, NEX ¼ 2 and DS ¼ 4.
With these parameters, a 64 diffusion, single-shell direc-
tion-encoding measurements can be acquired within 1 h with-
out respiratory triggering, or 2 h with respiratory triggering
[60]. An example of a FA map acquired using this protocol is
shown in Fig. 3.
3. To acquire a multi-shell HARDI dataset, the number of direc-
tions may be reduced to accommodate various shells and
reducing NEX to 1 to keep the acquisition time within 3 h.
The SNR should be tested and the image resolution will need
to be adjusted to ensure good data quality is maintained.
Fig. 3 FA map of an in vivo mouse brain acquired using DTI-EPI at 16.4 T. The brain white matter structures
are shown. This image is reproduced from [60]
3.1.3 T1-Weighted T1-weighted imaging is useful to obtain high-resolution images to

Imaging measure volumetric changes [62]. In vivo T1-weighted imaging can
be used in combination with a bolus contrast agent injection to
detect BBB damage, brain lesion [1, 63, 14] or to highlight tumor
area [64]. At 16.4 T, the T1 of mouse brain in vivo is approximately
2.5 s [60] and thus a short TR (0.5–1 s) can be used to obtain T1
weighted images (see Note 14).
2D T1-Weighted Acquisition 1. T1-weighted imaging can be acquired using spin-echo or fast

Parameters spin-echo sequences (turbo factor 2–4) with short TR
(~500 ms) and short TE (10–20 ms, or as short as possible,
see Note 15), NEX ¼ 1–2, with similar FOV/matrix para-
meters as described in Subheading 3.1.2.1. The acquisition
time is approximately 5–10 min.
2. Alternatively, T1-weighted imaging may also be acquired using
a gradient echo sequence with short TR (~200 ms), short TE
(~6 ms), and a large excitation flip angle pulse (>30 ). A large
bandwidth (~100 kHz) can be used to achieve a short TE
(see Note 15). NEX may be increased to 4 to improve SNR,

especially when acquiring high resolution images.
3. T1 relaxation map may be estimated from a T1 saturation spin
echo sequence, where the data are acquired at variable TRs, at
~250, 500, 750, 1000, 2500, 5000, and 10,000 ms (see Note
16). A minimum of five variable TRs are required to obtain a
good fit, in which the longest TR should be approximately 5
of T1 of the tissue (T1 is in the range of 2–3 s for the CNS). A
turbo or fast spin echo sequence (FSE) and Fourier accelera-
tion can be used to keep the acquisition time ~ 1 h. If using
FSE, the turbo factor should be kept small (2–4) to keep the
TE short.
4. T1 relaxation may also be measured using an inversion recovery
(IR) spin echo sequence (see Note 16), with the variable IR
time (TI) in the range of 100–5000 ms, with the TR is set
constant at 10,000 ms (approximately 5 T1).
5. For IR sequence, water suppression of the cerebrospinal fluid
(CSF) may be obtained with IR ~ 700 ms (CSF nulling) to
obtain better visualization of brain regions.
3.1.4 T2-Weighted T2-weighted imaging is often used to determine tissue edema, for
Imaging example, in stroke [65] and traumatic CNS injury [5], and ventric-
ular enlargement. At 16.4 T, the T2 of mouse brain in vivo is
approximately 25–30 ms [60], so that a TE of 40–60 ms can be
used to obtain T2-weighted images.
2D T2-Weighted Acquisition – T2-weighted imaging can be acquired using spin-echo with long
Parameters TR (>5 s) and long TE (>40 ms). This TR setting, however, will
result in a long acquisition time (>20 min). Shorter acquisition
time can be obtained using FSE with a turbo factor of 4–8,
however turbo factor >4 may result in blurring or motion
artifacts. Alternatively, a shorter TR (~2 s) may be used with
some T1-weighting effect in the image.
– Quantitative T2 MR images can be acquired using a multi-echo
spin-echo sequence with TR ¼ 2 s and the TE of 10, 20, 30, 40,
50, 60 ms, using similar FOV/matrix image parameters as
described above (see Note 16). The total acquisition time is
approximately 20 min.
3.1.5 T2*-Weighted T2*-weighted imaging is sensitive to the variation of microscopic

Imaging (local) magnetic field, which can be useful to detect iron deposi-
tion, hemorrhage [15], and white matter demyelination
[66]. Medium TE (~8–15 ms) can be used to obtain T2*-weighting
without significant distortion artifacts (see Note 17).
2D T2*-Weighted 1. Quantitative T2* map can be acquired using a multi-slice multi-

Acquisition Parameters gradient-echo (GE) sequence with flip angle of 30 ,
TR ¼ 500 ms and the TE of 5, 10, 15, 20, 25, and 30 ms,
NEX ¼ 2, using similar FOV/matrix image parameters as
described above (see Note 16). The total acquisition time is
approximately 5–10 min.
2. T2*-weighted images can be post-processed to calculate
susceptibility-weighted imaging (SWI) and multi-echo T2*-
weighted images can be used to calculate quantitative suscepti-
bility mapping (QSM) [67–69].
3.2 Ex Vivo Mouse Procedure

Brain Imaging 1. Prior to MRI, incubate PFA-fixed samples with an equivalent
of 20 sample volume (approximately 30 ml of solution for an
3.2.1 Sample adult mouse brain) in 1 mM Gd-DTPA (Magnevist, Bayer) in
Preparation PBS (0.2% v/v) over 4 days. The samples should not be washed
using PBS prior to MRI, as it will quickly remove the contrast
agent from the tissue. Alternatively, the skull may be kept intact
for MRI, and the samples are incubated for approximately
4 weeks in 1 mM Gd-DTPA/PBS to ensure the PFA has
been washed out and Gd-DTPA penetrates the tissues
(a number of small holes may be carefully drilled into the
skull at non-critical areas to facilitate buffer exchange). The
spinal cord in an intact spine requires only a short (~4 days)
of incubation due to its small sample size.
2. Ex vivo samples may need to be vacuumed in the PBS/Mag-
nevist incubation solution or in the Fomblin medium to
remove air bubble in the tissue (see Notes 18 and 19).
3.2.2 Ex Vivo Mouse 1. Adult mouse brain HARDI: 3D DWI SE sequence with the
Brain MRI parameters TE/TR ¼ 22.8/400 ms, 30 uniformly distributed
DW directions, b ¼ 5000 s/mm2, δ/Δ ¼ 2.5/12.5 ms, NEX ¼ 1.
3D DWI SE Acquisition
FOV ¼ 19.6 11.7 8.4 mm matrix size ¼ 19.6 11.7 8.4,
Parameters
resolution ¼ 100 μm isotropic. The acquisition time ~32 h
(without partial Fourier acceleration), or 15 h with 1.5 partial
Fourier encoding acceleration in the phase dimensions [37].
2. Embryonic and neonate mouse brain HARDI can be acquired
with similar parameters as above, but the FOV is reduced to fit
in the sample size, and the matrix size adjusted to obtain 80 μm
isotropic resolution.
3. Ultra-high resolution brain HARDI can be acquired using
similar parameters as above using a 3 mm brain section, with
the FOV and matrix size adjusted to obtain 48 μm isotropic
resolution, with NEX ¼ 2 and the acquisition time of
~24 h [37].
4. To improve the quality of fibertracking, the k-space Fourier

transform may be performed with the digital resolution
increased by 50% to reduce interpolation error in estimating
the fiber orientation distribution (FOD) for fibertracking
[70, 71].
3D T1/T2*-Weighted 1. Conventional T1/T2*-weighted anatomical images: 3D GE

Acquisition Parameters sequence with TR/TE/FA ¼ 50 ms/12 ms/30 , 82 kHz
spectral bandwidth, NEX ¼ 2–4, partial Fourier accelera-
tion ¼ 1.4. FOV is set the same as for HARDI, with the matrix
size adjusted to produce images at 30 μm isotropic resolution.
The acquisition time is 1.5–3 h. An example of this high-
resolution 3D GE image is shown in Fig. 1b.
2. For a 3 mm brain section sample, the resolution can be
increased to 10 μm isotropic resolution, with an acquisition
time of ~2.5 h with 1.3 factor of zero-fill Fourier encoding
accelerations in the phase-encoded dimensions [37].
3.3 Rodent Spinal Positioning

Cord Imaging 1. Mouse head is secured via a tooth bar and nose/head cone,
3.3.1 In Vivo DTI such that the animal dorsum is located parallel and as close as
possible to the surface coil.
2. The animal must be positioned such that the center of
the region of interest is in the center of the coil to allow a
sufficient coverage of the dorsal and caudal parts of the spinal
cord.
3. The slice for RF gain calculation must be corrected manually so
that the RF reference plane is positioned transversely (not
sagittally) along the length of the spinal cord, to ensure a
correct power reference calibration for the surface coil.
4. Axial images should be acquired in parallel to the section of the
spinal cord of interest (Fig. 4). Multiple slice packages may be
acquired along the curvature of the spinal cord to obtain per-
fect orthogonal axial slices (see Note 20).
2D DWI Acquisition 2D DTI spin-echo sequence with the acquisition parameters

Parameters TR/TE ¼ 2400/21 ms, 12 diffusion-encoding gradient direc-
tions, b ¼ 1500 s/mm2, δ/Δ ¼ 2/14 ms, 2 b ¼ 0 image, NEX ¼ 2,
partial Fourier acceleration factor ¼ 2. FOV ¼ 9 12 mm 1.0 mm
slice thickness, matrix ¼ 128 170 to produce 70 70 μm2
in-plane resolution [7, 12]. The phase encoding should be placed
in left-right and the read encoding in dorsal-ventral direction to
minimize respiratory motion artifacts. The total acquisition time
with respiratory gating 2.5–3 h.
Fig. 4 Animal positioning for mouse spinal cord imaging at 16.4 T using a surface coil. (a) Animal head is
placed inside a head restrain with a toothbar. The animal spine is placed in the center of the surface coil, with
the respiratory breathing pad (blue tubing) is placed on the abdomen and lightly secured with a surgical tape. A
foam may be placed on its feet to support its upward position in the vertical scanner. (b, c, d) are sagittal,
coronal, and axial MRI slices. A FOV saturation band (b) may be used to suppress signal from the abdomen
2D T1-, T2-, and T2*- 2D T1-, T2-, and T2*-weighted imaging can be acquired using
Weighted Acquisition similar parameters to that described in Subheadings 3.1.3–3.1.5
Parameters for in vivo brain imaging, but with the FOV and resolution adjusted
as above. The application of T2*-weighted imaging with long TE is
limited in the spinal cord, due to severe inhomogeneity artifacts
from the spine and the lung.
3.3.2 Ex Vivo DTI For tandem imaging, samples must be glued as close as possible
between them to improve shimming and reduce the scan time (see
Note 21).
3D Acquisition Parameters
1. 3D DTI spin-echo sequence with TR/TE ¼ 400/20.8 ms, δ/
Δ ¼ 2.5/10 ms, 12 non-collinear directions, two b0 images,
b ¼ 3000 s/mm2, NEX ¼ 1 and partial Fourier acceleration
factor ¼ 1.5. Images can be acquired at 60 μm isotropic reso-
lution with the acquisition time of 13–24 h depending on the
number of samples imaged [7, 8].
2. Ultra-high resolution HARDI and T1/T2* imaging can be
acquired using similar parameters as described in Subheading
3.2.2.
3.4 Data Processing Neurodegeneration processes can be correlated with specific

(See Note 22) changes in DTI parameters (see [49] for review). In WM, axonal
injury can be observed as a reduction in AD, demyelination as an
3.4.1 Characterization
increase in RD, and these pathologies generally result in a decrease
of CNS Tissue
of FA. Extensive WM and GM injuries can be observed as an
Microstructure Using DWI
increase in MD. In stroke, MD initially decreases in the acute
stage as the cell body swells due to hypoxic injury (cytotoxic
edema), as the proportion of restricted intracellular water diffusion
increased. At the later stage, cellular necrosis results in an increase
of water diffusion in the extracellular space which is detected as an
increase in MD [72, 73]. DTI parameter changes are more compli-

cated when the pathology is located in regions with significant
crossing fibers. For example, an increase in FA (instead of a decrease
of FA) can be observed if one population on the fibers at the
crossing is diminished [74].
New higher order diffusion models have been developed to
measure changes more accurately in CNS tissue microstructure,
and the development in this area is progressing rapidly. One exam-
ple is Neurite orientation dispersion and density imaging
(NODDI) [75]. NODDI separates the CNS compartments into
the CSF, intracellular (neurites) and extracellular space. Changes in
neurite orientation dispersion index (ODI) and neurite density
(intracellular volume fraction, ICVF) appear useful to measure
changes in the CNS more accurately especially in the GM. GM is
often difficult to analyze using DTI, as the FA values are low (~0.2).
For ex vivo datasets, an additional NODDI compartment of
restricted diffusion is required to model the effect of tissue fixation.
DWI can be used to measure the distribution of axon diameter
of the WM using AxCaliber [76] or ActiveAx [77]. AxCaliber has
been implemented to measure axon diameter distribution in the
spinal cord or the midline of the brain corpus callosum, but it
cannot be used to measure the whole brain WM structure. AxCa-
liber uses q-space imaging (DWI using both variable δ/Δ and b-
values) but using only diffusion gradient directions orthogonal to
the WM fiber. In comparison, ActiveAx has the potential to per-
form whole brain axon diameter distribution measurement, but it is
more demanding to use than AxCaliber as it involves multi-shell
HARDI acquisitions. These methods are promising to measure
changes in the axon size due to neurodegenerative injury, and
they are still being developed to improve their accuracy and imple-
mentation for in vivo settings [78, 79].
3.4.2 Data 1. Data conversion. MRI data are typically exported into DICOM
Pre-processing before they are converted into nifti (.nii) format using dcm2nii
program (MRIcron suite), or into .mih/dat format using
mrconvert (MRTrix).
2. Image transform correction. This step is often needed when
working with Bruker data with MRTrix (www.mrtrix.org).
The 3 3 image cosine orientation matrix shown in the .mih
header should be changed to reflect pure axial or sagittal or
coronal image transforms (integers of 0 or 1) (see
Note 23).
3. Masking. Masking is used to remove other parts of the image
which are not part of the study and improve the quality of
subsequent analyses. For DWI, the mask may be derived from
an average of all b-weighted images (excluding the b ¼ 0
images), as they normally have adequate suppression of
non-CNS tissues. The mask may be generated using region
growing using ITK-SNAP (itk-snap.org) using the average of

the diffusion-weighted images (excluding the b0) (see Note 24).
4. Image bias correction. N3 or N4 image bias correction can be
used using the modules within the program Slicer (slicer.org)
or Advanced Normalization Tools (ANTs). For the T1 anato-
mical image, bias correction can be done directly on the image.
For DWI data, the bias field is calculated from the b ¼ 0 image,
and the same bias correction is then applied to normalize the
diffusion-weighted images (i.e., by dividing all images with the
same bias field).
5. Eddy current and motion correction. These can be performed
using FSL’s eddy and MCFLIRT (motion correction linear
registration) programs (see Notes 25 and 26).
6. Correction of the diffusion encoding directions. During data
format conversion, the image orientation axes and their polar-
ity may become interchanged. This may be checked by com-
paring the matrix size of each read/phase/slice axes in the
method file and in the image header (in Bruker the r, p, s axes
are equal to x, y, z axes). Failure to use the correct the diffusion
gradient vector will result in false fibertracking and down-
stream parameter quantification (!) (see Note 27).
7. Calculations of diffusion parameters. Although the MRI con-
sole (e.g., Paravision) can often perform standard online DTI
processing (see Note 28), diffusion parameters are normally
processed offline using freely available programs such as Diffu-
sion Toolkit/TrackVis (trackvis.org) and MRTrix (mrtrix.org).
MRTrix software can also perform pre-processing steps for
correcting the motion and geometry distortion by calling the
appropriate FSL functions described above prior to the calcu-
lation of the diffusion parameters (see Notes 29 and 30).
DTK/Trackvis can perform diffusion modeling using DTI,
Q-ball and diffusion spectrum imaging (DSI) [80]. MRTrix
can calculate DTI and constrained spherical deconvolution
(CSD) modeling [70].
8. Fibertracking. Higher diffusion modeling using Q-ball, DSI
and CSD allow more accurate fibertracking through the areas
with crossing fibers. The reconstructed fibertracks/streamlines
can be directly quantified [81], or to generate a connectome
matrix for measuring changes in brain connectivity using
network-based analysis [82].
To perform fibertracking, it is important that the tensors
and orientation distribution function (ODF) have been recon-
structed correctly. An example of correct mouse brain MRTrix
ODF and fibertracking of the corpus callosum is shown in
Fig. 5. Perfect alignment of ODF can be carefully inspected
using the ODF tool in MRView, or by displaying the result of
FSL DTIFIT V1 map in FSLView.
Fig. 5 ODF map of an ex vivo mouse brain and generated streamlines of the corpus callosum. (a) and (b) are
ODF for the axial and transverse slices. Red, green, blue (RGB) indicate ODF directions in dorso/ventral,
medial/lateral, and rostro/caudal, respectively. Note that although the RGB colors themselves may be
interchangeable, it is the shape and the direction of the ODF (or the tensor) that are critical for fibertracking.
(c) Corpus callosum streamlines generated from the midline ROI using probabilistic iFOD2 fibertracking in
MRTRix 3. HARDI data was acquired ex vivo using 30 diffusion encoding direction at 100 μm isotropic
resolution at 16.4 T
Fig. 6 Ex vivo C56 BL6J mouse brain MRI template and atlas. Cortical and basal ganglia atlas are overlaid on
the Australian Mouse Brain Mapping Consortium (AMBMC) 15 μm resolution template [92, 93]
3.4.3 ROI Analyses ROI analyses may be performed directly using an offline version of
Bruker Paravision. Alternatively, MRI data can be converted into
DICOM for analysis using freely available programs such as MRI-
cron, Osirix, MIPAV, or ITK-SNAP. ROIs are typically segmented
using the freehand tools within these programs. Experts blinded to
group assignments and/or independent raters may be required to
do manual segmentation as this method can be subjective and the
results may vary between operators.
3.4.4 Voxel-Based 1. Choice of MRI atlases. There are several atlases available for
Analyses of Structural ex vivo mouse brain (brainatlas.mbi.ufl.edu; imaging.org.au/
Images AMBMC/AMBMC; mouseimaging.ca/research/mouse_atlas.
html; lbam.med.jhmi.edu) and in vivo mouse brain MRI
(brainatlas.mbi.ufl.edu). An example of the atlas segmentation
showing cortical and basal ganglia segmentations of an adult C57
BL6 brain atlas (AMBMC) is shown in Fig. 6 (see Note 31).
2. Registration of structural images. These steps may be done

using the modules within FSL (fsl.fmrib.ox.ac.uk/fsl/fslwiki/
FSL) or ANTs (stnava.github.io/ANTs/). Various levels of
registrations with increasing complexity need to be performed.
This typically involves reorientation of the study image into the
atlas space (or vice versa), and then followed by finer normali-
zation steps.
a. A linear rigid body is first used to obtain a global alignment
between the individual datasets and the atlas. Typically max-
imum orientation rotation sampling (180 ) in x, y, z axes
are performed to obtain a good initial alignment. This may
be done using FSL FLIRT, which is easy to use and achieve
good results.
b. A linear affine registration, followed by a nonlinear registra-
tion are then used to normalize the atlas into the individual
data. This can be done using FSL FLIRT-FNIRT, or using
ANTs antsIntroduction script.
c. Shadow registrations (FSL applywarp or ANTs ImageWarp-
MultiTransform) can then be used to normalize the atlas
segmented structures into the dataset, and the size of the
structures can be measured using FSL fslstats or ITK-SNAP.
Alternatively, a template study of T1-weighted image can be
generated using the whole study data using ANTs buildtem-
plateparallel.sh. To map the atlas onto the individual data,
the chosen MRI atlas is first normalized to the study tem-
plate, and it is subsequently transformed onto the individual
images.
3. Tensor-based morphometry. The Jacobian determinant from
nonlinear registration step (FNIRT or ANTS diffeomorphic
registrations) can be used to detect spatial expansion/shrink-
age [83]. The Jacobian determinant can be obtained by speci-
fying –jcout in FSL FNIRT or using ANTSJacobian, and group
comparison can be calculated using VBA (see Subheading
3.4.6).
3.4.5 Analysis of DTI There are several methods that may be used to normalize DTI data
Parametric Maps to the template.
1. Using the transformation matrix derived from the T1-weighted
image structural images to normalize the DTI data to the
template. This method may reduce registration bias due to
the disease, especially if there are significant morphological
changes in the WM structures that can affect registration.
2. Using one of DTI parametric maps (e.g., FA) to create the
study template (using a similar strategy as outlined for the
structural images, e.g., using ANTS buildtemplateparallel.sh),
and apply the resulting transformation matrices to the other

DTI parametric maps [84].
3. Using FSL tract-based spatial statistics (TBSS) to register ske-
letonized WM tracts (fsl.fmrib.ox.ac.uk/fsl/fslwiki/TBSS)
[85]. A study FA template will need to be first generated as
input for the registration process; this is then used to substitute
the human standard template in the FSL database (see Note
32). TBSS analysis is limited to the WM tracts as the FA value
>0.2 is usually used to generate the FA skeletons.
4. Using tensor-based registration DTI-TK program (dti-tk.
sourceforge.net/pmwiki/pmwiki.php) to calculate the study
template and normalization of DTI data [86]. DTI-TK
requires the tensor calculated using FSL FMRIB Diffusion
Toolbox (FDT) as the input for its registration process.
3.4.6 Statistical Analyses Statistical analyses may be performed using FSL randomize T-tests.
First, datasets need to be arranged as a 4D data (using FSL
fslmerge) and a design matrix needs to be created accordingly.
Randomize calculation is typically performed using >5000 permu-
tations, and the output is measured using a threshold-free cluster
enhancement with a family-wise error correction and p < 0.05.
4 Notes
1. For in vivo CNS imaging, electrocardiogram (ECG) monitor-

ing is normally not required. It often suffers from magnetic
field interference and gradient vibration during acquisition. To
circumvent deleterious current induction at high magnetic
field, an infrared monitoring apparatus (e.g., SA Instruments)
can be used to simultaneously monitor cardiac cycle, respira-
tory and PO2 monitoring.
2. To do cardiac perfusion fixation, animals are first euthanized
using Lethabarb (with a dose of ~1 μL/g body weight) or an
overdose of isoflurane. To ensure good perfusion, the proce-
dure should be performed when pedal reflexes are absent but
the cardiac pulsation is still present. The catheter is inserted
into the left ventricle, and cardiac perfusion is initiated with
room temperature phosphate buffered saline (PBS) at a flow-
rate of 5 mL/min (mouse), 10 mL/min (rat) or 1 mL/min for
neonates. Perfusion should be done with optimal flowrate so
that the capillaries are not expanded due to over pressure, but
not too low such that brain ventricle structures collapse. Foam
forming on the nose/mouth of the animal may indicate a
rupture in the lung capillaries, which can be avoided by reduc-
ing the flow rate.
The initial PBS perfusion is often performed using a vol-
ume at least equal to the animal body weight (v/w) to ensure
the blood is completely washed out. Subsequently, fixation is

performed using 4% paraformaldehyde (PFA) in PBS with the
same flow rate and volume.
3. Excess skin and muscle tissues should be removed from the
skull or the spine to reduce FOV. The quality of the fixation
should be inspected, where the tissues should appear pale and
rigid. The brain/spinal cord may be kept inside the skull/spine
and incubated in the fixation mixture for 16 h (overnight) at
4 C. Prolonged storage in PFA is possible, but this may result
in over fixation that can affect MRI and subsequent histology.
4. The following day, the brain/spinal cord can be extracted out
carefully using fine scissors to avoid damage. Damage not
visible with the eye will be pronounced in the MRI, it will
stop fiber tracking propagation and can affect image registra-
tion process. Keep the olfactory bulb, cerebellum and a short
segment of the spinal cord intact with the sample, as they will
be useful for downstream whole brain normalization. Samples
can be kept in PBS with 0.02% w/v sodium azide until MRI.
5. Ex vivo CNS tissue samples prepared using paraformaldehyde
cardiac perfusion fixation are normally preferred instead of
drop-fixed samples. Perfusion-fixed samples are free from red
blood cells and have better preservation of tissue morphology,
which will be important for subsequent histological study.
6. To monitor the animal respiration, the respiratory pad should
be placed just under the diaphragm, not close to the lung as the
breathing signals may be confounded by the fast heartbeat.
7. A high temperature water bath (set up at ~60 C) may be used
if the distance between the water bath and the probe in the
scanner is far, to keep the animal bed warm.
8. For tail vein catheterization insertion or bolus injection, a
heparin/PBS solution (50 unit/mL) may be used to wash the
syringe, needle and catheter tubing to avoid blood coagulation
in the catheter. The tail needs to be first warmed up to dilate
the veins. This may be done using a disposable glove prefilled
with warm water (~45 C) and placed over the tail for 2–5 min.
Correct insertion of the needle into the vein should be con-
firmed by the appearance of blood in the tubing when the vein
is gently pressed toward the needle. The catheter position can
be fixed using a small amount of acrylic glue or surgical tape.
9. In vertical scanners, the coronal, sagittal, and axial orientations
shown in the scan program may not match the animal anato-
mical axes, which can be confusing. These image orienta-
tions actually describe the image slices placed orthogonal to
the scanner x and y (transverse), and z (vertical) axes,
respectively.
10. The brain axial landmarks are the lines separating the left/right
of the brain cortical hemispheres (viewed dorso-ventrally, and
rostro-caudally), and the tip of the olfactory bulb to the center

of the cerebellum (viewed sagittally). Three-axes pilot scans
(Tripilot) can be used to iteratively refine these slice positions.
11. To obtain maximum signal and reduce geometrical distortion,
the brain should be first shimmed globally using a standard
free-induction decay (FID) first order and z2 shims, followed
by the second order shims. Subsequently, a fieldmap shim
protocol, which uses the scanner magnetic fieldmap (called
B0 mapping), can be used further to optimize the first and
second order shims using an ellipsoid shimming region placed
to cover the brain volume, or a cylinder to cover the spinal
cord. These shimming volumes should not include the skull
and spine structures.
12. The MRI parameters listed here, such as image resolution,
number of scan averages, diffusion weighting (b-values, δ/Δ),
TR, and TE, will need to be adjusted accordingly with respect
to the scanner sensitivity, the strength of imaging gradients and
tissue relaxation times to obtain optimal results [87, 88].
13. For higher order diffusion processing, the scripts and tutorials
for NODDI and ActiveAx can be found at http://mig.cs.ucl.
ac.uk/mig/mig/index.php/?n¼Tutorial.NODDImatlab/
and at http://camino.cs.ucl.ac.uk/index.php?n¼Tutorials.
ActiveAx.
14. In vivo T1-weighted imaging of mouse brain using conven-
tional hardware may produce images with low tissue contrast-
to-noise ratio (CNR) and SNR. T2-weighted imaging is often
used instead of T1-weighted imaging to produce better quality
data for segmentation [89]. These problems, however, can be
resolved when imaging is performed using a cryoprobe [28].
15. Shorter TE may be achieved by reducing the RF pulse dura-
tions (1–1.5 ms), decreasing the phase matrix size (<128),
increasing the image acquisition bandwidths (~100 kHz),
reducing the turbo/RARE factor in fast spin echo sequence,
and using partial Fourier acceleration.
16. T1 relaxation using partial saturation recovery can be calculated
using Eq. 1, using a very short TE (TE « T2).

TR TE
S ðt Þ ¼ S 0 1 e T 1 e T 2 ð1Þ
T1 relaxation using inversion recovery can be calculated using

Eq. 2, using TR » T1 or ~5 T1.

TI TR
S ðt Þ ¼ S 0 1 2e T 1 þ e T 1 ð2Þ
T2 and T2* relaxation times are obtained by fitting the signal

intensity at multiple TEs into an exponential decay Eq. 3. This
is derived the Eq. 1 using TR » T1.
TTE
SðtÞ ¼ S 0 e 2 ð3Þ
17. There is some limitation in using a long TE (approx. >15 ms)

for in vivo gradient echo sequence in high magnetic field, as it
can produce strong artifacts near the jaw, ear canal, and brain-
skull interfaces.
18. Occasionally air pockets may be trapped inside the tissues (e.g.,
in the ventricles or between the skull and the brain tissue).
Trapped air pockets produce local magnetic field susceptibility
artifacts, and thus they need to be removed. To remove the air,
place the Gd-DTPA/PBS in a tall beaker/tube in a vacuum
chamber. Apply the vacuum until all bubble disappear
(~5 min). Put the sample in this solution, and vacuum again
until significant bubbles appear (~1–2 min), but it is not nec-
essary to keep vacuuming until all bubbles disappear. Transfer
the sample into an imaging tube (see below) and fill with
Fomblin (Fig. 1a). If air is still found trapped inside the tissue
during the pilot MRI scans, the sample may be vacuumed again
while kept inside the imaging tube filled with Fomblin.
19. Fomblin is used to preserve ex vivo sample during scanning,
providing a homogenous magnetic field to improve shimming
and black background in the MR images. Fomblin is not sterile
but it may be reused for imaging several samples. After imag-
ing, samples should be washed in PBS and stored in PBS
solution containing sodium azide or PFA for a long-term
storage.
20. There is a mismatch between the levels of the spine and the
levels of the spinal cord [90]. The lumbar spinal cord can be
located approximately in the last ribcage segment, and it can be
identified as a segment with the largest butterfly structure in
the axial image.
21. For tandem ex vivo imaging, trimmed spinal cords can be
glued using an acrylic superglue (use sparingly) onto a plastic
board in 1 2 or 2 2 arrangement. This sample group can
then be inserted into a container tube and immersed in Fom-
blin. The samples may need to appear distinctly from the others
(at least when scanning for the first-time) so their positions can
be unambiguously identified.
22. The DWI processing methods described in this section are
generally sufficient to perform basic analyses. However, a
large number of higher DWI processing protocols and group
analyses are continuously developed and improved, and it is

difficult to provide complete protocol overview of all the pro-
grams available. Therefore, users are advised to check related
websites for more advanced protocols and recent variations.
23. Header corrections of the MR image transformation matrix
may be required for Bruker datasets, as the Paravision software
has already normalized the diffusion encoding directions with
respect to the read/phase/slice imaging axes. Therefore, an
oblique slice orientation in the header will need to be reset into
integer (0 or 1) cosine angles to ensure that correct diffu-
sion encoding directions are used for calculations.
24. An average of diffusion-weighted images only (without the b0
images) can be useful as a template to generate a mask as signals
for the muscle tissues have been suppressed by the diffusion
gradients. Masking must be done cautiously, especially if there
is a significant motion or eddy current distortion present. In
this case, the operator may consider to do eddy current and
motion corrections prior to masking.
25. The extent of eddy current artifact can be determined by image
shearing in the diffusion-weighted images compared to the b0
images, or by the presence of a halo around the brain in the
DTI processed images. The extent of eddy current distortions
in ex vivo and in vivo DWI spin-echo data at 16.4 T were
minimal, and they often did not require eddy correction.
26. Motion correction often needs to be done using rigid-body
registration with no rotation, as the animal motion is often
only associated with respiratory breathing (i.e., movement in
dorso/ventral direction). To obtain good results, these cor-
rections should be performed only for data which have good
SNR. Noisy data will result in poor spatial registration. Note
that at b-value ~2000 s/mm2, the signal from muscle tissue
surrounding the brain has largely disappear, such that FSL
eddy may mistakenly enlarge the brain tissue of the DW image
to match the size of the b0 image. This may be avoided by
masking the brain prior to eddy correction for each of the b0
and diffusion-weighted images.
27. The x, y, z diffusion encoding matrix in a sagittal DWI dataset
acquired using a vertical scanner and Paravision may be
swapped to z, y, x directions during data conversion, so that
the b-vector matrix will need to be corrected. Occasionally,
the polarity of the diffusion encoding axis may also need to be
corrected to obtain the final –z, y, x diffusion gradient vectors.
For data acquired in the axial plane using a horizontal scan-
ner, this swap may not be required.
28. Besides using the offline processing software, DTI FA and

eigenvalues (λ1, 2, 3) can also be calculated at the scanner
console. These eigenvalues can then be converted into AD
(¼λ1), RD (¼average of λ2, λ3) and MD (¼average of λ1, λ2,
λ3), and FA is calculated using the Eq. 4:
rffiffiffisffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
ffi
1 ðλ1 λ2 Þ2 þ ðλ1 λ3 Þ2 þ ðλ2 λ3 Þ2
FA ¼ ð4Þ
2 λ1 2 þ λ2 2 þ λ3 2
29. MR image intensity is often inhomogeneous due to the RF

and magnetic field inhomogeneities. For volume coil, the
signal intensity is typically maximum in the center of the coil
and dropping off toward the edge. For surface coils, the signal
intensity is usually maximum close to coil and dropping off
away from the coil.
30. DWI image intensity correction is especially important to
obtain accurate amplitude of the fiber orientation distribution
(FOD) in MRTrix. For statistical analysis using MRTrtix
apparent fiber density (AFD) [91], the image intensity of
the whole dataset needs to be normalized to allow group
comparison.
31. Brain MRI atlases were generated with variable methods,
pulse sequence, image contrasts and resolutions. These atlases
also have variable details in the structural segmentations.
Therefore, they need to be inspected individually in compari-
son to the acquired data, so that the appropriate template can
be chosen to fit the study requirement.
32. Many MRI processing programs have been optimized to work
with human MRI data whose image resolutions are typically
in the range of 1–2 mm. As a result, the animal MRI data
resolution may need to be “adjusted” to the human data
resolution (i.e., typically using a multiplication by a factor of
10 or 20 from the native MRI data resolution) prior to
calculations, if using the software default values. Alternatively,
the registration regularization parameters may be modified to
accommodate high-resolution animal data.
Acknowledgements
N.K. thanked the Queensland Government and Australian Federal

Government for funding and operational support of the 16.4T
NMR spectrometer through the QLD NMR Network (QNN)
and the National Imaging Facility (NIF).
References
1. Neumann-Haefelin T, Kastrup A, de cord injury. Brain Sci 7(2). https://doi.org/
Crespigny A, Yenari MA, Ringer T, Sun GH, 10.3390/brainsci7020021
Moseley ME (2000) Serial MRI after transient 10. Kim JH, Loy DN, Wang Q, Budde MD,
focal cerebral ischemia in rats: dynamics of tis- Schmidt RE, Trinkaus K, Song SK (2010) Dif-
sue injury, blood-brain barrier damage, and fusion tensor imaging at 3 hours after trau-
edema formation. Stroke 31(8):1965–1972; matic spinal cord injury predicts long-term
discussion 1972–1963 locomotor recovery. J Neurotrauma 27
2. Yan G, Dai Z, Xuan Y, Wu R (2015) Early (3):587–598. https://doi.org/10.1089/neu.
metabolic changes following ischemia onset in 2009.1063
rats: an in vivo diffusion-weighted imaging and 11. Cowin GJ, Butler TJ, Kurniawan ND,
1H-magnetic resonance spectroscopy study at Watson C, Wallace RH (2011) Magnetic reso-
7.0T. Mol Med Rep 11(6):4109–4114. nance microimaging of the spinal cord in the
https://doi.org/10.3892/mmr.2015.3283 SOD1 mouse model of amyotrophic lateral
3. Hutchinson EB, Schwerin SC, Radomski KL, sclerosis detects motor nerve root degenera-
Irfanoglu MO, Juliano SL, Pierpaoli CM tion. Neuroimage 58(1):69–74. https://doi.
(2016) Quantitative MRI and DTI abnormal- org/10.1016/j.neuroimage.2011.06.003
ities during the acute period following CCI in 12. Underwood C, Kurniawan ND, Butler T,
the ferret. Shock 46(3 Suppl 1):167–176. Cowin GJ, Wallace RA (2011) Non-invasive
https://doi.org/10.1097/SHK. diffusion tensor imaging detects white matter
0000000000000659 degeneration in the spinal cord of a mouse
4. Kharatishvili I, Immonen R, Grohn O, Pitka- model of amyotrophic lateral sclerosis. Neuro-
nen A (2007) Quantitative diffusion MRI of image 55(2):455–461. https://doi.org/10.
hippocampus as a surrogate marker for post- 1016/j.neuroimage.2010.12.044
traumatic epileptogenesis. Brain 130 13. Absinta M, Sati P, Reich DS (2016) Advanced
(Pt 12):3155–3168. https://doi.org/10. MRI and staging of multiple sclerosis lesions.
1093/brain/awm268 Nat Rev Neurol 12(6):358–368. https://doi.
5. Long JA, Watts LT, Chemello J, Huang S, org/10.1038/nrneurol.2016.59
Shen Q, Duong TQ (2015) Multiparametric 14. Hunger M, Budinger E, Zhong K, Angenstein
and longitudinal MRI characterization of mild F (2014) Visualization of acute focal lesions in
traumatic brain injury in rats. J Neurotrauma rats with experimental autoimmune encephalo-
32(8):598–607. https://doi.org/10.1089/ myelitis by magnetic nanoparticles, comparing
neu.2014.3563 different MRI sequences including phase imag-
6. Blomster LV, Cowin GJ, Kurniawan ND, Rui- ing. J Magn Reson Imaging 39(5):1126–1135.
tenberg MJ (2013) Detection of endogenous https://doi.org/10.1002/jmri.24280
iron deposits in the injured mouse spinal cord 15. Nathoo N, Agrawal S, Wu Y, Haylock-Jacobs S,
through high-resolution ex vivo and in vivo Yong VW, Foniok T, Barnes S, Obenaus A,
MRI. NMR Biomed 26(2):141–150. https:// Dunn JF (2013) Susceptibility-weighted imag-
doi.org/10.1002/nbm.2829 ing in the experimental autoimmune encepha-
7. Brennan FH, Cowin GJ, Kurniawan ND, Rui- lomyelitis model of multiple sclerosis indicates
tenberg MJ (2013) Longitudinal assessment of elevated deoxyhemoglobin, iron deposition
white matter pathology in the injured mouse and demyelination. Mult Scler 19
spinal cord through ultra-high field (16.4T) (6):721–731. https://doi.org/10.1177/
in vivo diffusion tensor imaging. Neuroimage 1352458512460602
82:574–585. https://doi.org/10.1016/j. 16. Tambalo S, Peruzzotti-Jametti L, Rigolio R,
neuroimage.2013.06.019 Fiorini S, Bontempi P, Mallucci G,
8. Brennan FH, Kurniawan ND, Vukovic J, Bart- Balzarotti B, Marmiroli P, Sbarbati A,
lett PF, Kasermann F, Arumugam TV, Basta M, Cavaletti G, Pluchino S, Marzola P (2015)
Ruitenberg MJ (2016) IVIg attenuates com- Functional magnetic resonance imaging of
plement and improves spinal cord injury out- rats with experimental autoimmune encephalo-
comes in mice. Ann Clin Transl Neurol 3 myelitis reveals brain cortex remodeling. J
(7):495–511. https://doi.org/10.1002/ Neurosci 35(27):10088–10100. https://doi.
acn3.318 org/10.1523/JNEUROSCI.0540-15.2015
9. Sun P, Murphy RK, Gamble P, George A, Song 17. Colgan N, Siow B, O’Callaghan JM, Harrison
SK, Ray WZ (2017) Diffusion assessment of IF, Wells JA, Holmes HE, Ismail O,
cortical changes, induced by traumatic spinal Richardson S, Alexander DC, Collins EC,
Fisher EM, Johnson R, Schwarz AJ, Ahmed Z, Longitudinal characterization of brain atrophy
O’Neill MJ, Murray TK, Zhang H, Lythgoe of a Huntington’s disease mouse model by
MF (2016) Application of neurite orientation automated morphological analyses of magnetic
dispersion and density imaging (NODDI) to a resonance images. Neuroimage 49
tau pathology model of Alzheimer’s disease. (3):2340–2351. https://doi.org/10.1016/j.
Neuroimage 125:739–744. https://doi.org/ neuroimage.2009.10.027
10.1016/j.neuroimage.2015.10.043 25. Yang S, Yang Z, Fischer K, Zhong K, Stadler J,
18. Li L, Wang XY, Gao FB, Wang L, Xia R, Li ZX, Godenschweger F, Steiner J, Heinze HJ, Bern-
Xing W, Tang BS, Zeng Y, Zhou GF, Zhou stein HG, Bogerts B, Mawrin C, Reutens DC,
HY, Liao WH (2016) Magnetic resonance T2 Speck O, Walter M (2013) Integration of ultra-
relaxation time at 7 Tesla associated with amy- high field MRI and histology for connectome
loid beta pathology and age in a double- based research of brain disorders. Front Neu-
transgenic mouse model of Alzheimer’s dis- roanat 7:31. https://doi.org/10.3389/fnana.
ease. Neurosci Lett 610:92–97. https://doi. 2013.00031
org/10.1016/j.neulet.2015.10.058 26. Spencer RG, Fishbein KW, Cheng A, Mattson
19. Wells JA, O’Callaghan JM, Holmes HE, MP (2006) Compatibility of Gd-DTPA perfu-
Powell NM, Johnson RA, Siow B, sion and histologic studies of the brain. Magn
Torrealdea F, Ismail O, Walker-Samuel S, Reson Imaging 24(1):27–31. https://doi.org/
Golay X, Rega M, Richardson S, Modat M, 10.1016/j.mri.2005.10.017
Cardoso MJ, Ourselin S, Schwarz AJ, 27. Webb AG, Van de Moortele PF (2016) The
Ahmed Z, Murray TK, O’Neill MJ, Collins technological future of 7 T MRI hardware.
EC, Colgan N, Lythgoe MF (2015) In vivo NMR Biomed 29(9):1305–1315. https://doi.
imaging of tau pathology using multi- org/10.1002/nbm.3315
parametric quantitative MRI. Neuroimage 28. Baltes C, Radzwill N, Bosshard S, Marek D,
111:369–378. https://doi.org/10.1016/j. Rudin M (2009) Micro MRI of the mouse
neuroimage.2015.02.023 brain using a novel 400 MHz cryogenic quad-
20. Antonsen BT, Jiang Y, Veraart J, Qu H, rature RF probe. NMR Biomed 22
Nguyen HP, Sijbers J, von Horsten S, Johnson (8):834–842. https://doi.org/10.1002/nbm.
GA, Leergaard TB (2013) Altered diffusion 1396
tensor imaging measurements in aged trans- 29. Bosshard SC, Baltes C, Wyss MT, Mueggler T,
genic Huntington disease rats. Brain Struct Weber B, Rudin M (2010) Assessment of brain
Funct 218(3):767–778. https://doi.org/10. responses to innocuous and noxious electrical
1007/s00429-012-0427-0 forepaw stimulation in mice using BOLD
21. Blockx I, De Groof G, Verhoye M, Van fMRI. Pain 151(3):655–663. https://doi.
Audekerke J, Raber K, Poot D, Sijbers J, org/10.1016/j.pain.2010.08.025
Osmand AP, Von Horsten S, Van der Linden 30. Wu D, Xu J, McMahon MT, van Zijl PC,
A (2012) Microstructural changes observed Mori S, Northington FJ, Zhang J (2013) In
with DKI in a transgenic Huntington rat vivo high-resolution diffusion tensor imaging
model: evidence for abnormal neurodevelop- of the mouse brain. Neuroimage 83:18–26.
ment. Neuroimage 59(2):957–967. https:// https://doi.org/10.1016/j.neuroimage.
doi.org/10.1016/j.neuroimage.2011.08.062 2013.06.012
22. Sawiak SJ, Wood NI, Williams GB, Morton AJ, 31. Adams S, Pacharinsak C (2015) Mouse anes-
Carpenter TA (2009) Voxel-based morphome- thesia and analgesia. Curr Protoc Mouse Biol 5
try in the R6/2 transgenic mouse reveals dif- (1):51–63. https://doi.org/10.1002/
ferences between genotypes not seen with 9780470942390.mo140179
manual 2D morphometry. Neurobiol Dis 33
(1):20–27. https://doi.org/10.1016/j.nbd. 32. Colby LA, Morenko BJ (2004) Clinical con-
2008.09.016 siderations in rodent bioimaging. Comp Med
54(6):623–630
23. Sawiak SJ, Wood NI, Williams GB, Morton AJ,
Carpenter TA (2013) Voxel-based morphome- 33. Lukasik VM, Gillies RJ (2003) Animal anaes-
try with templates and validation in a mouse thesia for in vivo magnetic resonance. NMR
model of Huntington’s disease. Magn Reson Biomed 16(8):459–467. https://doi.org/10.
Imaging 31(9):1522–1531. https://doi.org/ 1002/nbm.836
10.1016/j.mri.2013.06.001 34. McConville P (2011) Small animal preparation
24. Zhang J, Peng Q, Li Q, Jahanshad N, Hou Z, and handling in MRI. Methods Mol Biol
Jiang M, Masuda N, Langbehn DR, Miller MI, 771:89–113. https://doi.org/10.1007/978-
Mori S, Ross CA, Duan W (2010) 1-61779-219-9_5
35. Kim JH, Song SK (2013) Diffusion tensor model of Parkinson’s disease. PLoS One 11(3):
imaging of the mouse brainstem and cervical e0151884. https://doi.org/10.1371/journal.
spinal cord. Nat Protoc 8(2):409–417 pone.0151884
36. Hedlund LW, Cofer GP, Owen SJ, Allan John- 46. Kara F, Hofling C, Rossner S, Schliebs R, Van
son G (2000) MR-compatible ventilator for der Linden A, Groot HJ, Alia A (2015) In vivo
small animals: computer-controlled ventilation longitudinal monitoring of changes in the cor-
for proton and noble gas imaging. Magn Reson pus callosum integrity during disease progres-
Imaging 18(6):753–759 sion in a mouse model of Alzheimer’s disease.
37. Kurniawan ND, Richards KL, Yang Z, She D, Curr Alzheimer Res 12(10):941–950
Ullmann JFP, Moldrich RX, Liu S, Yaksic JU, 47. Rudko DA, Klassen LM, de Chickera SN, Gati
Leanage G, Kharatishvili I, Wimmer V, JS, Dekaban GA, Menon RS (2014) Origins of
Calamante F, Galloway GJ, Petrou S, Reutens R2* orientation dependence in gray and white
DC (2014) Visualization of mouse barrel cor- matter. Proc Natl Acad Sci U S A 111(1):
tex using ex-vivo track density imaging. Neuro- E159–E167. https://doi.org/10.1073/pnas.
image 87:465–475. https://doi.org/10. 1306516111
1016/j.neuroimage.2013.09.030 48. Thiessen JD, Zhang Y, Zhang H, Wang L,
38. Ullmann JF, Calamante F, Collin SP, Reutens Buist R, Del Bigio MR, Kong J, Li XM, Martin
DC, Kurniawan ND (2015) Enhanced charac- M (2013) Quantitative MRI and ultrastruc-
terization of the zebrafish brain as revealed by tural examination of the cuprizone mouse
super-resolution track-density imaging. Brain model of demyelination. NMR Biomed 26
Struct Funct 220(1):457–468. https://doi. (11):1562–1581. https://doi.org/10.1002/
org/10.1007/s00429-013-0667-7 nbm.2992
39. Huang S, Liu C, Dai G, Kim YR, Rosen BR 49. Zhang J, Aggarwal M, Mori S (2012) Struc-
(2009) Manipulation of tissue contrast using tural insights into the rodent CNS via diffusion
contrast agents for enhanced MR microscopy tensor imaging. Trends Neurosci 35
in ex vivo mouse brain. Neuroimage 46 (7):412–421. https://doi.org/10.1016/j.
(3):589–599. https://doi.org/10.1016/j. tins.2012.04.010
neuroimage.2009.02.027 50. Mori S, Zhang J (2006) Principles of diffusion
40. Kim S, Pickup S, Hsu O, Poptani H (2009) tensor imaging and its applications to basic
Enhanced delineation of white matter struc- neuroscience research. Neuron 51
tures of the fixed mouse brain using (5):527–539. https://doi.org/10.1016/j.neu
Gd-DTPA in microscopic MRI. NMR Biomed ron.2006.08.012
22(3):303–309. https://doi.org/10.1002/ 51. Choi IY, Lee SP, Guilfoyle DN, Helpern JA
nbm.1324 (2003) In vivo NMR studies of neurodegener-
41. D’Arceuil H, de Crespigny A (2007) The ative diseases in transgenic and rodent models.
effects of brain tissue decomposition on diffu- Neurochem Res 28(7):987–1001
sion tensor imaging and tractography. Neuro- 52. Robinson S, Berglass JB, Denson JL, Berkner J,
image 36(1):64–68. https://doi.org/10. Anstine CV, Winer JL, Maxwell JR, Qiu J,
1016/j.neuroimage.2007.02.039 Yang Y, Sillerud LO, Meehan WP 3rd,
42. Kim TH, Zollinger L, Shi XF, Rose J, Jeong EK Mannix R, Jantzie LL (2016) Microstructural
(2009) Diffusion tensor imaging of ex vivo and microglial changes after repetitive mild
cervical spinal cord specimens: the immediate traumatic brain injury in mice. J Neurosci
and long-term effects of fixation on diffusivity. Res. https://doi.org/10.1002/jnr.23848
Anat Rec (Hoboken) 292(2):234–241. 53. Ou X, Sun SW, Liang HF, Song SK, Gochberg
https://doi.org/10.1002/ar.20823 DF (2009) Quantitative magnetization transfer
43. Schwartz ED, Cooper, Chin CL, Wehrli S, measured pool-size ratio reflects optic nerve
Tessler A, Hackney DB (2005) Ex vivo evalua- myelin content in ex vivo mice. Magn Reson
tion of ADC values within spinal cord white Med 61(2):364–371. https://doi.org/10.
matter tracts. AJNR Am J Neuroradiol 26 1002/mrm.21850
(2):390–397 54. Perez-Torres CJ, Reynolds JO, Pautler RG
44. Alonso-Ortiz E, Levesque IR, Pike GB (2015) (2014) Use of magnetization transfer contrast
MRI-based myelin water imaging: A technical MRI to detect early molecular pathology in
review. Magn Reson Med 73(1):70–81. Alzheimer’s disease. Magn Reson Med 71
https://doi.org/10.1002/mrm.25198 (1):333–338. https://doi.org/10.1002/
45. Cong L, Muir ER, Chen C, Qian Y, Liu J, Biju mrm.24665
KC, Clark RA, Li S, Duong TQ (2016) Multi- 55. Biedermann S, Fuss J, Zheng L, Sartorius A,
modal MRI evaluation of the MitoPark mouse Falfan-Melgoza C, Demirakca T, Gass P,
Ende G, Weber-Fahr W (2012) In vivo voxel (1):26–37. https://doi.org/10.1179/

based morphometry: detection of increased 1743132813Y.0000000269
hippocampal volume and decreased glutamate 64. Leten C, Struys T, Dresselaers T, Himmelreich
levels in exercising mice. Neuroimage 61 U (2014) In vivo and ex vivo assessment of the
(4):1206–1212. https://doi.org/10.1016/j. blood brain barrier integrity in different glio-
neuroimage.2012.04.010 blastoma animal models. J Neurooncol 119
56. Sack M, Lenz JN, Jakovcevski M, Biedermann (2):297–306. https://doi.org/10.1007/
SV, Falfan-Melgoza C, Deussing J, s11060-014-1514-2
Bielohuby M, Bidlingmaier M, Pfister F, Stalla 65. Pham M, Helluy X, Braeuninger S, Jakob P,
GK, Sartorius A, Gass P, Weber-Fahr W, Fuss J, Stoll G, Kleinschnitz C, Bendszus M (2010)
Auer MK (2016) Early effects of a high-caloric Outcome of experimental stroke in C57Bl/6
diet and physical exercise on brain volumetry and Sv/129 mice assessed by multimodal ultra-
and behavior: a combined MRI and histology high field MRI. Exp Transl Stroke Med 2:6.
study in mice. Brain Imaging Behav. https:// https://doi.org/10.1186/2040-7378-2-6
doi.org/10.1007/s11682-016-9638-y 66. Liu C, Li W, Johnson GA, Wu B (2011) High-
57. Giuliani NR, Calhoun VD, Pearlson GD, field (9.4 T) MRI of brain dysmyelination by
Francis A, Buchanan RW (2005) Voxel-based quantitative mapping of magnetic susceptibil-
morphometry versus region of interest: a com- ity. Neuroimage 56(3):930–938. https://doi.
parison of two methods for analyzing gray mat- org/10.1016/j.neuroimage.2011.02.024
ter differences in schizophrenia. Schizophr Res 67. Argyridis I, Li W, Johnson GA, Liu C (2014)
74(2–3):135–147. https://doi.org/10.1016/ Quantitative magnetic susceptibility of the
j.schres.2004.08.019 developing mouse brain reveals microstructural
58. Testa C, Laakso MP, Sabattoli F, Rossi R, changes in the white matter. Neuroimage
Beltramello A, Soininen H, Frisoni GB 88:134–142. https://doi.org/10.1016/j.
(2004) A comparison between the accuracy of neuroimage.2013.11.026
voxel-based morphometry and hippocampal 68. Haacke EM, Liu S, Buch S, Zheng W, Wu D,
volumetry in Alzheimer’s disease. J Magn Ye Y (2015) Quantitative susceptibility
Reson Imaging 19(3):274–282. https://doi. mapping: current status and future directions.
org/10.1002/jmri.20001 Magn Reson Imaging 33(1):1–25. https://
59. Bammer R, Holdsworth SJ, Veldhuis WB, doi.org/10.1016/j.mri.2014.09.004
Skare ST (2009) New methods in diffusion- 69. Klohs J, Deistung A, Schweser F, Grandjean J,
weighted and diffusion tensor imaging. Magn Dominietto M, Waschkies C, Nitsch RM,
Reson Imaging Clin N Am 17(2):175–204. Knuesel I, Reichenbach JR, Rudin M (2011)
https://doi.org/10.1016/j.mric.2009.01. Detection of cerebral microbleeds with quanti-
011 tative susceptibility mapping in the ArcAbeta
60. Alomair OI, Brereton IM, Smith MT, Gallo- mouse model of cerebral amyloidosis. J Cereb
way GJ, Kurniawan ND (2015) In vivo high Blood Flow Metab 31(12):2282–2292.
angular resolution diffusion-weighted imaging https://doi.org/10.1038/jcbfm.2011.118
of mouse brain at 16.4 Tesla. PLoS One 10(6): 70. Tournier JD, Mori S, Leemans A (2011) Dif-
e0130133. https://doi.org/10.1371/journal. fusion tensor imaging and beyond. Magn
pone.0130133 Reson Med 65(6):1532–1556. https://doi.
61. Kharatishvili I, Shan ZY, She DT, Foong S, org/10.1002/mrm.22924
Kurniawan ND, Reutens DC (2014) MRI 71. Tournier JD, Calamante F, Connelly A (2012)
changes and complement activation correlate MRtrix: Diffusion tractography in crossing
with epileptogenicity in a mouse model of tem- fiber regions. Int J Imaging Syst Technol
poral lobe epilepsy. Brain Struct Funct 219 22:53–66
s00429-013-0528-4 72. Foley LM, Hitchens TK, Barbe B, Zhang F,
Ho C, Rao GR, Nemoto EM (2010) Quanti-
62. Badea A, Ali-Sharief AA, Johnson GA (2007) tative temporal profiles of penumbra and
Morphometric analysis of the C57BL/6 J infarction during permanent middle cerebral
mouse brain. Neuroimage 37(3):683–693. artery occlusion in rats. Transl Stroke Res 1
https://doi.org/10.1016/j.neuroimage. (3):220–229. https://doi.org/10.1007/
2007.05.046 s12975-010-0032-6
63. Frey L, Lepkin A, Schickedanz A, Huber K, 73. Zhang S, Yao Y, Shi J, Tang X, Zhao L, Zhu W
Brown MS, Serkova N (2014) ADC mapping (2016) The temporal evolution of diffusional
and T1-weighted signal changes on post-injury kurtosis imaging in an experimental middle
MRI predict seizure susceptibility after experi- cerebral artery occlusion (MCAO) model.
mental traumatic brain injury. Neurol Res 36
Magn Reson Imaging 34(7):889–895. 83. Lepore N, Brun CA, Chiang MC, Chou YY,
https://doi.org/10.1016/j.mri.2016.04.016 Dutton RA, Hayashi KM, Lopez OL, Aizen-
74. Douaud G, Jbabdi S, Behrens TE, Menke RA, stein HJ, Toga AW, Becker JT, Thompson PM
Gass A, Monsch AU, Rao A, Whitcher B, (2006) Multivariate statistics of the Jacobian
Kindlmann G, Matthews PM, Smith S (2011) matrices in tensor based morphometry and
DTI measures in crossing-fibre areas: increased their application to HIV/AIDS. Med Image
diffusion anisotropy reveals early white matter Comput Comput Assist Interv 9
alteration in MCI and mild Alzheimer’s dis- (Pt 1):191–198
ease. Neuroimage 55(3):880–890. https:// 84. Schwarz CG, Reid RI, Gunter JL, Senjem ML,
doi.org/10.1016/j.neuroimage.2010.12.008 Przybelski SA, Zuk SM, Whitwell JL, Vemuri P,
75. Zhang H, Schneider T, Wheeler-Kingshott Josephs KA, Kantarci K, Thompson PM, Peter-
CA, Alexander DC (2012) NODDI: Practical sen RC, Jack CR Jr, Alzheimer’s Disease Neu-
in vivo neurite orientation dispersion and den- roimaging Initiative (2014) Improved DTI
sity imaging of the human brain. Neuroimage registration allows voxel-based analysis that
61(4):1000–1016 outperforms tract-based spatial statistics. Neu-
76. Barazany D, Basser PJ, Assaf Y (2009) In vivo roimage 94:65–78. https://doi.org/10.1016/
measurement of axon diameter distribution in j.neuroimage.2014.03.026
the corpus callosum of rat brain. Brain 132 85. Smith SM, Johansen-Berg H, Jenkinson M,
(Pt 5):1210–1220. https://doi.org/10. Rueckert D, Nichols TE, Miller KL, Robson
1093/brain/awp042 MD, Jones DK, Klein JC, Bartsch AJ, Behrens
77. Dyrby TB, Sogaard LV, Hall MG, Ptito M, TE (2007) Acquisition and voxelwise analysis
Alexander DC (2013) Contrast and stability of multi-subject diffusion data with tract-based
of the axon diameter index from microstruc- spatial statistics. Nat Protoc 2(3):499–503.
ture imaging with diffusion MRI. Magn Reson https://doi.org/10.1038/nprot.2007.45
Med 70(3):711–721. https://doi.org/10. 86. Keihaninejad S, Zhang H, Ryan NS, Malone
1002/mrm.24501 IB, Modat M, Cardoso MJ, Cash DM, Fox
78. Daducci A, Canales-Rodriguez EJ, Zhang H, NC, Ourselin S (2013) An unbiased longitudi-
Dyrby TB, Alexander DC, Thiran JP (2015) nal analysis framework for tracking white mat-
Accelerated Microstructure Imaging via Con- ter changes using diffusion tensor imaging with
vex Optimization (AMICO) from diffusion application to Alzheimer’s disease. Neuro-
MRI data. Neuroimage 105:32–44. https:// image 72:153–163. https://doi.org/10.
doi.org/10.1016/j.neuroimage.2014.10.026 1016/j.neuroimage.2013.01.044
79. Sepehrband F, Alexander DC, Kurniawan ND, 87. DiFrancesco MW, Rasmussen JM, Yuan W,
Reutens DC, Yang Z (2016) Towards higher Pratt R, Dunn S, Dardzinski BJ, Holland SK
sensitivity and stability of axon diameter esti- (2008) Comparison of SNR and CNR for
mation with diffusion-weighted MRI. NMR in vivo mouse brain imaging at 3 and 7T
Biomed 29(3):293–308. https://doi.org/10. using well matched scanner configurations.
1002/nbm.3462 Med Phys 35(9):3972–3978. https://doi.
org/10.1118/1.2968092
80. Wedeen VJ, Wang RP, Schmahmann JD,
Benner T, Tseng WY, Dai G, Pandya DN, 88. van de Ven RC, Hogers B, van den Maagden-
Hagmann P, D’Arceuil H, de Crespigny AJ berg AM, de Groot HJ, Ferrari MD, Frants
(2008) Diffusion spectrum magnetic reso- RR, Poelmann RE, van der Weerd L, Kiihne
nance imaging (DSI) tractography of crossing SR (2007) T(1) relaxation in in vivo mouse
fibers. Neuroimage 41(4):1267–1277. brain at ultra-high field. Magn Reson Med 58
https://doi.org/10.1016/j.neuroimage. (2):390–395. https://doi.org/10.1002/mrm.
2008.03.036 21313
81. Moldrich RX, Pannek K, Hoch R, Rubenstein 89. Boretius S, Kasper L, Tammer R, Michaelis T,
JL, Kurniawan ND, Richards LJ (2010) Com- Frahm J (2009) MRI of cellular layers in mouse
parative mouse brain tractography of diffusion brain in vivo. Neuroimage 47(4):1252–1260.
magnetic resonance imaging. Neuroimage 51 https://doi.org/10.1016/j.neuroimage.
(3):1027–1036. https://doi.org/10.1016/j. 2009.05.095
neuroimage.2010.03.035 90. Harrison M, O’Brien A, Adams L, Cowin G,
82. Liu C, Li Y, Edwards TJ, Kurniawan ND, Ruitenberg MJ, Sengul G, Watson C (2013)
Richards LJ, Jiang T (2016) Altered structural Vertebral landmarks for the identification of
connectome in adolescent socially isolated spinal cord segments in the mouse. Neuro-
mice. Neuroimage 139:259–270. https://doi. image 68:22–29. https://doi.org/10.1016/j.
org/10.1016/j.neuroimage.2016.06.037 neuroimage.2012.11.048
91. Raffelt D, Tournier JD, Rose S, Ridgway GR, protocol and MRI atlas of the C57BL/6J
Henderson R, Crozier S, Salvado O, Connelly mouse neocortex. Neuroimage 78:196–203.
A (2012) Apparent Fibre Density: a novel mea- https://doi.org/10.1016/j.neuroimage.
sure for the analysis of diffusion-weighted mag- 2013.04.008
netic resonance images. Neuroimage 59 93. Ullmann JFP, Watson C, Janke AL, Kurniawan
(4):3976–3994. https://doi.org/10.1016/j. ND, Paxinos G, Reutens DC (2014) An MRI
neuroimage.2011.10.045 atlas of the mouse basal ganglia. Brain Struct
92. Ullmann JF, Watson C, Janke AL, Kurniawan Funct 219(4):1343–1353. https://doi.org/
ND, Reutens DC (2013) A segmentation 10.1007/s00429-013-0572-0
Chapter 22
MRI in the Study of Animal Models of Stroke

Pedro Ramos-Cabrer and Daniel Padro
Abstract
Stroke consists of the loss of cerebral functions resulting from the interruption of blood supply to a region
of the brain, and represents the second cause of death and the leading cause of major disability in adults in
Europe. Stroke is a very active field of research at preclinical and clinical levels, and Magnetic Resonance
Imaging (MRI) is one of the most powerful tools that scientist and clinicians have for the study of the onset,
evolution and consequences of this devastating disease, as well as for the monitoring of the success of
available treatments, or for the development of novel therapeutic strategies.
MRI can tackle the study of stroke from different points of view, and at scales ranging from subcellular to
systems biology level. Magnetic resonance spectroscopy (MRS) allows the noninvasive measurement of the
levels of principal metabolites in the brain, and how they change during the course of the disease, or in
response to therapy. Glutamate, in particular, is very important in the field of stroke. Several anatomical MR
techniques allow the characterization of the lesion volumes, the formation of cytotoxic and vasogenic
edema, changes in cerebral blood flow and volume, structural changes in gray and white matter, the
obtaining of the vascular architecture and status, etc. At functional level, diverse modalities of functional
MRI (fMRI) allow the assessment of the alteration in the function and organization of neuronal networks of
the subject under study, as a consequence of the disease or in response to treatment. Finally, emerging
imaging modalities that include temperature and pH mapping of the brain, imaging by chemical exchange
saturation transfer effect (CEST), all of them closely related to tissue status, or the use of contrast agents for
the targeting of tissue in theranostic approaches or for cell tracking studies in cell-based therapies, etc., are
only a few examples of the power and versatility of MRI as a definitive tool for the study of stroke.
In this work we will set our focus on preclinical imaging of stroke models, emphasizing the most
commonly used imaging modalities in a stroke-dedicated research laboratory. However, advanced techni-
ques will be briefly discussed, providing references to specialized literature for more advanced readers. Thus,
the aim of this chapter consist in the description of a simple imaging protocol for the study of the most
important and common aspects of stroke in a research laboratory.
Key words MRI, MRA, ADC, DWI, PWI, Stroke, Ischemia, Angiography, Perfusion, Diffusion
1 Introduction
Stroke is the second cause of death in Europe, and the leading cause
of major disability in adults (with 1484 DALY -Disability Adjusted
Life Years- per 100,000 population) [1, 2]. Ischemic stroke consists
of the loss of cerebral functions resulting from the interruption of
377
378 Pedro Ramos-Cabrer and Daniel Padro
blood supply to a region of the brain, following the occlusion of a

brain artery. Pharmacological thrombolysis (or surgical retrieving
of the obstruction) during the acute phase of stroke is the only
available effective treatment against cerebral ischemia. Although
very efficient, thrombolysis is not exempted from limitations, com-
plications, and risk of worsening patient’s outcome, leading even to
death [3, 4]. In this context, Magnetic Resonance Imaging (MRI)
plays an essential role for the study of the pathophysiology of the
disease, the diagnosis and management of ischemic subjects, and
for the development of alternative therapeutic approaches.
It is important to mention that stroke is a fast-changing condi-
tion. Therefore, tissue status and neuroimaging surrogates of such
status change within minutes after the onset of the disease [5]. The
knowledge of a full battery of imaging techniques, from vascular to
functional imaging, is as important as the knowledge of when to
apply which one, an aspect at which we will also pay special
attention.
An ischemic attack starts with the obstruction of a blood vessel;
thus the use of magnetic resonance angiography (MRA) is one of
the first imaging modalities to consider. MR angiograms can be
obtained either by contrast enhanced angiography (CE-MRA),
acquiring T1-weighted images after administration of an exoge-
nous vascular contrast agent (that is not able to cross the blood–-
brain barrier), or by arterial spin labeling techniques (ASL) on
which the RF-labeled water protons of blood are used as endoge-
nous contrast agent (see Note 1) [6]. Exogenous contrast agents
may be also used to study the integrity of the blood–brain barrier,
which may be transitorily or permanently disrupted during the
progression of stroke [7, 8].
The blocking of a brain blood vessel is followed by a drop in
tissue perfusion. The local reduction of the cerebral blood flow
(CBF) and the subsequent change in cerebral blood volume
(CBV) are key factors to determine the extension and severity of
stroke damage. Although technically challenging, Perfusion-
Weighted Imaging (PWI) is of paramount importance in the
study of stroke. We recommend a nice work from Barbier et al.
for a comprehensive review of perfusion imaging theoretical back-
ground, techniques, and applications [9].
The metabolism of brain tissue is so intense that within minutes
after a severe reduction in blood supply, profound regional meta-
bolic and structural changes are triggered in the brain. Brain metab-
olism may be noninvasively followed by using voxel-based magnetic
resonance spectroscopy (MRS) or chemical shift imaging (CSI)
spectroscopic techniques, with a particular focus on specific meta-
bolites, such as glutamate [10, 11] or with an overall analysis of
multiple metabolites [12]. In the particular case of glutamate, the
use of chemical exchange saturation transfer techniques (gluCEST)
allow the determination of glutamate levels though magnetization
MRI in Stroke 379
transfer effects between the amine protons of glutamate and bulk

water. This technique may play an important role in the near future
in the study of acute stroke [13].
When blood flow is reduced below certain threshold (see [5] for
a comprehensive correlation between blood flow thresholds and
physiopathological events during acute stroke) energy-dependent
pumps in cell membranes start to fail and cytotoxic edema takes
place (within minutes after stroke onset). This phenomenon alters
the diffusion behavior of water molecules in brain tissue. Although
these changes are sometimes minute, diffusion-weighted imaging
(DWI) is sensible enough to detect them. Indeed the estimation of
the apparent diffusion coefficient (ADC) of water molecules in the
brain is the most sensitive imaging technique to detect damaged
tissue in the very first moments after stroke onset. One can acquire
an MR image weighted by diffusion of water molecules, by applying
a pair of diffusion gradients in a Stejskal-Tanner approach [14],
which will show hyper-intense signal in the region of increased
cytotoxic edema (the ischemic core), where MR signal attenuation
by diffusion is lower due to restricted movement of water mole-
cules. Alternatively, one can acquire several images with different
signal attenuations, by applying diffusion gradients of different
magnitude, and then quantify absolute ADC (apparent diffusion
coefficient) values, which will lower in the region of increased
cytotoxic edema. These aspects will be analyzed in further detail
later in the experimental section (see Note 2).
A severe reduction of blood flow yields in irreversible damage
to brain tissue within minutes after the onset of stroke. This area,
referred as the infarct core, is well defined by using DWI-MRI, as
we have already mentioned. However, the area delineated as low
zones or reduced perfusion, by PWI, include not only the infarct
core but also areas where blood reduction levels are less severe
(areas of benign oligohemia) in which transitory effects in tissue
may happen, but that may be potentially reversed. Thus, during the
first hours after stroke onset there is a mismatch between PWI and
DWI images. The magnitude of such PWI-DWI mismatch (large at
the onset of stroke but progressively reduced to zero as the infarct
core grows with time) is a key factor that helps clinicians to take
decisions about submitting a patient to thrombolytic treatment,
despite the risks associated to it [15, 16].
Since diffusion is a tropic magnitude, the acquisition of diffu-
sion MRI data in a multidimensional space allows the creation of
diffusion tensors and tracks in tissue, providing maps of white
mater fibers of the brain. Original Diffusion Tensor Imaging
(DTI) approaches have evolved into a series of advanced white
matter imaging techniques, including Diffusion Kurtosis Imaging
(DKI), Diffusion Spectrum Imaging (DSI), Q-ball imaging, persis-
tent angular structure imaging (PAS-MRI), and others, reviewed in
[17] (see Note 3).
Hours after the onset of stroke, cytotoxic edema, accumulation

of glutamate and other pathologic events that take place at cellular
level, yield in the incapability of the cerebrovascular unit to main-
tain the integrity of the BBB, and fluid leaking from blood vessels to
the extracellular space result in the formation of vasogenic edema.
The presence of increased amounts of extracellular fluid in the
tissue can be visualized using T2- or T2*-weighted images (see
Note 5). Commonly, T2-weighted spin echo sequences are the
correct choice for imaging ischemic lesions from 4 to 6 h after the
onset of stroke to days or months later (see Notes 4 and 5, and
Fig. 1). Alternatively to conventional Spin-echo approaches,
T2-weighted images can be acquired by FLAIR (Fluid attenuation
by inversion recovery) sequences [18]. In FLAIR, signal of cere-
brospinal fluid (CSF) present at the ventricles and around the
meninges is attenuated, which may result advantageous (see Note
6). When multiple T2-weighted images are acquired at different
echo times it is possible to obtain parametric maps of the transverse
relaxation time (T2) that, besides the estimation of lesion volumes
can help us to distinguish different pathological statuses of the
tissue within the lesion area. Thus, it has been proposed that it is
possible to distinguish areas of necrosis from areas suffering selec-
tive neuronal death during the subacute phase of stroke, by follow-
ing the different T2 relaxation times evolution patterns [19]. Other
image contrast like T1 or T1ρ are less common in the study of
stroke models (see [12] for more details) (Fig. 1).
Finally, during the chronic phase of the disease most imaging
protocols are generally limited to the acquisition of T2- or T2*-
weighted imaging, to assess lesion volumes, and sometimes DTI, to
assess white matter alterations. Additionally, functional Magnetic
Resonance Imaging techniques (fMRI) represent also a remarkable
contribution of MRI for the diagnosis and treatment of stroke,
including the development of effective therapies [20]. Due to the
magnitude of the field, the contribution of fMRI to the study of
stroke will not be covered here.
Fig. 1 Evolution of T2 weighting contrast in the brain of a rat submitted to a transient (60 min) occlusion of the
middle cerebral artery (MCAo). The lesion is not (or barely) visible during the occlusion and the first hours after
it (hyper-acute and acute phases: t < 6–8 h). At subacute to chronic stages (t > 6–8 h), the lesion is clearly
visible as a hyper-intense area in the brain
MRI in Stroke 381
Table 1
Recommended imaging protocols at different stages of the progression of cerebral ischemia
MCAo stage Main assessments T2w T2* (SWI) DWI/ADC TOF-angio Total time
0 00
Pre- Discard abnormalities 9 36 – – 40 200 130 3800
During Cytotoxic edema, bleeding – 20 5900 80 000 40 200 150 100
<6 h post Reperfusion, hemorrhage – 20 5900 – 40 200 70 100
>6 h post Vasogenic edema, necrosis, 90 3600 20 5900 – – 130 3500
hemorrhage
In summary, stroke is a chain of events for which we can find

proper imaging surrogates through different MR imaging modal-
ities. The knowledge of which imaging modality needs to be used at
different stages of the progression of the diseases is the key to
success. In this work we describe simple imaging protocols at
different phases of ischemic stroke (see Table 1), that enable the
study of the most important and common aspects of stroke in a
research laboratory.
2 Materials
2.1 Rodent Models 1. Rats (250 50 g), preferred strains: Sprague Dawley, Wistar
of Cerebral Ischemia Kyoto.
2. Anesthetics: Isofluorane or Sevofluorane.
3. Appropriate anesthetic gas vaporizer.
4. Carrier gases for anesthesia: 1–1.5 L/min, N2:02
(70:30–80:20) mixtures, or medical air.
5. Set of general surgical tools.
6. Disposable material (syringes, needles, falcon tubes, etc.)
7. 0.9% NaCl solution for infusion (saline).
8. Betadine® dermic solution 10%.
9. Braided silk suture: 4/0, 6/0, 10/0.
10. Monofilament for MCA occlusion: set with 0.35 mm,
0.37 mm, 0.39 mm, 0.41 mm—4-0 medium MCAO suture
L34 PK10 (Doccol Corporation, Sharon, MA, USA).
11. Disposable low temperature ophthalmic cautery, 28 mm fine
tip (FIAB Spa, icchio—Firenze, Italy, ref. F7255).
12. Feedback controlled heating mat for surgery with rectal PT100
probe (Neoos Biotec S.L., Pamplona, Spain).
2.2 Magnetic 1. Anesthetics: Isofluorane or Sevofluorane.

Resonance Imaging 2. Appropriate anesthetic gas vaporizer.
3. Carrier gases for anesthesia: 1–1.5 L/min, N2:02
(70:30–80:20) mixtures, or medical air.
4. Bruker Biospec 70/30 USR scanner (Bruker Biospin GmbH,
Ettlingen, Germany).
5. 400 mT/m BGA-S 12 gradient insert (ID ¼ 114 mm, slew
rate: 3440 T/m/s) (Bruker Biospin GmbH, Ettlingen,
Germany).
6. MR compatible animal holder (Bruker Biospin GmbH, Ettlin-
gen, Germany).
7. Volumetric 7 cm internal diameter RF transmit coil combined
with 1 element or 4 arrayed elements surface RF Receive coils,
designed for rat head. Alternatively, 4 cm internal diameter
transmit/receive volumetric coil. (Bruker Biospin GmbH,
Ettlingen, Germany, or any other vendor).
8. SAII Model 1030 Monitoring & Gating System (Small Animal
Instruments Inc., Stony Brook, NY, USA).
9. Circulating warm water bath and mat.
2.3 Image Analysis 1. A general purpose computer (In our case, HP Z230 Worksta-
tion with an Intel® Xenon® CPU E3–1246 v3 @ 3.5 GHz,
8 Gb RAM running a 64-bit Windows 7 professional operative
system).
2. Image-J V1.50b, the open-source and publically available
image processing software package for JAVA of the National
Institutes of Health (Bethesda, MD, USA), developed by W. S:
Rasband (available at http://imagej.nih.gov/ij/, 1997–2016).
3. Bruker’s ParaVision 6® software suit (Bruker Biospin GmbH,
Ettlingen, Germany).
4. To build ADC maps (Fig. 2), DWI images acquired with b ¼ 0
were divided by DWI images acquired with b ¼ 800 mm2 s.
Then, the natural logarithm of the resulting image was calcu-
lated and divided by 0.8 (b ¼ 800 divided by 103). The result-
ing maps provide a pixel-by-pixel representation of 103 ADC
(mm2 s1).
3 Methods
3.1 Stroke Model: A detailed description of surgical procedures and different models
Occlusion of middle cerebral artery occlusion can be found elsewhere
of the Medial Cerebral [21, 22].
Artery (MCAo)
MRI in Stroke 383
Fig. 2 Diffusion-weighted images (DWI, top row) and Apparent Diffusion Coefficient maps (ADC, bottom row) at
the same planes and acquisition times of the rat brain shown in Fig. 1. During the occlusion (hyper-acute
phase) areas suffering cytotoxic edema are visible as hyper-intensities in DWI, or as areas of decreased ADC
value (darker blue color). In models of transient occlusion, the contrast in DWI and ADC vanishes briefly after
reperfusion of the MCA (acute phase), showing the false impression that the lesion is disappearing. At
subacute to chronic stages (t > 6–8 h) the lesion becomes visible again, but with opposite contrast (hypo-
intense signal in DWI and elevated ADC values in ADC maps). DWI/ADC imaging is particularly useful during
the occlusion, but may lead to false negatives during periods of reperfusion
1. Male Sprague-Dawley rats (250–300 g) anesthetized with iso-

flurane (3%) on a 70:30 N2:O2 carrier gas current of
1–1.5 L/min (alternatively, medical air). Weights of animals
should be within 50 g to ensure consistent lesion volume. The
use of male rats is generally justified in order to avoid con-
founding effects due to the hormonal cycles of females.
2. Prepare the surgical area in aseptic conditions and place the
animal in supine position over a heating mat. It is recom-
mended that the mat is connected to a heating device
connected to a PT100 thermal probe that will be introduced
in the rectum of the rat, to automatically control the tempera-
ture to 37 0.2 C. The control of temperature is essential for
reproducibility of lesion sizes.
3. Place an empty 5 mL syringe or similar under the neck of the
animal, this will slightly tilt down the head of the animal,
facilitating the exposure of arteries during surgery on the
neck. Gently tape the four extended limbs of the animal to
the heating mat to facilitate surgical procedures.
4. Use betadine® to impregnate the neck of the animal, pull the
skin at the midline of the ventral side of the neck with forceps,
and practice a 3–4 cm linear cut (vertical, in rostral-caudal
direction) on the skin with scissors, exposing the musculature
of the neck of the animal. At this point it is useful to place a
human eye speculum in the incision to keep the skin retracted

and keep an open and clean space for surgery.
5. Under a surgery microscope the common carotid artery
(CCA), internal carotid artery (ICA), and external carotid
artery (ECA) of one side (let’s say the left side from the ventral
view, but preferences may vary from left- or right-handed
surgeons) are dissected from connective tissue, with special
care to avoid excessive manipulation or lesion of the adjacent
vagus nerve, which generally yields in severe surgical complica-
tions, including the sudden death of the animal.
6. A clamp or a nonpermanent ligature is practiced in the left
CCA, before its bifurcation, the external carotid and pterygo-
palatine arteries are separated and coagulated with an ophthal-
mic cautery (alternatively tied with 6–0 silk sutures). A
nonpermanent suture is placed in the ECA, close to the origin
of the ICA, and the ECA is cut leaving a stump of several mm of
length, without blood flow.
7. A monofilament with engrossed end (size will depend on the
weight of the animal. A nylon filament with a rounded silicone
tip may do the job, but we strongly recommend commercial
suture L34 PK10 from Doccol—see materials section—to
increase success rate and reproducibility) is introduced in the
ECA stump and advanced to the nonpermanent tie. Then, in a
quick succession of movements (practice makes perfection!)
the suture is loosen, the monofilament is advanced toward
the origin of the ICA, and the suture knot is tighten again
compressing the walls of the ECA artery stump against the
thread of the monofilament. This will stop the bleeding caused
during the temporal opening of the ECA, to introduce the
filament. During this key step, bleeding may be considerable
(depending on the skills and experience of the surgeon) and
visibility may be seriously hampered. As usual, practice makes
perfection.
8. Finally the monofilament is advanced through the ICA
(approximately 18–20 mm) to occlude the origin of the MCA.
9. In permanent models, the knot of the ECA is strongly tighten
to keep in place the monofilament, the clamp or knot of the
CCA is removed (some people prefer to leave it permanently in
place) and the incision of the neck is closed by two or three
stitches. Then the animal is left for recovery in its cage, with
periodic observation to ensure absence of complications.
10. For transient models, the thread is left in place for 45–90 min
(typically 60 min), when the monofilament is cautiously
retrieved (the knot at the stump of the ECA must be transiently
loosen to allow the complete withdrawal of the filament, which
is a critical step, since bleeding may be considerable at this
MRI in Stroke 385
step). The knot at the stump of the ECA is strongly tighten, the
clamp of the ACA is retrieved, two or three stitches are used to
close the neck of the animal, and the animal is allowed to
recover in its cage.
11. A laser Doppler flowmeter may be used in the head of the
animal to assess the drop of blood flow during the MCA0
procedure (see [22, 23] for details).
12. Typical complications during surgery include excessive interac-
tion with the vagus nerve, internal bleeding by rupture of a
vessel during the introduction, but most likely during retriev-
ing of the thread, excessive bleeding during the transient loos-
ening of the ECA knot (to introduce/retrieve the threat from
the stump) and intensive increase of intracranial pressure due
to excessive brain edema (consequent to the occlusion of the
MCA). Most of them will result on the death of the animal
either during surgery or in the first hours after it.
13. The use of a pain killer such as buprenorphine (twice a day
during 48–72 h, 0.05–0.1 mg/kg) is recommended to treat
pain from surgical procedures.
14. Preparation of a food paste with pellet powder mixed with
water, and hydratation of the animal via i.p. or s.c. injections
of saline may be necessary to facilitate the feeding of the animal
during the first 78 h, since bad general condition is expected
for the animal (usually due to vasogenic edema in the brain
with displacement of the midline), which starts to resolve
48–72 h after occlusion.
3.2 Magnetic 1. Preparations. Anesthetize the animal, place it in the bed or

Resonance Imaging cradle of the MRI system, and attach the necessary monitoring
probes. Place the MRI coils in position and displace the bed to
locate the center of the head of the animal as close to the
isocenter of the magnet as possible.
2. Setting up the scanner. Tune and match the coils, perform
proper shimming, set the reference frequency, and calibrate
the pulses of the system according to manufacturer instruc-
tions, to achieve maximal signal-to-noise ratio.
3. Pilot scan. Acquire a pilot (scout) set of images to ensure the
proper positioning of the brain of the animal at the isocenter of
the scanner and to serve as reference for the positioning of
imaging planes in successive images. FLASH sequence (Fast
low-angle shot) with three orthogonal planes (axial, coro-
nal and sagittal) centered at (0,0) from the isocenter, Field of
view of FOV: 50 50 mm, Acquisition matrix: 256 256
points, echo time: TE ¼ 3.1 ms, repetition time: TR ¼ 100 ms,
5 dummy scans, averages ¼ 1, bandwidth: 78 KHz
(305.1758 Hz/pixel), gauss 30 degree excitation pulse of
1.11 ms length, fat suppression with a Gaussian pulse of

2.609 ms (90 degrees, 1050 Hz), total scanning time: 12.8 s.
4. T2 scan. RARE (rapid acquisition with relaxation enhance-
ment) sequence with the following parameters: RARE fac-
tor ¼ 4, TR ¼ 4500 ms, TE ¼ 20 ms (Effective
TE ¼ 40 ms), FOV ¼ 25.6 25.6 mm, Matrix ¼ 256 256
points, 14 slices of 1 mm thickness acquired interleaved,
averages ¼ 2, 1 dummy scan, bandwidth: 39 KHz
(152.5879 Hz/pixel), excitation sinc pulse of 2.1 ms
(90 degrees, 1050 Hz), fat suppression with a Gaussian pulse
of 2.61 ms (90 degrees), total scanning time: 9 min 36 s.
5. SWI scan. Flow compensated FLASH sequence (Fast low-angle
shot) with the following parameters: TR ¼ 350 ms, TE ¼ 18 ms,
FOV ¼ 25.6 25.6 mm, Matrix ¼ 256 256 points, 5 slices
of 1 mm thickness acquired interleaved, averages ¼ 2, no
dummy scan, bandwidth: 30 KHz (116.2574 Hz/pixel), exci-
tation sinc pulse of 2.67 ms (40 degrees), fat suppression with a
Gaussian pulse of 4.0 ms (90 degrees, 1050 Hz), total scanning
time: 2 min 59 s.
6. DWI scan. Bruker’s DtiStandard pulse sequence (spin echo
sequence with diffusion gradients) with the following para-
meters: TR ¼ 2500 ms, TE ¼ 22 ms, FOV ¼ 25.6 25.6 mm,
Matrix ¼ 128 96, zero-filled to 128 128 points, 5 slices of
1 mm thickness acquired interleaved, diffusion δ ¼ 4.5 ms,
diffusion Δ ¼ 10.6 ms, b values ¼ 28 (A0) and 800, diffusion
gradient amplitude ¼ 58.47%, diffusion direction (0, 0, 1),
averages ¼ 1, no dummy scan, bandwidth: 30 KHz
(241.1265 Hz/pixel), excitation sinc pulse of 2.318 ms
(90 degrees), fat suppression with a Gaussian pulse of
2.609 ms (90 degrees, 1050 Hz), total scanning time: 8 min.
7. TOF-Angiography. TOF 3D FLASH pulse sequence with the
following parameters: TR ¼ 20 ms, TE ¼ 2.77 ms,
FOV ¼ 32 32 12 mm, Matrix ¼ 256 192 63, zero-
filled to 256 256 96 points, excitation gauss pulse of 1 ms
(30 degrees), bandwidth: 90 KHz (348.7723 Hz/pixel),
averages ¼ 1, a single FOV saturation band located axial to
the scanned block, slab of 15 mm thickness located side to side
at the frontal (rostral) side of the acquisition matrix (sinc10h
pulse of 1.2 ms, 90 degrees and 5000 Hz), total scanning time:
4 min 2 s.
Not all image modalities are required at all imaging points (as it
can be deduced from the presented figures and their explanation at
the figure caption. Therefore, scanning times for the whole proto-
col are variable. In general we recommend to follow the imaging
scans summarized in Table 1, at different stages of the progression
for the disease.
MRI in Stroke 387
4 Notes
1. DCE-MRA methods require the use of an exogenous contrast

agent, which is a disadvantage versus truly noninvasive ASL
techniques, however the shorter scanning times and larger
fields of view (FOV) achievable in DCE-MRA may result
advantageous for certain situations. Besides more time con-
suming, the tuning process required in ASL may be challeng-
ing, since spin labeling vs. imaging timing and positioning of
imaging planes have to be carefully matched, taking into
account the speed of blood traveling through the vessels. Sub-
optimal imaging is common under pathological conditions,
such as stroke, where blood flow may be seriously slowed
down in ischemic regions (Figs. 3 and 4). An improperly
tuned angiographic sequence may hide a vessel with reduced
blood flow as a region where there is no blood supply at all [6].
2. DWI is one of the most useful and most used imaging techni-
ques during the acute phase of stroke, both in preclinical and
clinical practice, since this is one of the few sensitive techniques
that allow the delineation of the ischemic territory at this stage
of the disease. It is important to mention that if a successful
reperfusion can be achieved (for example, in animal models of
transient occlusion of the medial cerebral artery) [5], diffusion
coefficient values can be rapidly normalized showing apparently
normal tissue in DWI. This fact does not rule out the appear-
ance of delayed tissue damage, patent in MR images hours or
days after stroke onset. One should not confound an apparently
normal MR image of the brain with a normal/healthy brain
tissue, and DWI must be used at a proper timing after the onset
of stroke, and in combination with other imaging techniques.
In other words, DWI is very useful during, and hours after the
Fig. 3 Susceptibility-weighted images (SWI) showing enhanced T2* contrast of the brain. Areas of elevated
oxygen extraction (reduced blood supply) lead to marked susceptibility effects due to high levels of
deoxyhemoglobin. This sequence is very sensible to the presence of small bleedings or to hemorrhages
(not present in the images shown here), as strong hypo-intense signal. Otherwise, T2* contrast follows a
similar pattern as T2 contrast in cerebral ischemia. Nevertheless, during the occlusion of the MCA, the hypo-
perfused areas of the brain show a marked hypo-intense signal (see panel “occlusion” in Fig. 4). Although DWI
is the most powerful image modality for detecting acute cerebral ischemia, SWI may be used as an adjunct to
characterize the affected vascular territory during this phase
Fig. 4 Axial (top row, caudal to head view) and coronal (bottom row, top to bottom view) maximal intensity
projections (MIP) of MR angiographies of the head of a rat submitted to transient MCAo surgery. Arrows signal
the right MCA of the animal (the one occluded) while stars show the branching place of the MCA to the circle or
Willis. During the occlusion the right MCA is not seen, confirming a successful occlusion of this artery. When
the thread is withdrawn from the vessel, the angiography confirms a successful reperfusion of the MCA
occlusion of a blood vessel, but may result futile in the

in-between period. In situations of a permanent occlusion of
the blood vessels, without reperfusion, this phenomenon is not
observed, as DWI uninterruptedly reflects ischemic regions.
However it is also important to mention that the appearance
of DWI images or ADC maps changes with time after stroke as
cytotoxic edema (present at very early stages of ischemia) yields
in hampered diffusion, reflected as hyper-intense images on
DWI (hypo-intense in ADC), while tissue pannecrosis (days
after the onset of ischemia) yields in facilitated diffusion,
reflected as hypo-intense images on DWI (hyper-intense in
ADC).
3. Diffusion Tensor Imaging (DTI) is a useful tool to study the
progression of disease affecting white matter, or to characterize
the response of individuals to therapeutic procedures
[23]. However, in our experience, due to the low amount
and reduced size of white matter bundles in the rodent brain,
DTI is not used in routine explorations in preclinical stroke
research with rodents, and it is generally used for the study of
particular issues related with functional recovery and brain
plasticity following stroke.
4. Diffusion-weighted and T2-weighted imaging are the two
main modalities used for the study of ischemic lesions. The
choice of one or the other is based on temporal criteria. During
the occlusion and up to 4–6 h (in the case of permanent
occlusions, see Note 2) the correct choice for imaging is
DWI, since T2 is not sensible to pathological events happening
at that stage. From that point (4–6 h) both DWI and T2
MRI in Stroke 389
imaging can be indistinctively used to depict the extension of

the ischemic lesion. However, T2 is the most commonly used
approach, since it is easy to perform and leads to better quality
images in shorter imaging times than DWI-MRI (unless EPI
imaging is used for DWI, in which case imaging times are
comparable but images are more prone to contain distortions
and artifacts).
5. T2-weighted images can be easily acquired using a conven-
tional multi-slice multi-echo sequence (MSME) or Fast Spin
Echo (FSE) / Rapid Acquisition with Relaxation Enhancement
(RARE) sequences. T2*-weighted images acquired by conven-
tional gradient recalled echo sequences (GRE) or Fast Low
Angle Shot imaging (FLASH) sequences are equivalent for
the purpose of delineating the ischemic lesion. However gradi-
ent echo approaches are not the usual choice for ischemic
stroke imaging, since these sequences have lower signal-to-
noise ratios than SE, they require the use of short echo times
at high magnetic fields (sometimes challenging for conven-
tional gradient sets), and images are more prone to susceptibil-
ity artifacts and B1 inhomogeneities, respect to spin echo
approaches. Nevertheless gradient echo sequences have an
important role is stroke research since they are very sensitive
to the detection of bleeding. Indeed GRE is used for the study
of hemorrhagic stroke, and for the detection of hemorrhages
secondary to ischemic stroke events, as hemorrhagic transfor-
mation is one of the common complications observed in ische-
mic stroke patients.
6. FLAIR presents two main advantages over conventional Spin-
echo sequences. On the one side, flow artifacts arising from
CSF are suppressed, and lesion sizes are more accurately calcu-
lated since sometimes is difficult to differentiate hyper-intense
ischemic areas of tissue from the adjacent CSF related hyper-
intense regions, yielding in overestimated infarct volumes.
Thus, FLAIR is routinely used in clinical practice, but rarely
observed in experimental stroke studies in rodents. If one
searches for the terms Stroke AND flair AND human, in a
publications database such as Pubmed, 404 matches are
found (392 when using patient instead of human). When one
searches for stroke AND flair AND Rat, only five matches are
found (0 when changing rat by mouse or by mice, and 6 when
replacing rat by rodent). The explanation of this fact is not
really clear to us. It is true that the inversion-recovery prepara-
tion of the signal has to be finely tuned to effectively suppress
CSF signal, and that IR preparation requires longer acquisition
times, but that time may be recovered later, during the image
analysis process, since CSF signal will not interfere with the
quantification of ischemic volumes. On the other hand, cystic
components of lesions (specially present in the chronic phase of

stroke) may be suppressed by FLAIR sequence, as they present
characteristics of CSF regions. But in fact, we can speculate that
when FLAIR was introduced for the study of stroke, T2w SE
imaging was established for a long time as the standard preclin-
ical imaging procedure in many laboratories around the world.
Since then, we may have just “kept on going” the way we were
used to. But we always wondered if the generalized use of
FLAIR at preclinical level should some day be considered a
must, facilitating in this way the traslationality of preclinical
studies to clinical research.
Acknowledgements
We deeply acknowledge Ms. Marta Beraza, from the Molecular

Imaging Unit of CIC biomaGUNE for the performance of the
MCAo surgery of the rats scanned in this work. We acknowledge
the Spanish Ministry of Economy and Competence (Project
SAF2014-53413-R), The Basque Government (PI_2015_1_53),
Ikerbasque (the Basque Research Foundation), and the European
Union (FEDER funds) for financial support.
References
1. Nichols M, Townsend N, Scarborough P, Ray- de Vaccaro KC, Couser W, Cowie BC, Criqui
ner M (2013) Cardiovascular disease in Eur- MH, Cross M, Dabhadkar KC, Dahiya M,
ope: epidemiological update. Eur Heart J 34 Dahodwala N, Damsere-Derry J, Danaei G,
(39):3028–3034. https://doi.org/10.1093/ Davis A, De Leo D, Degenhardt L,
eurheartj/eht356 Dellavalle R, Delossantos A, Denenberg J,
2. Murray CJ, Vos T, Lozano R, Naghavi M, Flax- Derrett S, Des Jarlais DC, Dharmaratne SD,
man AD, Michaud C, Ezzati M, Shibuya K, Dherani M, Diaz-Torne C, Dolk H, Dorsey
Salomon JA, Abdalla S, Aboyans V, ER, Driscoll T, Duber H, Ebel B, Edmond K,
Abraham J, Ackerman I, Aggarwal R, Ahn SY, Elbaz A, Ali SE, Erskine H, Erwin PJ,
Ali MK, Alvarado M, Anderson HR, Anderson Espindola P, Ewoigbokhan SE, Farzadfar F,
LM, Andrews KG, Atkinson C, Baddour LM, Feigin V, Felson DT, Ferrari A, Ferri CP,
Bahalim AN, Barker-Collo S, Barrero LH, Bar- Fevre EM, Finucane MM, Flaxman S,
tels DH, Basanez MG, Baxter A, Bell ML, Flood L, Foreman K, Forouzanfar MH,
Benjamin EJ, Bennett D, Bernabe E, Fowkes FG, Fransen M, Freeman MK, Gabbe
Bhalla K, Bhandari B, Bikbov B, Bin BJ, Gabriel SE, Gakidou E, Ganatra HA,
Abdulhak A, Birbeck G, Black JA, Garcia B, Gaspari F, Gillum RF, Gmel G,
Blencowe H, Blore JD, Blyth F, Bolliger I, Gonzalez-Medina D, Gosselin R, Grainger R,
Bonaventure A, Boufous S, Bourne R, Grant B, Groeger J, Guillemin F, Gunnell D,
Boussinesq M, Braithwaite T, Brayne C, Gupta R, Haagsma J, Hagan H, Halasa YA,
Bridgett L, Brooker S, Brooks P, Brugha TS, Hall W, Haring D, Haro JM, Harrison JE,
Bryan-Hancock C, Bucello C, Buchbinder R, Havmoeller R, Hay RJ, Higashi H, Hill C,
Buckle G, Budke CM, Burch M, Burney P, Hoen B, Hoffman H, Hotez PJ, Hoy D,
Burstein R, Calabria B, Campbell B, Canter Huang JJ, Ibeanusi SE, Jacobsen KH, James
CE, Carabin H, Carapetis J, Carmona L, SL, Jarvis D, Jasrasaria R, Jayaraman S,
Cella C, Charlson F, Chen H, Cheng AT, Johns N, Jonas JB, Karthikeyan G,
Chou D, Chugh SS, Coffeng LE, Colan SD, Kassebaum N, Kawakami N, Keren A, Khoo
Colquhoun S, Colson KE, Condon J, Connor JP, King CH, Knowlton LM, Kobusingye O,
MD, Cooper LT, Corriere M, Cortinovis M, Koranteng A, Krishnamurthi R, Laden F,
MRI in Stroke 391
Lalloo R, Laslett LL, Lathlean T, Leasher JL, 3. Dyken ML (2010) What lessons have we
Lee YY, Leigh J, Levinson D, Lim SS, Limb E, learned in the past 40 years? Stroke 41
Lin JK, Lipnick M, Lipshultz SE, Liu W, (6):1073–1075. https://doi.org/10.1161/
Loane M, Ohno SL, Lyons R, Mabweijano J, STROKEAHA.110.584011
MacIntyre MF, Malekzadeh R, Mallinger L, 4. Hachinski V (2007) The 2005 Thomas Willis
Manivannan S, Marcenes W, March L, Margo- Lecture: stroke and vascular cognitive
lis DJ, Marks GB, Marks R, Matsumori A, impairment: a transdisciplinary, translational
Matzopoulos R, Mayosi BM, McAnulty JH, and transactional approach. Stroke 38
McDermott MM, McGill N, McGrath J, (4):1396. https://doi.org/10.1161/01.STR.
Medina-Mora ME, Meltzer M, Mensah GA, 0000260101.08944.e9
Merriman TR, Meyer AC, Miglioli V, 5. Hossmann KA (2008) Cerebral ischemia:
Miller M, Miller TR, Mitchell PB, Mock C, models, methods and outcomes. Neurophar-
Mocumbi AO, Moffitt TE, Mokdad AA, macology 55(3):257–270. https://doi.org/
Monasta L, Montico M, Moradi-Lakeh M, 10.1016/j.neuropharm.2007.12.004
Moran A, Morawska L, Mori R, Murdoch
ME, Mwaniki MK, Naidoo K, Nair MN, 6. Runck F, Steiner RP, Bautz WA, Lell MM
Naldi L, Narayan KM, Nelson PK, Nelson (2008) MR imaging: influence of imaging
RG, Nevitt MC, Newton CR, Nolte S, technique and postprocessing on measurement
Norman P, Norman R, O’Donnell M, of internal carotid artery stenosis. AJNR Am J
O’Hanlon S, Olives C, Omer SB, Ortblad K, Neuroradiol 29(9):1736–1742. https://doi.
Osborne R, Ozgediz D, Page A, Pahari B, Pan- org/10.3174/ajnr.A1179
dian JD, Rivero AP, Patten SB, Pearce N, 7. Ewing JR, Knight RA, Nagaraja TN, Yee JS,
Padilla RP, Perez-Ruiz F, Perico N, Nagesh V, Whitton PA, Li L, Fenstermacher
Pesudovs K, Phillips D, Phillips MR, Pierce K, JD (2003) Patlak plots of Gd-DTPA MRI data
Pion S, Polanczyk GV, Polinder S, Pope CA yield blood-brain transfer constants concor-
3rd, Popova S, Porrini E, Pourmalek F, dant with those of 14C-sucrose in areas of
Prince M, Pullan RL, Ramaiah KD, blood-brain opening. Magn Reson Med 50
Ranganathan D, Razavi H, Regan M, Rehm (2):283–292. https://doi.org/10.1002/
JT, Rein DB, Remuzzi G, Richardson K, Rivara mrm.10524
FP, Roberts T, Robinson C, De Leon FR, 8. Abo-Ramadan U, Durukan A, Pitkonen M,
Ronfani L, Room R, Rosenfeld LC, Marinkovic I, Tatlisumak E, Pedrono E,
Rushton L, Sacco RL, Saha S, Sampson U, Soinne L, Strbian D, Tatlisumak T (2009)
Sanchez-Riera L, Sanman E, Schwebel DC, Post-ischemic leakiness of the blood-brain bar-
Scott JG, Segui-Gomez M, Shahraz S, Shepard rier: a quantitative and systematic assessment
DS, Shin H, Shivakoti R, Singh D, Singh GM, by Patlak plots. Exp Neurol 219(1):328–333.
Singh JA, Singleton J, Sleet DA, Sliwa K, https://doi.org/10.1016/j.expneurol.2009.
Smith E, Smith JL, Stapelberg NJ, Steer A, 06.002
Steiner T, Stolk WA, Stovner LJ, Sudfeld C, 9. Barbier EL, Lamalle L, Decorps M (2001)
Syed S, Tamburlini G, Tavakkoli M, Taylor Methodology of brain perfusion imaging. J
HR, Taylor JA, Taylor WJ, Thomas B, Thom- Magn Reson Imaging 13(4):496–520
son WM, Thurston GD, Tleyjeh IM, 10. Campos F, Sobrino T, Ramos-Cabrer P,
Tonelli M, Towbin JA, Truelsen T, Tsilimbaris Argibay B, Agulla J, Perez-Mato M,
MK, Ubeda C, Undurraga EA, van der Werf Rodriguez-Gonzalez R, Brea D, Castillo J
MJ, van Os J, Vavilala MS, (2011) Neuroprotection by glutamate oxaloac-
Venketasubramanian N, Wang M, Wang W, etate transaminase in ischemic stroke: an exper-
Watt K, Weatherall DJ, Weinstock MA, imental study. J Cereb Blood Flow Metab 31
Weintraub R, Weisskopf MG, Weissman MM, (6):1378–1386. https://doi.org/10.1038/
White RA, Whiteford H, Wiebe N, Wiersma jcbfm.2011.3
ST, Wilkinson JD, Williams HC, Williams SR,
Witt E, Wolfe F, Woolf AD, Wulf S, Yeh PH, 11. Ramos-Cabrer P, Campos F, Sobrino T, Cas-
Zaidi AK, Zheng ZJ, Zonies D, Lopez AD, tillo J (2011) Targeting the ischemic penum-
AlMazroa MA, Memish ZA (2012) Disability- bra. Stroke 42(1 Suppl):S7–11. https://doi.
adjusted life years (DALYs) for 291 diseases org/10.1161/STROKEAHA.110.596684
and injuries in 21 regions, 1990–2010: a sys- 12. Weber R, Ramos-Cabrer P, Hoehn M (2006)
tematic analysis for the Global Burden of Dis- Present status of magnetic resonance imaging
ease Study 2010. Lancet 380 and spectroscopy in animal stroke models. J
(9859):2197–2223. https://doi.org/10. Cereb Blood Flow Metab 26(5):591–604.
1016/S0140-6736(12)61689-4 https://doi.org/10.1038/sj.jcbfm.9600241
13. Cai K, Haris M, Singh A, Kogan F, Greenberg Use of fluid attenuated inversion recovery
JH, Hariharan H, Detre JA, Reddy R (2012) (FLAIR) pulse sequences in MRI of the brain.
Magnetic resonance imaging of glutamate. Nat J Comput Assist Tomogr 16(6):841–844
Med 18(2):302–306. https://doi.org/10. 19. Wegener S, Weber R, Ramos-Cabrer P,
1038/nm.2615 Uhlenkueken U, Sprenger C, Wiedermann D,
14. Stejskal EO, Tanner JE (1965) Spin diffusion Villringer A, Hoehn M (2006) Temporal pro-
measurements: spin echoes in the presence of a file of T2-weighted MRI distinguishes between
time dependent field gradient. J Chem Phys pannecrosis and selective neuronal death after
42:288–292 transient focal cerebral ischemia in the rat. J
15. Heidenreich JO, Hsu D, Wang G, Jesberger Cereb Blood Flow Metab 26(1):38–47.
JA, Tarr RW, Zaidat OO, Sunshine JL (2008) https://doi.org/10.1038/sj.jcbfm.9600166
Magnetic resonance imaging results can affect 20. Van der Linden A, Van Camp N, Ramos-
therapy decisions in hyperacute stroke care. Cabrer P, Hoehn M (2007) Current status of
Acta Radiol 49(5):550–557. https://doi.org/ functional MRI on small animals: application
10.1080/02841850801958320 to physiology, pathophysiology, and cognition.
16. Chen F, Ni YC (2012) Magnetic resonance NMR Biomed 20(5):522–545. https://doi.
diffusion-perfusion mismatch in acute ischemic org/10.1002/nbm.1131
stroke: an update. World J Radiol 4(3):63–74. 21. Dirnagl U (2016) Rodent models of stroke. In:
https://doi.org/10.4329/wjr.v4.i3.63 Neuromethods, vol 47. Humana Press,
17. Tournier JD, Mori S, Leemans A (2011) Dif- New York
fusion tensor imaging and beyond. Magn 22. Wang-Fischer Y (2008) Manual of stroke mod-
Reson Med 65(6):1532–1556. https://doi. els in rats. CRC Press, Boca Raton, FL
org/10.1002/mrm.22924 23. Le Bihan D, Mangin JF, Poupon C, Clark CA,
18. Hajnal JV, Bryant DJ, Kasuboski L, Pattany Pappata S, Molko N, Chabriat H (2001) Dif-
PM, De Coene B, Lewis PD, Pennock JM, fusion tensor imaging: concepts and applica-
Oatridge A, Young IR, Bydder GM (1992) tions. J Magn Reson Imaging 13(4):534–546
Part VI
Other Applications
Chapter 23
Assessment of Blood Brain Barrier Leakage

with Gadolinium-Enhanced MRI
Min-Chi Ku, Sonia Waiczies, Thoralf Niendorf, and Andreas Pohlmann
Abstract
The integrity of the blood–brain barrier (BBB) can be noninvasively monitored by magnetic resonance
imaging (MRI). Conventional MR contrast agents (CAs) containing gadolinium are used in association
with MRI in routine clinical practice to detect and quantify BBB leakage. Under normal circumstances CAs
do not cross the intact BBB. However due to their small size they extravasate from the blood into the brain
tissue even when the BBB is partially compromised. Here we describe an MR method based on
T1-weighted images taken prior to and after CA injection. This MR method is useful for investigating
BBB permeability in in vivo mouse models and can be easily applied in a number of experimental disease
conditions including neuroinflammation disorders, or to assess (un)wanted drug effects.
Key words Magnetic resonance imaging (MRI), Mouse, Blood brain barrier (BBB), contrast agent
(CA)
1 Introduction
The blood brain barrier (BBB) is a multicellular vascular structural

interface that selectively restricts the blood-to-brain traffic of com-
pounds and cells, allowing a tight regulation of the central nervous
system (CNS) microenvironment [1]. Such an intrinsic CNS
homeostasis is fundamental for maintaining proper neuronal net-
work function and communication.
The BBB is a key player in many CNS pathologies, particularly
in disorders with a neuroinflammatory component such as cerebral
stroke [2] and multiple sclerosis [3], but also in traumatic brain
injury [4] and in disorders with a primary neurodegenerative com-
ponent such as Alzheimer’s disease [5]. Transporter dysfunction
impairs glucose transport [6] and has been related to psychiatric
disorders [7], and together with an altered hormone transport in
diabetes mellitus, has been associated with increased BBB perme-
ability [8]. These pathologies may result in a compromised BBB
and altered homeostasis. Such “BBB opening” event is an
395
important indicator of CNS disease status. Therefore, early detec-

tion of BBB permeability would be a precursor to diagnosing and
monitoring CNS abnormalities.
While loss in BBB integrity is part of the pathogenesis of a
number of neurological disorders, some drug therapies that are
meant to act deep within the brain parenchyma necessitate a tran-
sient disruption of the BBB in order to penetrate from the circula-
tion into the brain [9, 10]. Treatment of brain tumors remains a
great challenge because of the presence of BBB; the partially intact
blood-tumor barrier limits the delivery of drug therapies as well as
immunotherapies from the blood circulation into brain paren-
chyma. An accurate and noninvasive quantification of the BBB
permeability can provide a useful guide for CNS drug delivery, as
well as for measuring the response of the BBB to therapy, allowing
for rational drug design [1, 11].
Recent preclinical studies have highlighted the potential of sev-
eral approaches for detecting and monitoring BBB leakage. Methods
based on MRI are currently the gold standard, and are the most
commonly used techniques that allow real-time and noninvasive
imaging. Measuring BBB permeability by MRI holds great potential
for translation into the clinic and may also be utilized for monitoring
of treatment response [12, 13]. The most MRI methods involve an
intravenous injection of a gadolinium-containing CA. The leakage of
CA through a disrupted BBB has effects on the water proton signal in
the extravascular space [14] (blood-to-brain transfer). Determination
of BBB permeability by MRI can be carried out qualitatively or
quantitatively. A qualitative assessment involves a comparison of
T1-weighted images pre- and post- the CA injection. A quantitative
assessment involves either calculating T1 changes from pre- and post-
CA T1-maps, or using dynamic contrast-enhanced MRI (DCE-MRI)
[15], which differentiates between changes in vascular permeability
and changes in the extravascular extracellular space (EES).
DCE-MRI, however, requires pharmacokinetic modeling [16, 17]
and knowledge of the transient CA concentration in the arteries
(so-called arterial input function). This is, for several reasons, rather
challenging to achieve in the small mouse brain.
Here we describe the method of quantitative mapping of T1
using a steady-state like regime of CA administration. The reduc-
tion of T1 values in the brain after CA administration is used as an
indicator for BBB leakage.
2 Materials
2.1 Animals Mice. All mice are housed under standard conditions according to
the animal regulations in our institute. We used female mice
(C57BL/6 J) with a body weight of 16–20 g. Other strains, ages,
and genders might require some changes in the experimental
design (see Note 1).
MRI assessment of BBB 397
2.2 Preparation 1. Anesthesia: Mice should be anesthetized by an initial inhalation

of Animals for MRI narcosis using 2% isoflurane for about 3 min with air and
oxygen, then maintaining in 0.5–1.5% isofluorane during MR
scans.
2. Vital sign control: Core body temperature should be main-
tained at 37 C with a small animal electrical warming pad
while in preparation. Respiration rate and temperature can be
monitored using a remote monitoring system (Model 1025,
SA Instruments Inc.).
3. Catheterization of the vein: CA has to be injected employing
polythene tubing catheter (Ref 800/100/100; inner diameter
0.28 mm, Smiths Medical™ Portex™, Kent, UK), 30 G nee-
dles, 1 mL syringes, saline (0.9% NaCl solution), heparinized
saline (100 IU/mL), and instant glue.
4. When injecting contrast agent with a step-down manner we
strongly recommend using a syringe pump (Model PHD
22/2000, Harvard Apparatus, MA, USA) (see Note 2).
5. Temperature measurement: Interferometric measurement sys-
tem (ACS-P4-N-62SC, Opsens, Quebec City, Canada),
including a fiber-optical temperature probe (OTP-M, Accu-
Sens, Opsens) (see Note 3).
2.3 Magnetic This method requires access to an high field MRI system in addi-
Resonance Imaging tion suitable accessories for the MR acquisition such as radio fre-
(MRI) quency (RF) antennas; equipment for animal positioning,
anesthetizing, warming, and monitoring of physiological para-
meters of the animals to be imaged; and trained personnel for
operating the MRI system.
1. MRI system: A dedicated small animal MR system with a
magnetic field strength of 4.7 Tesla or higher is recommended.
Here we describe the use of a 9.4 Tesla 20 cm bore system
(Biospec 94/20, Bruker Biospin, Ettlingen, Germany)
equipped with a gradient system integrated with shim set
(B-GA12S2, Bruker Biospin, Ettlingen, Germany; gradient
amplitude 440 mT/m, max. Slew rate 3440 T/m/s).
2. Radio frequency (RF) coils: Use RF volume coils (reception
house made birdcage) suitable for mouse brain imaging, such
as a 16 mm volume resonator (custom-made), or surface RF
coils, such as a 2 2 element mouse brain RF coil array linear
reception-only surface coil (e.g., model: T11071 V3, Bruker
Biospin, Ettlingen, Germany) combined with a volume reso-
nator for transmit (model: T10325_V3, Bruker Biospin,
Ettlingen, Germany); for high resolved imaging cryogenically
cooled RF coil (CryoProbe, Bruker) is recommended
[18, 19]. Here we used the first mentioned RF coil setup, as
it also provides high spatial resolution.
3. Gases: O2 and compressed air, as well as a gas-mixing system

(FMI Föhr Medical Instruments GmbH, Seeheim-Ober Beer-
bach, Germany).
4. Device for warming of animal while scanning: Use a circulating
warm-water-based heating system, consisting of a plastic cover
or rubber mat with integrated tubing connected to a conven-
tional warm water bath (SC100-A10, ThermoFisher, Dreieich,
Germany). For alternative coil setups, water pipes may be built
into the animal holder.
5. Monitoring of physiological parameters: For monitoring of
respiration and core body temperature throughout the entire
MR experiment, use a small animal monitoring system (Model
1025, Small Animal Instruments, Inc., Stoney Brook, NY,
USA), including a rectal temperature probe and pneumatic
pillow. Maintain body temperature by water heater/circulator
set up to 42–45 C.
6. Data analysis: Quantitative analysis of the MR data requires a
personal computer and MATLAB software (R13 or higher;
The Mathworks, Natick, MA, USA), ImageJ (Rasband, W.S.,
ImageJ, U.S. National Institutes of Health, Bethesda, Mary-
land, USA, http://imagej.nih.gov/ij/, 1997–2016), or a simi-
lar software development environment. Analysis steps
described in the Methods section can be performed manually
by using the functions provided by the software development
environment. Most of these steps benefit from (semi-)
automation by creating software programs/macros—these
steps are indicated by the computer symbol (:).
3 Methods
3.1 Preparation 1. Start the ParaVision software (here we used ParaVision 5.0)
of MRI and—only once prior to the first experiment—create and store
the following MR protocols.
(a) Protocol_TriPilot (pilot scan for checking the mouse head
position): conventional FLASH pilot with three slices in
each direction (axial, coronal, sagittal).
(b) Protocol_Pilot_Axial (axial pilot scan): FLASH sequence,
repetition time (TR) ¼ 560 ms, effective echo time
(TE) ¼ 3 ms, the geometry define as an axial field of
view (FOV) ¼ 25 25 mm2, matrix size ¼ 256 256,
ten slices with a thickness of 0.7 mm.
(c) Protocol_T1 map (for T1-mapping): RARE sequence,
TR ¼ 9000, 2400, 1480, 940, 650, 418 ms (set up six
separate scans or use the RAREVTR protocol, which
automatically sets six different TRs for the 6 consecutive
scans). Echo times ¼ 8.1, rare factor ¼ 4, define as
geometry a coronal oblique image slice with a

FOV ¼ (18 18) mm2, matrix size ¼ 128 128, and a
slice thickness of 1.0 mm.
(d) Protocol_IR-GRE: Inversion recovery gradient echo
sequence (IR-GRE), use Bruker “MDEFT” sequence in
mode “inversion,” TR ¼ 15 ms, TE ¼ 3.89 ms, flip
angle ¼ 20 degree, FOV ¼ (18 18) mm2, matrix
size ¼ 128 128, with 12 slices of 1.0 mm thickness.
Inversion time ¼ 950 ms, segmentation time ¼ 2600 ms,
segmentation duration ¼ 735 ms, number of
segments ¼ 4.
(e) Protocol_T1w-GRE (to quickly determine whether CA
injection succeeded): FLASH sequence, TR ¼ 5.95 ms,
TE ¼ 2.15 ms, flip angle ¼ 20 degree,
FOV ¼ (18 18) mm2, matrix size ¼ 128 128.
2. Switch on the gradient amplifiers of the MR system; this will
also switch on the automatic animal positioning system
AutoPac.
3. Connect the animal holder to the animal positioning system
(AutoPac).
4. Attach the face mask unit (either commercial or custom-made)
to the animal holder and connect it to the inspiratory gas
providing system (luer tubing) (see Note 4).
5. Place the plastic mat (warm-water-based heating system) on
top of the mouse and then connect it to the warm-water
circulation system (water bath).
6. Switch on the water bath. Adjust the temperature to approxi-
mately 42–45 C (see Note 5).
7. Attach the rectal temperature probe and pneumatic pillow to
the small animal monitoring system and place the probes on the
animal bed, approximately at lower abdominal position of the
mouse. Attach all tubes and cables along the length of the
animal bed, fixing with masking (or autoclave marker) tape
(see Note 6).
8. Prepare the CA: Dissolve 50 mg gadodiamide powder in PBS
to make a 50 mM working solution for injection. Fill this
working solution in a 200 cm long polythene tube that was
pre-attached with a 30 G injection needle. This long polythene
tube will be then connected to an injection pump (see Note 7).
3.2 Transport, 1. Switch on the small animal monitoring system.

Fixation, 2. For the purpose of injecting CA while the animal is in the MR
Positioning, Mask scanner, mice need to be catheterized to allow CA injections
and infusions during the MR measurements. Construct a
catheter from a 30 cm polythene tubing catheter (Ref

800/100/100; inner diameter 0.28 mm, Smiths Medical™
Portex™, Kent, UK). Pull a 30 G needle out of its hub and
insert the dull end into the polythene tube, securing it with
instant glue. Attach a 1 mL syringe pre-filled with saline (0.9%
NaCl solution) to the catheter (avoid air bubbles) into the
other end. Dip the catheter needle tip with heparinized saline
(100 IU/mL). Rotate the mouse tail to illuminate the lateral
tail vein, then insert the needle with the bevel side up at the
distal end of the tail. Once the needle is inserted, slowly with-
draw blood to ensure the needle is inside the tail vein. Glue the
catheter onto the tail by a small drop of instant glue. Wait for
3 min and once the glue is dried, carefully transport the mouse
with the tubing for catheter to the MRI scanner. Connect
longer tubing (about 200 cm) to the short tubing to allow an
extension of the tubing outside of the scanner (see Note 8).
3. Carefully transfer the animal with the attached catheter into the
MR scanner room, avoiding any damage to the vein or dis-
placement of the needle.
4. Position the mouse on the MRI animal bed holder and use the
AutoPac laser system to align the mouse brain position with the
center of the RF volume coil.
5. Place a respiratory mask loosely around the muzzle of the
spontaneously breathing mouse. Turn on air supply to a rate
of 250 mL min1.
6. Insert the rectal temperature probe after cleaning with alcohol
and coating with Vaseline™.
7. Cover the animal with the warming plastic mat. Watch the
respiration trance on the monitor of the small animal monitor-
ing system. Press the “Out” button of the AutoPac system to
make sure that the animal bed is in the reference position.
Switch on the Laser position marker and drive the animal bed
until the anatomical region of interest (in this case the brain) is
aligned with the laser position (this ascertains the positioning
of the brain to the center of the RF volume coils). Switch off
the laser (this point is now stored as the point of interest).
8. Double check the entire animal bed to make sure nothing
protrudes.
9. Press the Work Position button of the AutoPac system to drive
the point of interest (point on the animal bed marked by the
Laser) to the isocenter of the magnet.
3.3 MRI Pre-scans 1. Start the ParaVision software and register a new subject and
study. For the first scan select the Protocol_TriPilot.
2. Tune and match the mouse brain RF coils using the Wobble
function.
3. Start the first pilot scan and verify on the acquired images
whether the brain is in the center of the magnet (field of
view) and well positioned within the signal intensity profile. If
necessary correct the animal or/and laser marker position.
Repeat the tuning/matching and redo this pilot scan (see
Note 9).
4. Load the Protocol_Pilot_Axial (axial pilot scan), edit the geom-
etry based on the TriPilot scan to position the brain slices
correctly in the center (Fig. 1) and run the scan.
5. Clone the Protocol_Pilot_Axial, change the slice orientation to
coronal, and run the scan.
6. Clone the Protocol_Pilot_Axial, change the slice orientation to
sagittal, and run the scan.
3.4 Baseline 1. Load the Protocol_T1 map and edit the geometry based on the
(Pre-contrast) Images coronal and sagittal views such that the slice package encloses
the brain (Fig. 2).
2. Clone Protocol_T1 map and run scans with GOP. Result from
pre-CA scans are shown in Fig. 5 (upper row) (see Note 10).
3. Load the Protocol_IR-GRE and copy the geometry from pre-
vious scan. Run the scan and check image geometry and
quality.
4. Load the Protocol_T1w-GRE and position one axial slice such
that it crosses the brain in the center. Run the scan.
Fig. 1 Coronal and sagittal image of a mouse brain from the Protocol_TriPilot
(pilot scan) and show how to place the slice position for Protocol_Pilot_Axial
Fig. 2 Position planning for running Protocol_T1 map based on the coronal and
sagittal pilot scans. Adjust the angle according to the brain orientation (see Note 11)
Table 1
Step-down injection method for injecting the CA over 20 mins
For 1 mmol/kg
20 g Rate (μL/min) Duration (mins) Volume (μL)
50 0.5 25
40 0.5 20
30 1 30
20 1 20
16 1 16
14 1 14
10 2 20
8 3 24
6 5 30
4 5 20
Total 20 219
3.5 Post-contrast 1. There are two ways of CA injection. The most common one is a
Images bolus injection (single fast inject) and the other is an infusion in
a step-down manner. Table 1 shows an example (for a 20 g
mouse) for injecting CA (1 mmol/kg) in a step-down manner
over 20 min via the tail vein using a syringe pump, which can
regulate the injection volume and injection rates. For mouse
with different weights, adjust the CA amount and flow rates
accordingly (see Note 12).
Fig. 3 Axial view images acquired prior to (Pre-CA) and after a successful
injection (Post-CA) with CA. Protocol_T1w-GRE images (right panel; Post-CA)
depict a succeeded injection as shown by the significant signal intensity
changes within the head vessels and muscle tissue but not in the healthy
brain (see Note 14)
Fig. 4 Axial view images acquired prior to (Pre-CA) and after (Post-CA) a failed
injection of CA. Protocol_T1w-GRE images (right panel; Post-CA) depict an
incomplete CA injection, the hyperintensity is only seen in bigger vessels of
the head but not in muscle tissue (see Note 15)
2. Start step-down contrast agent injection with a syringe pump.

3. Start Protocol_T1w-GRE (run with GOP) 5 mins after starting
injection, while infusion is still ongoing.
4. Check the images: A change in the signal intensity in muscle
and vessels is indicative of a successful injection (Fig. 3); if no
change or only minimal changes (Fig. 4) in signal intensity is
observed, the injection failed (see Note 13).
5. End of injection.
6. Clone Protocol_T1w-GRE (and run with GOP) and check the
images. If there is no change in signal intensity the injection
failed (in this case the experiment would have to be
terminated).
Fig. 5 Axial view images acquired prior to (Pre-CA; upper row) and after (Post-CA; lower row) CA injection and
scanned with six different TRs. Protocol_T1map images show the signal intensity changes between the scans
acquired with different TR values. Lower row shows increased signal intensity for each TR value in the Post-CA
images, compared to Pre-CA images
Fig. 6 Axial view images acquired prior to (Pre-CA) and after (Post-CA) injection. Protocol_IR-GRE provides
contrast between white/gray matters without CA (left panel; Pre-CA). This scan is informative about the
integrity of BBB following a successful CA injection, and can be performed prior to a T1 map, which is usually
more time consuming. The presence of CA is shown as increased signal intensity within the head region and
here also in the brain, indicating BBB leakage labeled with yellow arrows (right panel; Post-CA)
7. Clone Protocol_T1 map from previous scans and run scan with
GOP. Results from post-CA scans shown in Fig. 5 (lower row).
8. Clone Protocol_IR-GRE and run with GOP. Scan result shown
in Fig. 6.
3.6 End 1. Carefully remove mouse from the mouse holder.

of Experiment 2. Move the animal into the preparation room and place in a
supine position on a pre-warmed mat.
3. Carefully remove the catheter from mouse tail vein. Apply zinc
oxide cream to help heal any extravasation injuries.
Fig. 7 Representative T1 maps before (Pre-CA) and after (Post-CA) CA injection. T1 maps were created from
images by running Protocol_T1 map and post processed. T1 maps were calculated by fitting a mono-
exponential curve to the signal intensities of the six T1-weighted images using custom-made software
program (: MATLAB). Color coded ΔT1 map of relative change in T1 were calculated as ΔT1 ¼ |T1,postCA
/ T1,preCA| - 1 100. The result revealed that the BBB leakage detected by post-pre CA approach can be
seen in paraventricular space
3.7 Data Analysis 1. Open the reconstructed MR images in MATLAB (:; see Mate-
rials) by loading the binary 2dseq files of the RAREVTR scan
and import TR parameters from the method text file.
2. Perform pixel-wise exponential curve fitting of the equation:
S ðt Þ ¼ A þ B ñexp:ð1 t=T 1 Þ ð1Þ
to the image intensities S(t) versus time t (the 6 repetition

times) (: MATLAB).
3. To visually inspect the T1 maps display the matrix containing
fitted T1 parameter as images using a gray scale in MATLAB (:
MATLAB) with the limits 0–5000 ms. (see Fig. 7) (:
MATLAB)
4. For determining the change of regional T1 due to the injected
CA, calculate the ΔT1 map (in percentage) by dividing the
post-CA map by the pre-CA map, then subtract 1 and multiply
by 100 (examples see Fig. 7). (: MATLAB and :ImageJ) The
ΔT1 map was color coded in :ImageJ (Fig. 7, ΔT1 map).
5. For further quantification place a region-of-interest (ROI)
within ΔT1-maps (see Note 16) and use histogram analysis
(see Note 17).
6. The statistical significance of group differences in the ΔT1
histograms can be further determined using the Kolmogorov-
Smirnov test.
4 Notes
1. When doing a study, make sure the body weight of each mouse
within a study group is similar.
2. A syringe pump is not necessary when injecting contrast agents
as bolus.
3. When using a GaAs crystal-based system, be aware of the offset

caused by the magnetic field (approx. 4.7 C at 9.4 Tesla).
4. If the animal holder does not provide a respiratory (anesthetic)
mask, such a mask can be easily built: take a 20 mL plastic
syringe, cut off the tip approximately 15 mm from the bottom
of the syringe, yielding a funnel shaped mask. Finally deflate
and smooth the cutting edges of the mask using a file.
5. The temperature of the water will be much higher than the
temperature of the rubber mat and depends on the length and
material of the tubing used. Hence the temperature of the bath
must be adapted to the local setup.
6. It is important to make sure that all tubes or cables are kept
close to the animal bed in order to avoid them from getting
caught while driving the animal bed into the magnet.
7. Other examples of gadolinium-based CAs: Magnevist® (Bayer
Schering Pharma, Berlin, Germany). In this case the CA is
already in solution, dilutions need to be made to reach the
desired concentration.
8. Longer tubing should allow extra extension from the injection
pump and the catheter in mouse tail vein.
9. To repeat the pilot scan, undo it and then start it with the
Traffic Light button while holding the shift key on the key-
board in order to force all adjustments to run again.
10. When using an ordinary RARE scan that does not automati-
cally provide several scans with different repetitions (TRs),
prepare and run a set of six separate scans: first run one scan
with TR ¼ 9000 ms, then clone this scan 5 times, adjust the
TRs, and run these scans with GOP.
11. Adjust the slice position and angle according to the mouse
brain orientation and make sure the positioning for each
mouse is the same.
12. Keep the injection time (20 mins) constant. Adjust the injec-
tion rate only when the body weight changes. Here we suggest
giving a mouse a dose of 1 mmol/kg for the step-down
approach.
13. Using Protocol_T1w-GRE to check if the CA injection is fine,
you can already perform this scan while the injection is still
ongoing. From this, you can already after the first 5 mins of
injection get an idea about the success or failure of the CA
injection. Doing this scan earlier can save you time.
14. If there is a change in BBB permeability, you should also see
signal intensity changes within the brain parenchyma and not
only the vessel when running Protocol_T1w-GRE.
15. Failure of CA injection can have several reasons: the catheter is

displaced from tail vein; there is blood clot in the injection site;
the injection pump failed (e.g., syringe is no longer fixed
properly to the pump).
16. Make sure that the mouse head is correctly fixed throughout all
scans. If there is even a slight shift in the image pixel, the head
motion between post-contrast T1 maps can be corrected dur-
ing post processing by Avant’s Normalization Tool “ANTs.”
17. Histogram analysis function can be performed using :
ImageJ.
Acknowledgement
The authors wish to thank Till H€

ulnhagen, Henning Reimann, and
Joao Periquito for the development of all custom made analysis
tools with MATLAB program.
References
1. Cecchelli R, Berezowski V, Lundquist S, dysfunction. Mol Neurodegener 6:23.

Culot M, Renftel M, Dehouck MP, Fenart L https://doi.org/10.1186/1750-1326-6-23
(2007) Modelling of the blood-brain barrier in 7. Patel JP, Frey BN (2015) Disruption in the
drug discovery and development. Nat Rev blood-brain barrier: the missing link between
Drug Discov 6(8):650–661. https://doi.org/ brain and body inflammation in bipolar disor-
10.1038/nrd2368 der? Neural Plast 2015:708306. https://doi.
2. Fernandez-Lopez D, Faustino J, Daneman R, org/10.1155/2015/708306
Zhou L, Lee SY, Derugin N, Wendland MF, 8. Banks WA (2012) Brain meets body: the
Vexler ZS (2012) Blood-brain barrier perme- blood-brain barrier as an endocrine interface.
ability is increased after acute adult stroke but Endocrinology 153(9):4111–4119. https://
not neonatal stroke in the rat. J Neurosci 32 doi.org/10.1210/en.2012-1435
(28):9588–9600. https://doi.org/10.1523/ 9. Banks WA (2016) From blood-brain barrier to
JNEUROSCI.5977-11.2012 blood-brain interface: new opportunities for
3. Cramer SP, Modvig S, Simonsen HJ, Frederik- CNS drug delivery. Nat Rev Drug Discov 15
sen JL, Larsson HB (2015) Permeability of the (4):275–292. https://doi.org/10.1038/nrd.
blood-brain barrier predicts conversion from 2015.21
optic neuritis to multiple sclerosis. Brain 138 10. Blockx I, Einstein S, Guns PJ, Van
(Pt 9):2571–2583. https://doi.org/10.1093/ Audekerke J, Guglielmetti C, Zago W,
brain/awv203 Roose D, Verhoye M, Van der Linden A, Bard
4. Shlosberg D, Benifla M, Kaufer D, Friedman A F (2016) Monitoring blood-brain barrier
(2010) Blood-brain barrier breakdown as a integrity following amyloid-beta immunother-
therapeutic target in traumatic brain injury. apy using gadolinium-enhanced MRI in a
Nat Rev Neurol 6(7):393–403. https://doi. PDAPP mouse model. J Alzheimers Dis 54
org/10.1038/nrneurol.2010.74 (2):723–735. https://doi.org/10.3233/JAD-
5. Marques F, Sousa JC, Sousa N, Palha JA 160023
(2013) Blood-brain-barriers in aging and in 11. Chacko AM, Li C, Pryma DA, Brem S,
Alzheimer’s disease. Mol Neurodegener 8:38. Coukos G, Muzykantov V (2013) Targeted
https://doi.org/10.1186/1750-1326-8-38 delivery of antibody-based therapeutic and
6. Abdul Muneer PM, Alikunju S, Szlachetka imaging agents to CNS tumors: crossing the
AM, Murrin LC, Haorah J (2011) Impairment blood-brain barrier divide. Expert Opin Drug
of brain endothelial glucose transporter by Deliv 10(7):907–926. https://doi.org/10.
methamphetamine causes blood-brain barrier 1517/17425247.2013.808184
12. Absinta M, Sati P, Reich DS (2016) Advanced blood volume, and blood-brain barrier perme-
MRI and staging of multiple sclerosis lesions. ability, using dynamic contrast-enhanced T(1)-
Nat Rev Neurol 12(6):358–368. https://doi. weighted MRI at 3 tesla. Magn Reson Med 62
org/10.1038/nrneurol.2016.59 (5):1270–1281. https://doi.org/10.1002/
13. Lin Y, Pan Y, Wang M, Huang X, Yin Y, mrm.22136
Wang Y, Jia F, Xiong W, Zhang N, Jiang JY 17. Feng Y, Jeong EK, Mohs AM, Emerson L, ZR
(2012) Blood-brain barrier permeability is pos- L (2008) Characterization of tumor angiogen-
itively correlated with cerebral microvascular esis with dynamic contrast-enhanced MRI and
perfusion in the early fluid percussion-injured biodegradable macromolecular contrast agents
brain of the rat. Lab Invest 92 in mice. Magn Reson Med 60(6):1347–1352.
(11):1623–1634. https://doi.org/10.1038/ https://doi.org/10.1002/mrm.21791
labinvest.2012.118 18. Niendorf T, Pohlmann A, Reimann HM,
14. Heye AK, Culling RD, Valdes Hernandez Waiczies H, Peper E, Huelnhagen T,
Mdel C, Thrippleton MJ, Wardlaw JM (2014) Seeliger E, Schreiber A, Kettritz R, Strobel K,
Assessment of blood-brain barrier disruption Ku MC, Waiczies S (2015) Advancing cardio-
using dynamic contrast-enhanced MRI. A sys- vascular, neurovascular, and renal magnetic res-
tematic review. NeuroImage Clin 6:262–274. onance imaging in small rodents using
https://doi.org/10.1016/j.nicl.2014.09.002 cryogenic radiofrequency coil technology.
15. Barnes SR, Ng TS, Montagne A, Law M, Zlo- Front Pharmacol 6:255. https://doi.org/10.
kovic BV, Jacobs RE (2016) Optimal acquisi- 3389/fphar.2015.00255
tion and modeling parameters for accurate 19. Wagenhaus B, Pohlmann A, Dieringer MA,
assessment of low Ktrans blood-brain barrier Els A, Waiczies H, Waiczies S, Schulz-Menger J,
permeability using dynamic contrast-enhanced Niendorf T (2012) Functional and morpho-
MRI. Magn Reson Med 75(5):1967–1977. logical cardiac magnetic resonance imaging of
https://doi.org/10.1002/mrm.25793 mice using a cryogenic quadrature radiofre-
16. Larsson HB, Courivaud F, Rostrup E, Hansen quency coil. PLoS One 7(8):e42383. https://
AE (2009) Measurement of brain perfusion, doi.org/10.1371/journal.pone.0042383
Chapter 24
In Vivo Pharmacokinetics of Magnetic Nanoparticles

Carlos Caro, M. Carmen Muñoz-Hernández, Manuel Pernia Leal,
and Marı́a Luisa Garcı́a-Martı́n
Abstract
Over the past few years, many papers have been published on the nanomedical applications of magnetic
nanoparticles. However, most studies lack important information about the in vivo behavior of these
nanoparticles, which is a critical aspect for their rational design. In this chapter we describe a simple
protocol for the in vivo characterization of the pharmacokinetics of magnetic nanoparticles intravenously
injected in mice, using basic MRI sequences.
Key words Magnetic nanoparticles, Dynamic T2-weighted MRI, T2 mapping, CPMG, Pharmacoki-
netics, Relative enhancement
1 Introduction
Magnetic nanoparticles (MNPs) have been investigated over past

two decades for multiple purposes in nanomedicine such as drug
delivery nanocarriers, hyperthermia, and magnetic resonance imag-
ing contrast agents [1–7]. In the field of MRI contrast agents, many
efforts have been undertaken to develop more sensitive and specific
materials that on the one hand could improve the current contrast
agents in terms of relaxation times, and on the other hand could
specifically target the diseased tissue, thus allowing for MR-based
molecular imaging. One of the major limitations of these nanoma-
terials for in vivo applications is their distribution within the body.
Most of the nanomaterials injected intravenously are rapidly
sequestered by the mononuclear phagocyte system (MPS) [8],
which reduces their efficacy as diagnostic or therapeutic agents
[9]. Many efforts have been made to enhance blood circulation
times and reduce accumulation in non-desired tissues, by modify-
ing the physicochemical characteristics of the nanomaterial, such as
Carlos Caro and M. Carmen Muñoz-Hernández contributed equally to this work.
409
410 Carlos Caro et al.
size, surface charge, and hydrophilic/hydrophobic balance, which

are directly involved in their in vivo distribution [10–14]. Another
important parameter that has to be taken into account when
designing nanomaterials for intravenous administration is the
plasma protein adsorption [15–17]. Thus, several in vitro studies
have reported the importance of the charges at the material surface
since very likely they determine the nonspecific interactions with
plasma proteins, and also cell uptake [18–23]. In vivo, the injected
nanomaterial could be altered by such interactions resulting in
stability loss and reduced stealth properties, which could lead to
rapid clearance by the MPS [24]. Different approaches have been
developed to minimize plasma proteins adsorption, such as the
synthesis of neutral or negatively charged nanomaterials [25], coat-
ing with polyethylene glycol layers [14, 26, 27] or, as reported
recently, the use of zwitterionic materials [20, 23, 28, 29]. Interest-
ingly, in vivo studies have recently shown that the charge at the
outermost layer, rather than the overall charge, plays a predominant
role in these undesirable interactions with plasma proteins and that
neutrally charged surface exhibits the best in vivo behavior, with
long blood circulation times, very good stealth properties, and the
highest bioavailability [30]. In conclusion, the rational design of a
viable nanomaterial for in vivo diagnostic or therapeutic applica-
tions, involves that variables such as blood circulation times, pro-
tein adsorption, passive or active targeting, etc., are taken into
account from the beginning and are fully characterized in vivo
using animal models. In this sense, MRI offers unique possibilities
for the in vivo study of pharmacokinetics and biodistribution of
magnetic nanoparticles.
2 Materials
2.1 Animals 1. Balb/c mice with ca. 22 g in weight, provided by Janvier Labs
and Small Material (Le Genest-Saint-Isle, France).
2. 26G catheters for tail vein cannulation.
3. Heparinized saline (50 U/mL).
2.2 Magnetic The studies reported in this chapter were performed with 6 nm
Nanoparticles manganese ferrite nanoparticles (MnFe2O4) coated with a 3 kDa
polyethylene glycol (PEG) chain.
2.3 MR Equipment 1. Horizontal 9.4 T small animal MR imaging and spectroscopy

system (Biospec 94/20) equipped with an Avance III console
(Bruker BioSpin, Ettlingen, Germany).
2. Actively shielded gradient coil BGA-12S with integrated shims,
providing gradient amplitudes up to 400 mT/m.
Pharmacokinetics of Magnetic Nanoparticles 411
3. A 40 mm diameter quadrature bird-cage resonator for rat head

and mouse body applications.
4. AutoPac-motorized positioning system, which includes: ani-
mal bed with teeth bar to fix the head of the animal, integrated
waterbeds to maintain the body temperature during anesthesia,
and anesthesia mask for continuous delivery of isoflurane dur-
ing the procedure.
5. MRI scanner workstation running ParaVision 6.0 (Bruker
BioSpin, Ettlingen, Germany).
2.4 Other Equipment 1. Vaporizer and induction chamber to anesthetize the animals
outside the MR spectrometer.
2. Small Animal Monitoring and Gating system (Small Animal
Instruments Inc. Stony Brook, NY, USA), provided with respi-
ration and temperature sensors.
3. Animal physiology monitoring software (PC-SAM 32 v8.02,
Small Animal Instrument Inc.).
3 Methods
Detailed requirements to carry out in vivo pharmacokinetics and

biodistribution studies of intravenously injected magnetic nanopar-
ticles are described in this section. The protocol has been optimized
for mice, but it would also be valid for rats.
3.1 Synthesis 1. Mix and magnetically stir: 2 mmol of iron acetylacetonate,

of PEGylated MNPs 1 mmol of manganese acetylacetonate, 10 mmol of hexadeca-
nediol, 6 mmol of dodecylamine, 6 mmol of oleylamine, and
3.1.1 Synthesis of 6 nm
20 mL of benzyl ether.
MnFe2O4 Nanoparticles.
The synthesis is based 2. Heat the solution to 200 C for 2 h under a flow of nitrogen,
on the protocol reported by then heat the mixture to reflux (300 C) for 1 h.
Sun et al. [31] 3. Cool down the mixture to room temperature by removing the
heating mantle.
4. Wash the sample several times using ethanol, acetone, and
isopropanol as precipitation agents, centrifuge, and
re-disperse in toluene.
3.1.2 Functionalization 1. In a separation funnel, shake a solution of: 1 mL

of MNPs. The GA-PEG3kDa-OH (0.1 M in CHCl3), triethylamine (50 μL),
functionalization is carried and 1 mL oleic acid capped MnFe2O4 NPs (10 g L1 in
out by ligand exchange toluene).
with a previously 2. Dilute the solution with 5 mL toluene, 5 mL water, and 10 mL
synthesized gallol derived acetone.
ligand (GA-PEG3kDa-OH)
[32]
3. Collect the water phase and place it in a rotary evaporator to

remove the residual organic solvents.
4. Purify the PEGylated MNPs using centrifuge filters (MWCO:
100 kDa).
3.2 Animal Handling 1. Turn on the gas flow (oxygen or air at a flow rate of 1–1.5 mL/
min) and anesthesia (4–5% isoflurane), directed to the induc-
tion chamber, and place the mouse inside.
2. Once the animal is asleep, redirect the anesthesia to the outside
animal bed, reduce the isoflurane to 2–2.5%, and place the
mouse on the outside animal bed for tail vein cannulation.
3. Proceed with tail vein cannulation using a 26G catheter and
heparinized saline.
4. Redirect the anesthesia to the animal holder already placed on
the scanner, and quickly move the animal to the animal cradle,
which should have a temperature control system (water or air)
to maintain the temperature between 36.5 and 37.5 C during
the experimental procedure.
5. Once sensor for breathing and rectal temperature probe are
positioned correctly, check the animal monitoring unit. Anes-
thesia should be adjusted throughout the experiment to main-
tain a breathing rate between 50 and 80 breaths per minute.
6. At the end of the experiment, remove the animal from the
animal holder and place it on a warm location until it is awake
(2–3 min). Then return the animal to its cage.
3.3 Reference 1. Position the 40 mm-diameter resonator so that the mouse

Images abdomen is located at the center of the RF coil.
2. Adjust the tune and match of the resonator.
3. Introduce the animal holder inside the magnet making sure
that the center of the resonator is placed at the isocenter of the
magnet.
4. Load a positioning protocol and acquire the image with a field
of view (FOV) of 60 mm. Use these images to reposition the
animal if necessary (see Note 1).
5. Acquire axial and sagittal T1-weighted images using the
(FLASH) gradient echo sequence with the following or
approximate parameters: FOV ¼ 45 mm, matrix
size ¼ 256 256, slice thickness ¼ 1.5 mm, number of con-
tinues slices ¼ 19–20, repetition time (TR) ¼ 300 ms, echo
time (TE) ¼ 2.5 ms. Select the Fat Suppression option in the
Trigger Sub-card (see Note 2).
3.4 Pharma- The acquisition scheme used for the pharmacokinetics studies is as
cokinetics Study follows: high resolution T2-weighted images, T2 map, intravenous
injection of magnetic nanoparticles, dynamic T2-weighted
sequence (35 min), T2-weighted images and T2 map (see Note 3).
Fig. 1 Representative T2-weighted images with respiratory gating acquired at different experimental times
1. T2-weighted images with respiratory gating. Coronal

T2-weighted images using the RARE (rapid acquisition with
relaxation enhancement) sequence with the following or
approximate parameters: repetition time (TR) ¼ 1 s, echo
time (TE) ¼ 16 ms, rare factor ¼ 4, number of averages ¼ 2,
FOV ¼ 40 mm, matrix size ¼ 256 256, slice thick-
ness ¼ 1 mm, trigger for respiratory gating and fat suppression
activated. In Paravision 6.0 both options can be activated in the
“Contrast” card (see Note 4).
2. Quantitative T2 measurements. T2 is measured by using a
32-echo CPMG (Carl-Purcell-Meiboom-Gill) imaging
sequence, which is based on the so-called “MSME” (multi
slice multi echo) in Paravision. The geometry used is the
same as that of the T2-weighted images and the acquisition
parameters are as follows: TE values from 7 ms to 224 ms,
TR ¼ 3000 ms, FOV ¼ 4 cm, matrix size ¼ 128 128, slice
thickness ¼ 1 mm.
3. Dynamic T2-weighted sequence. Based on the RARE sequence,
but with reduced spatial resolution to improve temporal reso-
lution: TR ¼ 1 s, TE ¼ 16 ms, rare factor ¼ 4, number of
averages ¼ 1, number of repetitions ¼ 60 (acquisition time per
image ¼ 32 s), FOV ¼ 40 mm, matrix size ¼ 128 128, slice
thickness ¼ 1 mm. Start the acquisition, wait 2 min 8 s
(4 images) and then inject the MNPs (see Note 5).
4. Long-term pharmacokinetics. Repeat the T2-weighted and the
CPMG images at different experimental times (in the example
reported herein images were acquired at 0 h, 1 h, 24 h, 48 h,
and 1 week, Fig. 1).
3.5 Data Analysis 1. Load the T2-weighted sequence in the Paravision image display
and processing program and look for the image on the series
3.5.1 Dynamic
that gives good contrast between the different tissue regions
T2-Weighted Sequence
(see Note 6).
Fig. 2 Short-term pharmacokinetics by dynamic T2-weighted imaging of intravenously injected magnetic

nanoparticles. Selected regions of interest (ROIs) are outlined with dashed white lines on one of the series
images that provides good contrast. The right panels show the time-courses of the signal intensity (in arbitrary
units) corresponding to ROIs in the liver (upper) and kidney cortex (bottom)
2. Draw regions of interest (ROIs) on the tissues from which you

want to obtain the pharmacokinetics information (liver,
kidneys, spleen, muscle, etc.) (see Note 7). Get the average
intensity values over time (Fig. 2).
3. Normalize the arbitrary intensity values to obtain relative
enhancement values using the following equation (Fig. 3):

I t I 0
RE ¼ 100
I0
where RE is the modulus of relative signal enhancement, It is

the signal intensity at any given time after the nanoparticles
Fig. 3 Pharmacokinetics of magnetic nanoparticles injected intravenously in mice. The upper panel shows the
signal intensity in arbitrary units, and the lower panel shows the normalized signal expressed as relative
enhancement (RE)
injection, and I0 is the signal intensity before the injection (see

Note 8).
3.5.2 Quantitative T2 1. Repeat steps 1 and 2 as in the dynamic analysis above (Fig. 4).
Imaging 2. Calculate the T2 values by fitting the following equation
(Fig. 4)
Mðt Þ ¼ M0 eTE=T2 þ n
where M(t) is the signal intensity at time TE, M0 is the signal

intensity at equilibrium, and n is the background noise.
3. For long-term pharmacokinetics, measure the T2 values using
the approximate same ROIs at all different experimental times
(Table in Fig. 4).
4 Notes
1. The positioning protocol we refer to is generally called “Loca-

lizer” in Paravision 6.0 or above and “Tripilot” in previous
versions. It is based on the fast low angle shot (FLASH) gradi-
ent echo sequence and uses three orthogonal slices that inter-
sect at the isocenter producing a signal drop, due to partial
saturation, which can be used as a reference to reposition the
animal.
Fig. 4 Long-term pharmacokinetics by quantitative T2 measurements. Selected regions of interest (ROIs) are
outlined with dashed lines on one of the series images that provides good contrast. The right panel shows the
exponential fittings for T2 calculation in the different ROIs. Table at the bottom shows the average T2 values
(standard deviation) and the average increments in T2 (standard deviation) of three different animals at
different experimental times
2. These axial and sagittal T1-weighed images will be used as

anatomic reference to define the geometry for the pharmaco-
kinetic study.
3. T2*-weighted images are more sensitive to the field distortions
caused by superparamagnetic agents, however, they are prone
to susceptibility artifacts caused by motion, air, etc., which are
especially pronounced in the abdomen. Consequently,
T2-weighted, instead of T2*-weighted, images are used in this
type of studies.
4. The protocol described herein has been optimized to provide
good contrast and anatomical detail of the mice abdomen at
9.4 T.
5. This sequence is used to study the short-term pharmacokinet-

ics (30 min) following the intravenous injection of magnetic
nanoparticles. The reproducibility of the injection is a critical
aspect in this type of studies, thus you can either use an auto-
matic injector or define a very strict protocol for manual
injection.
6. This type of analysis can be performed with many other soft-
ware packages, some of them freely available, such as ImageJ
(National Institute of Health, USA).
7. Selection of ROIs must be carried out carefully, avoiding the
edge of the organs, since dynamic images are acquired without
motion correction and therefore measures close to the edge of
the organ can give overestimation or underestimation of the
real values.
8. Data normalization is needed for statistical analysis. There are
different ways of normalizing data. We use the modulus of the
relative enhancement because the results thereby obtained are
easier to interpret when compared to experiments using posi-
tive contrast, such as dynamic contrast enhancement with
gadolinium-based contrast agents [32].
Acknowledgements
The MRI system used in this work has been funded by the Spanish
Ministry of Science and Innovation (National Plan for Scientific
Research, Development and Technological Innovation 2008-
2011) and the European Regional Development Fund
(PCT-420000-2010-3).
References
1. Petros RA, Desimone JM (2010) Strategies in 5. Pernia Leal M, Torti A, Riedinger A, La
the design of nanoparticles for therapeutic Fleur R, Petti D, Cingolani R, Bertacco R, Pel-
applications. Nat Rev Drug Discov 9 legrino T (2012) Controlled release of doxo-
(8):615–627 rubicin loaded within magnetic thermo-
2. Lee JH, Jang JT, Choi JS, Moon SH, Noh SH, responsive nanocarriers under magnetic and
Kim JW, Kim JG, Kim IS, Park KI, Cheon J thermal actuation in a microfluidic channel.
(2011) Exchange-coupled magnetic nanopar- ACS Nano 6(12):10535–10545
ticles for efficient heat induction. Nat Nano- 6. Gazeau F, Lévy M, Wilhelm C (2008) Opti-
technol 6(7):418–422 mizing magnetic nanoparticle design for
3. Mornet S, Vasseur S, Grasset F, Duguet E nanothermotherapy. Nanomedicine 3
(2004) Magnetic nanoparticle design for med- (6):831–844
ical diagnosis and therapy. J Mater Chem 14 7. Figuerola A, Di Corato R, Manna L, Pellegrino
(14):2161–2175 T (2010) From iron oxide nanoparticles
4. Lee JH, Huh YM, Jun YW, Seo JW, Jang JT, towards advanced iron-based inorganic materi-
Song HT, Kim S, Cho EJ, Yoon HG, Suh JS, als designed for biomedical applications. Phar-
Cheon J (2007) Artificially engineered mag- macol Res 62(2):126–143
netic nanoparticles for ultra-sensitive molecular 8. Challenor M, Gong P, Lorenser D, House MJ,
imaging. Nat Med 13(1):95–99 Woodward RC, St Pierre T, Fitzgerald M,
Dunlop SA, Sampson DD, Iyer KS (2014) The 20. Garcı́a KP, Zarschler K, Barbaro L, Barreto JA,
influence of NaYF4:Yb,Er size/phase on the O’Malley W, Spiccia L, Stephan H, Graham B
multimodality of co-encapsulated magnetic (2014) Zwitterionic-coated "stealth" nanopar-
photon-upconverting polymeric nanoparticles. ticles for biomedical applications: Recent
Dalton Transactions (Cambridge, England: advances in countering biomolecular corona
2003) 43(44):16780–16787. https://doi. formation and uptake by the mononuclear
org/10.1039/c4dt01597e phagocyte system. Small 10(13):2516–2529.
9. Wang B, He X, Zhang Z, Zhao Y, Feng W https://doi.org/10.1002/smll.201303540
(2013) Metabolism of nanomaterials in vivo: 21. Moyano DF, Saha K, Prakash G, Yan B,
blood circulation and organ clearance. Acc Kong H, Yazdani M, Rotello VM (2014) Fab-
Chem Res 46(3):761–769 rication of corona-free nanoparticles with tun-
10. Karakoti AS, Das S, Thevuthasan S, Seal S able hydrophobicity. ACS Nano 8
(2011) PEGylated inorganic nanoparticles. (7):6748–6755. https://doi.org/10.1021/
Angewandte Chemie - International Edition nn5006478
50(9):1980–1994 22. Jiang Y, Huo S, Mizuhara T, Das R, Lee YW,
11. Tong S, Hou S, Zheng Z, Zhou J, Bao G Hou S, Moyano DF, Duncan B, Liang XJ,
(2010) Coating optimization of superpara- Rotello VM (2015) The interplay of size and
magnetic iron oxide nanoparticles for high T2 surface functionality on the cellular uptake of
relaxivity. Nano Lett 10(11):4607–4613 sub-10 nm gold nanoparticles. ACS Nano 9
12. Howard MD, Jay M, Dziubla TD, Lu X (2008) (10):9986–9993. https://doi.org/10.1021/
PEGylation of nanocarrier drug delivery sys- acsnano.5b03521
tems: state of the art. J Biomed Nanotechnol 23. Elci SG, Jiang Y, Yan B, Kim ST, Saha K,
4(2):133–148 Moyano DF, Yesilbag Tonga G, Jackson LC,
13. Yoo JW, Chambers E, Mitragotri S (2010) Rotello VM, Vachet RW (2016) Surface charge
Factors that control the circulation time of controls the suborgan biodistributions of gold
nanoparticles in blood: Challenges, solutions nanoparticles. ACS Nano 10(5):5536–5542.
and future prospects. Curr Pharm Des 16 https://doi.org/10.1021/acsnano.6b02086
(21):2298–2307 24. Gustafson HH, Holt-Casper D, Grainger DW,
14. Pernia Leal M, Rivera-Fernandez S, Franco Ghandehari H (2015) Nanoparticle uptake:
JM, Pozo D, de la Fuente JM, Garcia-Martin the phagocyte problem. Nano Today 10
ML (2015) Long-circulating PEGylated man- (4):487–510. https://doi.org/10.1016/j.
ganese ferrite nanoparticles for MRI-based nantod.2015.06.006
molecular imaging. Nanoscale 7 25. Alexis F, Pridgen E, Molnar LK, Farokhzad
(5):2050–2059. https://doi.org/10.1039/ OC (2008) Factors affecting the clearance and
c4nr05781c biodistribution of polymeric nanoparticles.
15. Peer D, Karp JM, Hong S, Farokhzad OC, Mol Pharm 5(4):505–515
Margalit R, Langer R (2007) Nanocarriers as 26. Pelaz B, Del Pino P, Maffre P, Hartmann R,
an emerging platform for cancer therapy. Nat Gallego M, Rivera-Fernández S, De La Fuente
Nanotechnol 2(12):751–760 JM, Nienhaus GU, Parak WJ (2015) Surface
16. Dobrovolskaia MA, McNeil SE (2007) Immu- functionalization of nanoparticles with poly-
nological properties of engineered nanomater- ethylene glycol: effects on protein adsorption
ials. Nat Nanotechnol 2(8):469–478 and cellular uptake. ACS Nano 9
17. Dobrovolskaia MA, Aggarwal P, Hall JB, acsnano.5b01326
McNeil SE (2008) Preclinical studies to under-
stand nanoparticle interaction with the 27. Suk JS, Xu Q, Kim N, Hanes J, Ensign LM
immune system and its potential effects on (2016) PEGylation as a strategy for improving
nanoparticle biodistribution. Mol Pharm 5 nanoparticle-based drug and gene delivery.
(4):487–495 Adv Drug Deliv Rev 99(Part A):28–51.
https://doi.org/10.1016/j.addr.2015.09.
18. Verma A, Stellacci F (2010) Effect of surface 012
properties on nanoparticle-cell interactions.
Small 6(1):12–21. https://doi.org/10.1002/ 28. Ye L, Zhang Y, Yang B, Zhou X, Li J, Qin Z,
smll.200901158 Dong D, Cui Y, Yao F (2016) Zwitterionic-
modified starch-based stealth micelles for
19. Zhu M, Nie G, Meng H, Xia T, Nel A, Zhao Y prolonging circulation time and reducing mac-
(2013) Physicochemical properties determine rophage response. ACS Appl Mater Interfaces
nanomaterial cellular uptake, transport, and 8(7):4385–4398. https://doi.org/10.1021/
fate. Acc Chem Res 46(3):622–631. https:// acsami.5b10811
doi.org/10.1021/ar300031y
29. Huang P, Liu J, Wang W, Zhang Y, Zhao F, 31. Sun S, Zeng H, Robinson DB, Raoux S, Rice
Kong D, Liu J, Dong A (2016) Zwitterionic PM, Wang SX, Li G (2004) Monodisperse
nanoparticles constructed from bioreducible MFe2O4 (M ¼ Fe, Co, Mn) nanoparticles. J
RAFT–ROP double head agent for shell shed- Am Chem Soc 126(1):273–279. https://doi.
ding triggered intracellular drug delivery. Acta org/10.1021/ja0380852
Biomater 40:263–272. https://doi.org/10. 32. Pernia Leal M, Munoz-Hernandez C, Berry
1016/j.actbio.2015.11.038 CC, Garcia-Martin M (2015) In vivo pharma-
30. Pernia Leal M, Caro C, Garcia-Martin ML cokinetics of T2 contrast agents based on iron
(2017) Shedding light on zwitterionic mag- oxide nanoparticles: optimization of blood cir-
netic nanoparticles: limitations for in vivo culation times. RSC Adv 5(94):76883–76891.
applications. Nanoscale 9(24):8176–8184. https://doi.org/10.1039/C5RA15680G
https://doi.org/10.1039/c7nr01607g
Part VII
Anesthesia and Advanced Contrast Agents

Chapter 25
Anesthesia and Monitoring of Animals During MRI Studies

Jordi L. Tremoleda, Sven Macholl, and Jane K. Sosabowski
Abstract
The use of imaging represents a major impact on the refinement and the reduction of in vivo studies in
animal models, in particular for allowing longitudinal monitoring of the onset and the progression of
disease within the same animal, and studying the biological effects of drug candidate and their therapeutic
effectiveness. But the use of imaging procedures can affect animal physiology, and the need to anesthetize
the animals for imaging entails potential health risks. During anesthesia, there is an inevitable autonomic
nervous system depression which induces cardiovascular depression, respiratory depression, and hypother-
mia. Also other procedures associated with imaging such as animal preparation (e.g., fasting, premedica-
tion), blood sampling, and dosage/contrast agent injections can also affect physiology and animal welfare.
All these factors are likely to have confounding effect on the outcome of the imaging studies and pose
important concerns regarding the animal’s well-being, particularly when imaging immune deprived animals
or diseased animals. We will discuss these challenges and considerations during imaging to maximize
efficacious data while promoting animal welfare.
Key words Anesthesia, Physiological monitoring, Animal welfare, 3Rs
1 Introduction
The use of imaging technologies is instrumental in biomedical

research due to their great scope for noninvasively studying bio-
chemical and biological processes in the living animal. Their appli-
cation represents a major impact on the “3Rs” principles
particularly in the reduction of animals used for in vivo studies,
allowing longitudinal monitoring of the onset and the progression
of disease within the same animal through serial imaging and also to
test the therapeutic effectiveness of new treatment. Preclinical
magnetic resonance imaging (MRI) remains the most versatile
imaging modality which has been extensively used for anatomical,
functional, and physiological characterizations of tissues/organs
[1]. Current MRI systems can routinely achieve a spatial resolution
of 100 μm in all dimensions in living animals providing high-quality
anatomical detail. But MR techniques can also provide information
on chemical composition (spectroscopy) and other parameters such
423
424 Jordi L. Tremoleda et al.
as cardiovascular function [2] and neuronal network alterations

[3]. Its nonionizing 3D imaging functionality is truly noninvasive
avoiding any exposure of the animals to any ionizing radiation. This
is especially important during serial imaging giving an advantage
over modalities such as CT, SPECT, and PET. A challenge is the
higher spatial resolution required with rodents compared to
humans. In order to segment each body into the same number of
voxels, the rodent voxel volume must be approximately 3000 times
smaller than that of a human. In MRI, the answer to this challenge
lies in higher magnetic fields system (e.g., 11.7 T), strong magnetic
field gradient systems (e.g., 1000 mT/m) [4] and preclinical-
specific radio-frequency coils and cryogenic probes [5]. Overall,
MRI methodology provides a major preclinical refinement and
reduction tool, delivering clinically relevant outcomes in a mini-
mally invasive manner in a living organism, allowing for repetitive
monitoring in the same animal, maximizing biological relevance,
thus delivering a rapid, efficacious, and reasonably cost-effective use
of animal disease models. A major difference with human MRI
remains which is the need to anesthetize the animals for imaging,
which per se entails potential health risks in the imaged animals.
Indeed, during anesthesia there is inevitable autonomic nervous
system depression which induces cardiovascular and respiratory
depression and hypothermia. All these factors have a profound
effect on the animal’s homeostasis and may thereby confound the
image quality and interpretation, or statistically speaking lead to
bias and increased variability. In addition, other experimental pro-
cedures occasionally performed along with imaging such as fasting,
premedication, blood sampling and the administration of contrast
agents or imaging biomarkers can also affect physiology and animal
welfare. Furthermore functional pharmacological blood oxygen
level dependent (BOLD) MRI can be markedly affected by anes-
thesia due to its effects on blood flow, blood oxygenation levels,
and cardiac and respiratory functions [6]. Similarly, the need for
lengthy acquisition times for ultra-high resolution MRI also poses
challenges for adequate physiological monitoring. All these factors
are likely to have confounding effects on the outcome of the
imaging studies and pose important concerns regarding the ani-
mal’s well-being, particularly when imaging disease-modeling
animals [7].
In the remainder of this chapter we will discuss the challenges
associated with setting up and carrying out preclinical MRI proce-
dures, addressing the protocols related to animal handling, induc-
tion and maintenance of anesthesia and physiological monitoring,
to promote appropriate animal care and to minimize potential
confounding effects on imaging outcomes.
MRI and Anaesthesia 425
2 Methods
2.1 Planning When thinking on undertaking a preclinical MRI study, it is impor-

and Setting Up an MRI tant to allocate enough time for planning, addressing all the animal
Study regulatory guidelines and defining all scientific objectives. The
following protocol may be used as guidelines.
1. Define specific targets and procedures
l What are the scientific objectives and how are they relevant
to the clinical scenario and/or in vivo mechanistic model-
ing? Include a clear hypothesis to be tested via the primary
(and optionally secondary) objectives.
l What is the rationale for using imaging, and in particular
MRI?
l Which stage of disease models will be used and why? Define
why the animal species and model is being selected.
l Describe the rationale for the specific imaging timeframes
used (onset and/or progression of disease and/or response
to treatments).
l Define imaging outcomes (both qualitative and quantita-
tive) and data analysis algorithms.
l Describe how data quality control is handled (e.g., imaging
system calibration).
l Draw up prospective plans for further imaging sessions
and/or co-registration with other imaging modalities.
2. Ensure that all animal procedures are approved by the specific
institutional animal welfare body (AWERB-UK [8]; IACUC-
USA [9]) and are conform to regulatory frameworks
(EU Directive 2010/63/EU [10]; US-PHS “The Guide for
the Care and Use of Laboratory Animals; Health Research
Extension Act of 1985 [11]) for the care and use of animals.
3. For budget preparation, include costs for animal purchase,
transport, housing, time for technical and veterinarian support,
imaging acquisition, data storage, processing and analysis.
4. Define experimental procedures, ensuring (1) the animal
model can be set up and specific MRI experiments can be
done in the facility, (2) biosafety requirements are fulfilled
(e.g., need for quarantine), (3) general and specific housing
capabilities are available, and (4) supporting staff is trained and
available.
5. Design the study including the number of experimental and
control groups, steps taken for randomization and blinding, a
workflow to work out the experimental procedures that will be
carried out within the same animal for severity assessment.
Check that this is all in accordance with the ARRIVE

guidelines [12].
The experimental protocol should include:
l Health surveillance/general check prior to study.
l Selection of animal strain, sex, age, weight (range), and geno-
type (e.g., knock-out or transgenic).
l Plan specifics for animal housing (individual/in groups); IVC
caging; isolators; specific SPF housing; specific bedding; cage
enrichment.
l Define experimental group, including appropriate any sample
size calculation used and number of independent replications of
each experiment.
l Selection of statistical methods to be used for each analysis,
including data normalization and a measure of precision (e.g.,
standard error or confidence interval).
l Pretreatment: fasting, specific feeding, premedication, contrast
agents including dose and route of administration.
l Anesthesia regime: selection of drug, dosage and route of
administration.
l Physiological monitoring (for depth of anesthesia and imaging
monitoring) with clear definition of adverse effects, care proce-
dures, and humane endpoints.
l Record keeping: reporting all the animal procedures, usage, and
discharge records according to regulatory requirements.
2.2 Transport Animals often have to be transported from general holding rooms
of Animals to the imaging facility. One consequence is that this will require full
to the Imaging Unit compliance with the institutional health screening controls to min-
and Acclimatization imize disease transfer to any other animals in the facility. A second
consequence is that for the transported animals themselves experi-
ence stress leading to an increase of the glucocorticoid level, a loss
in body weight and immune response suppression which have been
observed for up to 48 h after transport in rodents [13]. Therefore
an acclimatization period, typically between 3 and 7 days depend-
ing on the procedure and transport duration, is required after
animal transport. Imaging facilities should make housing capacity
available to allow for the acclimatization of transferred animals and
reduce stress before imaging takes place. Housing in the imaging
facility is also useful for monitoring animal recovery after anesthe-
sia. A sufficiently long recovery period should also be included after
imaging, depending on the type of anesthetic, the length of anes-
thesia, and any experimental procedures carried out. If an imaging
unit is integrated within a centralized animal unit, transport time
and acclimatization may be considerably reduced.
2.3 Animal Different procedures will be required depending on the animal

Preparation model used and the in vivo experimental protocol required prior
for Imaging to imaging. This may include procedures such as fasting, specific
treatments/dosing, and behavior assessments. Such additional pro-
cedures need to be well planned ahead and appropriately moni-
tored. As regulated procedures, these need to be properly reported
as per specific regulatory guidelines (Note, fasting is a regulated
procedure). Regular assessments of the animals include regular
body weight measurements and standard clinical checks on such
aspects like fur status, general behavior, and peer interaction, alert-
ness, good food/water uptake, feces control and absence of diar-
rhea, and face expression including healthy eyes. Passing these
criteria, animals can be accepted into the study. Then, these mea-
surements and assessments prior to any imaging serve as physiolog-
ical baseline values.
On the day of imaging, animals may also require more specific
interventions such as vessel cannulation, implantation of ECG
probes, and tracheal intubation. These need to be carried out
according to institutional guidelines for aseptic surgery [14] with
good thermoregulation control and a suitable analgesic and anes-
thesia regime to ensure minimal physiological impact of these
interventions. The pre-anesthetic exam should contain, but is not
limited to, confirmation of animal’s identification, sex, age, body
weight, body/fur condition, hydration/color of mucous mem-
branes status, heart rate and rhythm, respiratory rate, signs of
diarrhea, evidence of normal food and water consumption and
normal production of urine and feces in the cage. The body condi-
tion of the animal will impact on anesthesia induction, mainte-
nance, and recovery (e.g., obese animals may respond more
slowly to anesthetics).
Prior to imaging acquisition animals will be placed on the
“imaging bed” which will need to be fitted with anesthesia supply,
thermoregulation equipment for the anesthetized animals and
physiological monitoring sensors. Anesthesia regimes and monitor-
ing systems are discussed below.
2.4 Induction Anesthesia is generally required during imaging to ensure humane

of Anesthesia restraint of the animals. Most anesthetics induce a certain degree of
autonomic nervous system depression, which triggers cardiovascu-
lar depression (e.g., reduction in cardiac output, blood pressure)
and respiratory depression (e.g., hypoxia, hypercapnia) and induces
hypothermia, affecting whole body metabolism. Such effects can be
critical in laboratory rodents which have a small body size, high
body surface-to-weight ratio, and high metabolic rate, compromis-
ing the pharmacological efficacy of anesthetic agents and their
ability to thermoregulate. It is important to ensure that anesthe-
tized animals remain in a stable physiological state with consistent
cardiovascular, respiratory function and body temperature during

imaging.
The experimental protocol should include:
l Animal transport—ensuring biosafety based on health surveil-
lance program.
l Acclimatization period (3–7 days; shorter if integrated imaging
& housing units).
l Animal preparation and pretreatments (e.g., dosing/fasting).
l Regular general health status monitoring (body weight) with
good record keeping.
2.5 Anesthesia Injectable and inhaled anesthetics are commonly used in preclinical
Regimes imaging. Gas anesthetics are most suitable and recommended for
imaging due to their rapid onset and recovery times, and faster
elimination. Inhaled anesthetics with medical air or medical oxygen
as gas carrier allow for a better control of depth of anesthesia and
degree of oxygenation than injectables [15]. This is particularly
critical when undertaking long-term imaging procedures due to
the risk of developing hypoxia, respiratory depression, hypercapnia,
and acidosis. Dosing with injectables remains challenging due to
the lack of interventional management and when additional dosing
is required which can lead to overdosing and intermittent changes
on the depth of anesthesia. Injectables also often cause a prolonged
recovery time, in particular opioids with their strong residual effect
[16]. Some injectables can be administered via infusion which may
allow to achieve a more steady plasma concentration over the
course of imaging.
Anesthetic dose rates for injectable agents will depend on spe-
cies used, administration route, age, sex, strain, body condition,
environment, experimental setup, previous drug treatments, and
the level of anesthesia required. During the initial period of use, it is
important to monitor animals closely and make any adjustments
necessary in the protocol for subsequent experiments.
2.5.1 Injectables The most commonly used injectable anesthesia agents are listed in
Anesthetics Table 1 with their individual advantages and disadvantages and
typical dosing regimes.
2.5.2 Inhalation Inhalation anesthesia is the recommended method for imaging

Anesthesia laboratory animals as this provides a rapid induction and recovery
and since inhalation anesthesia agents are safe, nonirritant, and
nonexplosive. Medical oxygen is commonly used as the carrier
with a flow rate between 0.5 and 1.5 L/min. Non-rebreathing
circuits are usually used to ensure minimum dead space and resis-
tance, and resulting waste/excess gas is removed by a scavenger
system to protect lab staff. The most commonly used anesthetic
Table 1
Commonly used injectable anesthesia agents
Agent Advantages Disadvantages Dosing

Fentanyl/fluanisone l Good analgesic l Cardiovascular and respiratory Mouse: 10 mL/kg; rat: 2.7 mL/kg (i.p.)
(Hypnorm™) based l Sedative depression Hypnorm™/Hypnovel™ (midazolam)/water
combination l Need to “top up” for long-term l Poor muscle relaxation alone mixture (1:1:2 vol) (120–140 min sleep time)
anesthesia. l Prolonged recovery time HypnormTH top up 0.3 mL/kg (mouse)
l Sedative effect can be reverse with l Hypersensitivity to noise 0.1 mL/kg (rat) i.p. (30–40 min sleep time)
buprenorphine to speeds up
recovery time
Ketamine based l Analgesic effects l High muscle rigidity unless Mouse and rats:
combination l Light sedation combined with other agents. Ketamine þ medetomidine: 75 mg/
l Wide safety margin l Increases intracranial pressure. kg þ 0.5–1 mg/kg i.p.
l Can increase blood pressure l Recovery often with ataxia and hyper Ketamine þ xylazine 75–100 mg/kg/10 mg/k
responsiveness. i.p. (60–120 min sleep time)
Atipamezole: 1 mg/kg i.p. (reverse agent)
Alfaxalone (Alfaxan) l Minimal respiratory/ l Administration route IV (rodents, 15–20 mg/kg (mouse) 10–12 mg/kg (rat) iv
cardiovascular depression. cats) or IM (primates) (10–15 min sleep time after bolus)
l Rapidly metabolized: good for 0.25–0.75 mg/kg/min iv infusion (long term)
repeat dosing
l Suitable for long-term anesthesia
Propofol (Rapinovet®, l Rapidly metabolized, l IV use only 26 mg/kg (mouse) 10–12 mg/kg (rat) iv
Diprivan®) l Good for continuous infusion for l No analgesic properties (10–15 min sleep time after bolus)
long-term anesthesia. l Severe respiratory depression: risk of 2–2.5 mg/kg/min iv infusion (long term-
l Rapid recovery apnea mouse)
l High safety: can be used in animals (0.5–1 mg/kg/min iv infusion (long term-rat)
with hepatic or renal impairment.
Barbiturates products l Sedative effect l No analgesic properties Pentobarbitone: 40–50 mg/kg i.p. (mouse)
l Hypnotic l Severe respiratory depression and (120–180 min sleep)
l Reasonable muscle relaxation hypotensive Thiopentone: 30 mg/kg iv 15 min sleep) (rat)
MRI and Anaesthesia
l Easy to over does

l High metabolites accumulation
l Caustic substances: use only iv route
429
(continued)
Table 1
430
(continued)

Chloral hydrate l Sedative effect l No analgesic properties 300–400 mg/kg i.p. (1–2 h sleep time) (mouse
l Hypnotic l Paralytic noted in rats and rats)
l Minimal CVS and respiratory l Terminal/non-recovery work only
depression
a-Chloralose l Sedative effect l No analgesic properties 50–60 mg/kg i.p. (rats) 120 mg/kg i.p. (mouse)
l Hypnotic. l IV use only (8–12 h; for non-recovery only)
Jordi L. Tremoleda et al.
l Suitable for long-term anesthesia. l Slow induction and recovery 50 mg/kg iv bolus followed by 25–40 mg/kg/
l Minimal CVS and respiratory associated with involuntary hr (rats)
depression excitement
l Terminal/non-recovery work only
Urethane l Suitable for long-term anesthesia. l Carcinogenic: only allowed to be 0.8–1.3 g/kg, i.p. (mouse and rats) duration of
l Minimal CVS and respiratory used with special justification! action 8–10 h (nonrecovery only)
depression l Terminal/non-recovery work only
® l l
Avertin Wide safety margin Local irritation/peritonitis 0.015 mL/g body wt of 2.5% i.p.
(tribromoethanol) l Good muscle relaxation l Handling and storage safety issues 30 min—supplemental doses of anesthesia:
l Rapid induction and recovery l Toxic effects minimum of one-half of the initial does up to
l Pharmaceutical-grade TBE (e.g., 1 mL max vol per animal
Avertin) is no longer available:
IACUC/AWERB
circuit type for laboratory rodents is Bain’s coaxial T-piece coupled

with an open facemask system mounted to the imaging bed.
The most commonly used inhalation anesthesia agents are
listed in Table 2 with their individual advantages and disadvantages
and typical dosing regimes.
It is important to appreciate the variation in response to anes-
thetics between different animal strains and thus, to reassure and
adjust the anesthesia protocol to the particular needs of a given
strain and experimental setup.
2.6 Monitoring Most anesthetic drugs will impact on respiration, the cardiovascular
and Impact system, and/or thermoregulation. It is important to monitor the
of Anesthesia During animals during imaging and provide any physiologic support to
Imaging Procedures ensure animal welfare and also to minimize any confounding effects
on data acquisition. The recommended approach is to monitor
respiratory function and body temperature, as the minimum stan-
dard applicable during imaging to control depth of anesthesia.
Cardiac function is also highly recommended. Direct visualization
is not possible and it is greatly suggested to use the existing moni-
toring equipment for laboratory rodents which is tailored to their
small body size, fast cardiac and respiratory rates, and importantly
in the case of MRI, is non-ferromagnetic to avoid any interference
with the magnetic field [17, 18]. Most MRI-compatible systems are
based on fiber-optic or carbon fiber equipment and power source or
batteries are adequately filtered/isolated to avoid magnetic
interference [19].
While protocols do not specifically define how frequently one
should monitor the animals during anesthesia, it is obvious that the
more invasive the procedure and/or the longer time under anes-
thesia, the more likely it is to interfere with normal homeostasis and
thus greater the need for more frequent/constant monitoring. Also
it is instrumental to support the acquisition of gated imaging,
minimizing the effects due to biological motion and also targeting
specific imaging sequences, in a given respiratory or cardiac phase.
The position of the animals and the imaging bed systems used
are also crucial, to ensure that the neck and head are well positioned
to avoid restricting the breathing, and also the body and extremities
to avoid restricting the circulation and to avoid any bruising, strains
or avulsion in the body structures.
2.6.1 Temperature Most anesthetic agents profoundly depress thermoregulation

[20]. Rodents are highly susceptible to hypothermia due to their
large surface area-to-body mass ratio and rapid metabolism. This is
particularly critical when imaging nude or hairless mice. Hypother-
mia will have confounding effects on glucose metabolism and heart
rate, which can significantly affect, e.g., FDG-PET studies and
echocardiography outcomes. Therefore monitoring body core
temperature is crucial. This is mostly done using rectal
Table 2
Commonly used inhalation anesthesia agents
432

Halothane l Potent anesthetic l Highly metabolized (hepatotoxic) Induction: 3–4%
l High therapeutic index l Cardiovascular depressant Maintenance 1–2% (rats
l Rapid induction and recovery (1–3 min) l Moderate hypotension: reduction in cardiac output and mice)
l Adequate muscle relaxation and peripheral vasodilatation)
l Nonirritant, nonflammable nor explosive l Respiratory depressant
l Easy to vaporize l Halothane sensitizes the heart to catecholamines
(sympathetic stimulation)
Jordi L. Tremoleda et al.
Isoflurane l Similar physical properties to halothane l Decreases arterial blood pressure (vasodilatation) Induction: 3–4%
l Rapid induction and recovery l More expensive than halothane Maintenance 1–2% (mice);
l Low toxicity and metabolic activity: highly safe l Strong smell; aversive response for repetitive use 1.5–2.5% (rats)
l Suitable for high frequency and log-term anesthesia l More potent respiratory depressant than halothane
l Minimal cardiovascular depression
l Moderate respiratory depression
l Good muscle relaxation
Sevoflurane l Faster induction and recovery times (3 times faster l Expensive but cost coming down Induction: 1–8%
than isoflurane) l Costly specific vaporizers Maintenance: 3–5% (mice
l Less respiratory depression that isoflurane l Provides an equally reliable anesthesia in and rats)
l Less struggling and excitement during induction laboratory mice.
l Metabolism similar to isoflurane: good for repetitive
and/or long-term anesthesia
l Blood glucose homeostasis is better maintained
l Method of choice for PET FDG myocardium uptake
studies
thermometers or thermocouples which are well fitted to the MRI

systems and generally do interface with an external heat source such
as circulating hot water blankets or blowing air systems that help to
maintain the animal’s body temperature. Hyperthermia is as dan-
gerous as hypothermia, thus the external heat source should be
“thermostatically controlled” and linked to the core temperature
reading. Heat loss must also be minimized during the animal
preparation before imaging (e.g., hair removal, alcohol application)
and fluid replacement, if required, should be warmed to 37 C.
The eyes should be protected both from drying off and expo-
sure keratitis by regular application of ophthalmic ointment.
2.6.2 Respiratory System Most anesthetics are known to cause respiratory depression. There-
fore it is important to reduce variability due to poor ventilation
including hypoxia, hypercapnia, and acidosis [21, 22]. Respiratory
monitoring is generally carried out via detecting the breathing
motion registered as compressions of a respiratory sensor placed
in contact with the animal’s chest. These systems are extensively
used during imaging and highly compatible with MRI. Motion
artifacts due to breathing can be eliminated from the images by
employing gated imaging. Typically the most prominent motion of
the diaphragm and abdomen occurs during inspiration, and acqui-
sitions are generally carried out during expiration.
Other more advanced approaches include the use of arterial
blood gas analysis, which is very valuable during functional MRI.
The partial measurements of oxygen (pO2) and carbon dioxide
(pCO2) and the pH of the blood are detected from a single blood
sample allowing levels of oxygenation, imbalance of CO2 produc-
tion and acid-base balance to be monitored. Impaired gas
exchange, or hyper- or hypoventilation can be corrected by chang-
ing the anesthetic regime, and when critical through using artificial
ventilation. Indeed, some studies will require the animal to be
mechanically ventilated and it is important to ensure that the animal
does not develop hypercapnia or hypoxia. Other advanced respira-
tory monitoring systems include digitized systems such as capno-
graphs that measure the CO2 level through a highly sensitive
infrared spectroscopy CO2 sensor in the inhaled and exhaled gas,
based on the CO2 values in the venous return to the heart and the
efficacy of breathing. This equipment provides a very good indica-
tor of the respiratory function by continuous measurements of the
CO2 level. Capnographs that measure the CO2 level in the inhaled
and exhaled gas between inspiration and expiration at the endotra-
cheal tube connector or face masks level are very good indicator of
the respiratory function [23].
2.6.3 Cardiovascular Basic cardiovascular monitoring, including heart rate and blood
System pressure, is highly recommended. Electrocardiograms (ECG) are
generally used to monitor heart rate and rhythm and help to detect
arrhythmias, myocardial ischemia, or metabolic disorders. ECG

measurements are also used to synchronize the heart rate to the
image acquisition, during gated imaging. Non-ferromagnetic elec-
trodes, MRI–compatible needles or patch electrodes are mostly
used. Pulse oximeters are also very useful to monitor arterial oxy-
genation and pulse during anesthesia, detecting any changes long
before the animal becomes cyanosed. The system provides real-time
continuous measurements of arterial O2 saturation, pulse strength,
breathing rate, blood flow and effort to breathe. The systems
available for rodents are not invasive and are MRI compatible [24].
Blood pressure measurements are also used during imaging
[25]. The mean arterial pressure (MAP) is the overall judge of the
state of the circulation, being the best indicator on how well tissues
are perfused. As the MAP falls, vital organ auto regulation and
perfusion is quickly compromised. Measuring MAP stability is
very important during functional MRI or when assessing the per-
fusion of contrast agents through specific tissues/organs during
contrast enhanced MRI. The direct blood pressure measurement
involves placing a catheter in an artery and connecting it to a
transducer. The procedure is generally invasive as the artery, femo-
ral or carotid usually, has to be exposed surgically. Indirect methods
typically use an inflatable tail cuff pressure sensor to detect the
arterial blood flow [26] but are less accurate than direct ones and
generally data acquisition is intermittent.
2.6.4 Repeated Some imaging protocols may require long acquisition times and/or
and Long-Term Anesthesia high frequency imaging with considerable impact on cumulative
time of anesthesia. In these circumstances it is recommended to use
minimally metabolized volatile anesthetics like isoflurane or sevo-
flurane, which allow for a quick induction, good surgical anesthesia
and fast recovery, so that animals can regain full physiological
functions quickly. Repeat imaging is likely to induce some
hypothalamic-pituitary-adrenal (HPA) axis and autonomic nervous
system responses to the chronic stress with prolonged elevation of
corticosterone and impairments to homeostasis [27]. It is of
utmost importance to monitor body temperature. Hypo- or hyper-
thermia will interfere with many electrophysiological outcome
measurements. The extremities and tail should be covered when-
ever possible. Long imaging procedures with lengthy anesthesia
may seriously compromise the hydration statues of the animals.
Therefore it is important to compensate for the fluid lost and
prevent animal dehydration. Parental administration of warmed
fluids (0.9% saline or lactated Ringer’s) may be applied and ideally
also the inspired gases may be humidified to avoid desiccation via
the airways. Eye protection may also be used to help maintain good
lubrication of the cornea and also to protect the eyes from exposure
keratitis.
2.6.5 Ventilation Some studies will require the animal to be mechanically ventilated,
which may have important effects on thoracic hemodynamics and
may also override the autonomic reflex control of breathing, which
normally maintains blood gas homeostasis. When monitoring a
ventilated animal, ideally, one should have oxygen/carbon dioxide
monitoring equipment (e.g., pulse oximeter, end tidal capno-
graph). The ventilation settings can then be adjusted to try to
maintain physiological levels of O2/CO2. Blood gas and electrolyte
analysis are useful but confer intermittent monitoring, and a local
analyzer is needed.
The most commonly observed physiological effects and health
risks during anesthesia and troubleshooting management are listed
in Table 3.
2.7 Materials Table 4 lists physiological parameters and exemplary MRI-

compatible systems for monitoring. Respiration and heart rate
signals can be fed into the MR console for gated imaging, e.g.,
imaging only during exhale phases. Temperature readings can be
used as input to a feedback loop for temperature control.
2.8 Recovery Monitoring should be carried out until the animal has fully recov-
ered from anesthesia, i.e., until the animal regained full conscious-
ness with essential physiological functions back to normal. Checks
should include in particular respiratory and cardiovascular function
and the ability to control body temperature. Only then the animal
may be transferred back to its housing cage.
During recovery, it is important to maintain the core tempera-
ture of the animal. Good practice is to place the animals on their
right side or in sternal recumbent position in a warm recovery cage
with no sawdust or suchlike bedding that may be inhaled. If the
animals have been under anesthesia for a relatively long time, it may
be worth giving oxygen, fluids and nutrition supplement. This can
include oral or parenteral support, e.g., high-energy moist foods
such as nutrient agar, jelly or crushed rodent pellets mixed with
water, presented in a Petri dish or other manner that does not
require the animal to reach up high. Additional analgesia should
be provided if the animal has undergone a painful procedure or if
there appears to be any sign of pain [28].
3 Conclusions and Outlook
Imaging technologies have dramatically increased the efficiency of

preclinical studies, providing a powerful, noninvasive, and clinically
translatable means of monitoring disease progression and therapeu-
tic response. The noninvasive acquisition of detailed in vivo anato-
mical and functional data represents an important milestone in
refinement and reduction in the use of animal models. However,
Table 3
Commonly physiological effects and management during anesthesia
Heath risks Troubleshooting issues

Respiration Hypoxia, hypercapnia and acidosis High breathing rate:
Effect exacerbated during prolonged l Increase anesthesia dose
Impact on HbO2/02 l Check anesthetic system
Low breathing rate/apnea:

l Animal may be too deep: Lightening or
reversing anesthesia
l Hypothermia: Check temperature
l Hyperventilation in oxygen
l Short term reaction to injectables (e.g.,
thiopentone)
Respiration fails:
l Manual ventilation if animal is intubated
l Gently massage the chest side to side-not very
effective (if animals is accessible outside the

MRI magnet)
l Supply O even if using injectable anesthesia
2
l Use specific reverse agents
Cardiac Rodents high heart rate (10 faster Tachycardia:

function than humans) and short circulation l May be a pain response: Too “light”
time anesthesia, increase anesthesia
Profound effect of anesthetics l Correct any fluid deficit (check for any
Heart rate, rhythm and pulse intensity bleeding)

Bradycardia:
l Animal may be too deep: Lightening or
reverse anesthesia
l Animal too cold: Control body temperature
l Can be surgically induced by vagal reflexes or
by anesthetic drugs
l Atropine may be indicated
Cardiac arrest
l Gently massage the chest side to side-not very
effective
l Stop anesthesia and transfer to O supply only
2
l Some drugs are effective in larger animals
(adrenalin) but seldom practical for

emergency procedures in rodents
Risk of Heart rate (high or slow at critical l Turn off gas anesthetics/administer reversal
anesthesia stage) anesthetic agent.
overdose Arrythmias l Maintain animal with O2 only (or ventilate
Respiration (high and/or shallow)- with oxygen)
diaphragmatic l Administer isotonic fluids.
Pulse weak l Warm up the animal slowly to increase body
Membranes pale temperature/metabolism
Table 4
Exemplary systems for MRI-compatible physiological monitoring
Physiological parameter Examples

Respiration monitoring and gating, ECG BioVet (m2m Imaging Corp., USA)
monitoring and gating Low-cost solution Uni Jena [28]
Model 1030 monitoring and gating system (Small Animal
Instruments, Inc., USA)
Physioguard II (Minerve, France)
Temperature monitoring BioVet (m2m imaging corp, USA)
Imaging cells (Minerve, France)
Model 1030 monitoring and gating system (Small Animal
OTG-M (Opsens, Canada)
Heating systems Hot air fan with feedback loop (Small Animal
Hot air tubes built into animal bed (Minerve, France)
Hot water tubes built into animal bed (aspect imaging,
Bruker)
Resistive heating system Oxford Inst. Radiation oncology
[28]
Pulse oximetry, heart rate monitoring MouseOx (STARR Life Sciences Corp., USA)
MouseSTAT (Kent Scientific Corp., USA)
Respiratory CO2 monitoring Capnograph V9004 (Harvard Apparatus, Ltd., UK)
Capnoscan (Kent Scientific Corp., USA)
Blood pressure monitoring Samba Preclin 420/360 transducer (Harvard Apparatus,
Ltd., UK)
TSD104A blood pressure transducer (Biopac Systems,
Inc., USA)
it is important to continue refining all the rodent bio-imaging

protocols including appropriate anesthetic regimen and monitor-
ing systems to ensure the animal’s well-being and to minimize
stress-related responses that would compromise the imaging
outcomes.
It is important to consolidate good protocols for handling,
anesthesia and monitoring the animals suitable to all the specific
needs of a wide range of imaging experiments. This is a rapidly
advancing field that holds great opportunities for further research
and technology development. Technological challenges are addres-
sing faster acquisitions, faster analysis software, and more versatile
integrated multimodality imaging. Imaging companies continue
developing and integrating physiological monitoring systems in
preclinical equipment, and strategies are needed to continue
supporting further investment and developments in the field. Simi-
larly, further studies on the impact of frequent and repetitive
anesthesia/imaging are needed to ensure the appropriate severity

assessments and its impact on lifetime experience for each studied
animal and to improve their welfare.
References
1. Driehuys B, Nouls J, Badea A et al (2008) 10. Directive 2010/63/EU (2016) Legislation for
Small animal imaging with magnetic resonance the protection of animals used for scientific
microscope. ILAR J 49:35–53 purposes. Available via http://ec.europa.eu/
2. Bakermans AJ, Abdurrachim D, Moonen RP environment/chemicals/lab_animals/legisla
et al (2015) Small animal cardiovascular MR tion_en.htm. Accessed 7 Nov 2016
imaging and spectroscopy. Prog Nucl Magn 11. Public Health Service Policy on Humane Care
Reson Spectrosc 88-89:1–47. https://doi. and Use of Laboratory Animals National Insti-
org/10.1016/j.pnmrs.2015.03.001 tutes of Health Office of Laboratory Animal
3. Febo M, Foster TC (2016) Preclinical mag- Welfare (2016) Available via http://grants.
netic resonance imaging and spectroscopy nih.gov/grants/olaw/references/
studies of memory, aging, and cognitive phspolicylabanimals.pdf. Accessed 7 Nov 2016
decline. Front Aging Neurosci 8:158. https:// 12. The ARRIVE Guidelines: Animal Research:
doi.org/10.3389/fnagi.2016.00158 Reporting of In Vivo Experiments (2016)
4. Sasser TA, Bahadur A, Frederick E et al (2016) Available via https://www.nc3rs.org.uk/sites/
Considerations When Setting Up a Preclinical default/files/documents/Guidelines/NC3Rs
In Vivo Imaging Laboratory. Available via %20ARRIVE%20Guidelines%202013.pdf
https://www.bruker.com/fileadmin/user_ Accessed 7 Nov 2016
upload/8-PDF-Docs/PreclinicalImaging/ 13. Obernier JA, Baldwin RL (2016) Establishing
Brochures/Considerations_Preclinical_Lab_ an appropriate period of acclimatization fol-
T158053.pdf. Accessed 7 Nov 2016 lowing transportation of laboratory animals.
5. Sack M, Wetterling F, Sartorius A et al (2014) ILAR J 47(4):364–369
Signal-to-noise ratio of a mouse brain (13) C 14. LASA (2010) Guiding Principles for Preparing
CryoProbe™ system in comparison with room for and Undertaking Aseptic Surgery. In:
temperature coils: spectroscopic phantom and Jennings M, Berdoy M (eds) A report by the
in vivo results. NMR Biomed 27(6):709–715. LASA Education, Training and Ethics section.
https://doi.org/10.1002/nbm.3110 www.lasa.co.uk/publications.html. Accessed
6. Ferrari L, Turrini G, Crestan V et al (2012) A 7 Nov 2016
robust experimental protocol for pharmaco- 15. Wolfensohn S, Lloyd M (2003) Anaesthesia of
logical fMRI in rats and mice. J Neurosci laboratory animals. In: Wolfensohn S, Lloyd M
Methods 204:9–18. https://doi.org/10. (eds) Handbook of laboratory animal manage-
1016/j.jneumeth.2011.10.020 ment and welfare, 3rd edn. Blackwell Publish-
7. Tremoleda JL, Kerton A, Gsell W (2012) ing Ltd, Oxford. https://doi.org/10.1002/
Anaesthesia and physiological monitoring dur- 9780470751077.ch7
ing in vivo imaging of laboratory rodents: con- 16. Lovell DP (1986) Variation in pentobarbitone
siderations on experimental outcomes and sleeping time in mice. 2. Variables affecting test
animal welfare. EJNMMI Res 2(1):44. results. Lab Anim 20:91–96. https://doi.org/
https://doi.org/10.1186/2191-219X-2-44 10.1258/002367786780865089
8. Berdoy M, Hawkins P et al (2015) Guiding 17. Data Acquisition System – BIOPAC Systems,
principles on good practice for animal welfare Inc. (2016) Available via http://www.biopac.
and ethical review bodies. In: Jennings M com/magnetic-resonance-imaging-mri-com
(ed) A report by the RSPCA Research Animals patible. Accessed 7 Nov 2016
Department and LASA Education, Training 18. SAII Small Animal Instruments, Inc. (2016)
and Ethics Section. Available via http://www. Available via http://www.i4sa.com/web_app/
lasa.co.uk/PDF/AWERB_Guiding_ main/defaultProduct.aspx?ID¼76&PT¼3.
Principles_2015_final.pdf. Accessed 7 Nov Accessed 7 Nov 2016
2016 19. Peden CJ, Menon DK, Hall AS et al (1992)
9. IACUC Resources (2016) Available via Magnetic resonance for the anaesthetist. Part
https://www.aalas.org/iacuc/iacuc_res II: Aanaesthesia and monitoring in MR units.
ources#.WADzWfkrJD8 Accessed 7 Nov 2016 Anaesthesia 47:508–517
20. Colby LA, Morenko BJ (2004) Clinical con- 25. Wagner AE, Brodbelt DC (1997) Arterial
siderations in rodent bioimaging. Comp Med blood pressure monitoring in anesthetized ani-
54(6):623–630 mals. J Am Vet Med Assoc 210:1279–1285
21. Hauber HP, Karp D, Goldmann T et al (2010) 26. Thal SC, Plesnila N (2007) Non-invasive
Effect of low tidal volume ventilation on lung intraoperative monitoring of blood pressure
function and inflammation in mice. BMC Pulm and arterial pCO2 during surgical anesthesia
Med 10:21. https://doi.org/10.1186/1471- in mice. J Neurosci Methods 159(2):261–267
2466-12-7 27. Wu XY, Hua YT, Lua L et al (2015) Effect of
22. Pecchiari M, Monaco A, Koutsoukou A et al pentobarbital and isoflurane on acute stress
(2014) Effects of various modes of mechanical response in rat. Physiol Behav 145:118–121
ventilation in normal rats. Anesthesiology 120 28. Kohn DF, Martin TE (eds) (2006) Guidelines
(4):943–950 for the assessment and management of pain in
23. Larach DR, Schuler G, Skeehan TM et al rodents and rabbits. Available via https://
(1988) Mass spectrometry for monitoring wwwaclamorg/Content/files/files/Public/
respiratory and anaesthetic gas waveforms in Active/position_pain-rodent-rabbitpdf
rats. J Appl Physiol 65:955–963 Accessed 7 Nov 2016
24. Available via http://www.starrlifesciences.
com/conscious-monitoring. Accessed 7 Nov
2016
Chapter 26
Advanced Contrast Agents for Multimodal Biomedical

Imaging Based on Nanotechnology
Daniel Calle, Paloma Ballesteros, and Sebastián Cerdán
Abstract
Clinical imaging modalities have reached a prominent role in medical diagnosis and patient management in
the last decades. Different image methodologies as Positron Emission Tomography, Single Photon Emis-
sion Tomography, X-Rays, or Magnetic Resonance Imaging are in continuous evolution to satisfy the
increasing demands of current medical diagnosis. Progress in these methodologies has been favored by the
parallel development of increasingly more powerful contrast agents. These are molecules that enhance the
intrinsic contrast of the images in the tissues where they accumulate, revealing noninvasively the presence of
characteristic molecular targets or differential physiopathological microenvironments. The contrast agent
field is currently moving to improve the performance of these molecules by incorporating the advantages
that modern nanotechnology offers. These include, mainly, the possibilities to combine imaging and
therapeutic capabilities over the same theranostic platform or improve the targeting efficiency in vivo by
molecular engineering of the nanostructures. In this review, we provide an introduction to multimodal
imaging methods in biomedicine, the sub-nanometric imaging agents previously used and the development
of advanced multimodal and theranostic imaging agents based in nanotechnology. We conclude providing
some illustrative examples from our own laboratories, including recent progress in theranostic formulations
of magnetoliposomes containing ω-3 poly-unsaturated fatty acids to treat inflammatory diseases, or the use
of stealth liposomes engineered with a pH-sensitive nanovalve to release their cargo specifically in the acidic
extracellular pH microenvironment of tumors.
Key words Imaging agents, Image Guided Drug Delivery, Magnetic Resonance Imaging, Nanotech-
nology, Positron Emission Tomography, Single Photon Emission Tomography, Theranostic agents, X-
Ray computed tomography
1 Introduction
In the last decades, clinical imaging modalities have reached a

prominent role in preclinical research, medical diagnosis, and
patient management [1, 2]. Progress has evolved from the techno-
logical advances in different imaging modalities as Positron Emis-
sion Tomography (PET), Single Photon Emission Tomography
(SPECT), X-Ray computerized tomography (CT), and Magnetic
Resonance Imaging (MRI), among others, which have become
441
442 Daniel Calle et al.
widely available both preclinical and clinical versions. These tech-

nological advances have been paralleled by the complementary
development of increasingly more powerful, specific, and safe con-
trast agents [3–6]. These are molecules that enhance the intrinsic
contrast of the images in the regions where they accumulate,
improving both sensitivity and specificity in tissue characterization
and disease diagnosis. However, despite their well-known advan-
tages, the first generation of contrast agents remained limited to
single imaging modalities or suffered from limited sensitivity and
specificity qualifications.
More recently, the advent of hybrid imaging methodologies has
triggered the need for contrast agents active within more than one
imaging modality [7–9]. Nanotechnological formulations have
paved the way to multimodal agents, able to target in vivo specific
receptor molecules, enzymes, or even whole tissue environments,
revealing their presence and activity by complementary imaging
methodologies. Moreover, the possibility to combine noninvasive
disease detection with individualized treatments has become a real-
ity in personalized medicine, through recent theranostic formula-
tions [10, 11]. Together, these improvements have configured the
development of Image Guided Drug Delivery, a methodology
promising to improve the efficiency of pharmaceuticals by visualiz-
ing noninvasively their delivery, their effects on specific molecular
targets, and the corresponding therapeutic consequences [12–15].
This review provides an introduction to these developments
with some practical examples implemented recently in our own
laboratories. We begin with an overview of the most utilized bio-
medical imaging modalities and the first generation of the asso-
ciated contrast agents, the nanotechnological formulations
implemented to improve their performance, and conclude with
some practical examples of the development of theranostic magne-
toliposomes to detect and treat tissue inflammation and of
pH-sensitive liposomes to target the acidic pH microenvironment
in cancer.
2 Multimodal Imaging in Biomedicine
Biomedical images originate from the interaction between an elec-

tromagnetic radiation and the biological specimen [16]. The elec-
tromagnetic spectrum (Fig. 1) spans from the shortest wavelengths
and highest energies associated to the cosmic rays, to the longest
wavelengths and lowest energies associated to the heating radia-
tion. The γ-radiation, X-rays, ultraviolet, visible and infrared lights,
micro- or radiofrequency waves, lay in between these two extremes.
Nuclear medicine approaches involve the use of highly
penetrating γ-radiations, generating high sensitivity (1012 M)
images with limited resolution (mm). Optical images may become
Multimodal Imaging Agents Using Nanotechnology 443
Fig. 1 The electromagnetic spectrum expressed as energy (top) or associated wavelength (bottom). The
middle panel shows some of the initial images obtained from each modality including radiofrequency (MRI)
[17], infrared [18], X-Rays [19], and γ-rays (PET) [20]
very sensitive (1015 M), as in microscopy, but their penetration

capacity within the biological specimen is limited in vivo (mm) and,
only very thin objects or highly superficial processes, can be imaged.
Finally, radiofrequency imaging provides excellent penetration
capacity and resolution (μm), at the expense of limited sensitivity
(103 M). Together, these considerations suggest that there is no
single optimal imaging modality, all of them presenting advantages
and limitations. This provides a sound basis to combine the differ-
ent imaging modalities in a hybrid manner to obtain a single
multimodal image including the information derived from the
different modalities and overcoming the specific limitations of
each modality with the advantages of the others [21].
The physical interaction between the radiation and the speci-
men involves, most of the times, its transmittance though the
biological medium and the detection of the transmitted radiation
across the object with suitable detectors [22]. This is the case of
optical and nuclear imaging. The image generated follows then the
Lambert-Beer law
I t ¼ I 0 eε d ð1Þ
where It is the intensity of the image detected, I0 the intensity of the

radiation source before the interaction, ε is the extinction coeffi-
cient revealing the density of the specimen toward the transmitted
radiation per unit of length, and d the distance traveled by the
radiation through the object. In the case of nuclear imaging, the
source of radiation is inside the specimen and decays rapidly with
time because of the radioactive decay law, so corrections are need to
account for this circumstance, particularly in the case of positron

emitters with very short half-lives.
In general, a planar image is obtained after detecting the trans-
mitted radiation either with a digital camera in the case of optical
imaging, or a tomographic section through a coronal array of
highly efficient scintillation detectors in nuclear or X-ray imaging
[22, 23]. The tomographic reconstruction process, normally used
in X-ray computerized tomography and nuclear medicine
approaches, was first proposed by Godfrey Hounsfield and plays a
fundamental role in nuclear imaging [24]. Since then a variety of
reconstruction algorithms have been proposed with increasing res-
olution and computing time efficiency [25].
The discovery of Magnetic Resonance Imaging disclosed a
completely new imaging strategy, not based on the transmittance
phenomenon [17]. Rather, it took advantage of the physical inter-
action between the external magnetic fields applied and the mag-
netic properties of biological nuclei in the sample, yielding high
resolution images (μm) despite the fact that the wavelength asso-
ciated to nuclear transitions may reach 10–100 m. The method
used highly penetrating, nonionizing radiofrequency radiations,
presented improved resolution when compared to nuclear medi-
cine and in vivo optical imaging approaches and depicted no detect-
able harmful effects on the body, all these advantages at the expense
of a more limited sensitivity.
3 Sub-nanometric Contrast Agents
Despite the intrinsic image contrast determined by the physical

properties of the tissues, the use of exogenous contrast enhancing
molecules has been generalized in all imaging modalities
[26–30]. These molecules are designed to enhance selectively and
sensitively the intensity of the images in regions of tissue presenting
particular physiological, cellular, or molecular properties, favoring
enormously the recent development of the molecular and cellular
imaging field [31].
The initial contrast agents were chemicals with sub-nanometric
dimensions active in the different regions of the electromagnetic
spectrum. Briefly, periodinated benzenes, Gadolinium chelate-
sand99mTc derivatives were among the first contrast agents used
in X-ray, MRI, and nuclear medicine, respectively
[27,32,33]. More specifically, a large variety of SPECT or PET
tracers have been proposed in the nuclear medicine field, with
dominant applications in oncologic and neurodegenerative dis-
eases. In general, PET or SPECT tracers contain two parts, a
targeting moiety providing the selectivity to the imaged process
and a radionuclide moiety, containing the radionuclide emitting
gamma rays with 140 meV (SPECT) or 511 meV (PET), enabling
the detection. The targeting moieties are vectorial reagents, includ-
ing either small chelates or even peptides, proteins, or antibodies
with molecular specificity for the desired target[34]. The radio-
nuclides involve mainly99mTc,68Ga, 111In or131I for SPECT
or18F,11C,15O for PET, among others.
A large variety of SPECT probes are commercially available to
monitor noninvasively tissue perfusion [35], hypoxia [36], inflam-
mation [37], thyroid function [38], or even cerebral activation
[39]. Their main advantage is the possibility to use relatively low
cost gamma cameras for detection, available in many hospitals and
the relatively long half-lives of the SPECT emitters (in the hour
range), while their main limitation is the reduced resolution of the
SPECT method [21, 22]. This limitation has been overcome, in
large part, through the use of PET probes, that offer increase spatial
and temporal resolution, at the expense of the more expensive PET
equipment and the reduced half-life (in the min. range) of the
positron emitters.18F-2-deoxiglucose (FDG) is probably the most
widely used PET tracer, despite its poor selectivity. FDG is taken up
by virtually all tissues in an analogous manner to glucose, but it is
not degraded by glycolysis, resulting in FDG accumulation in those
regions with enhanced glucose uptake. FDG reveals spectacularly
the presence of tumors or metastases, regions of cerebral activation
or even inflammatory lesions, providing probably the most success-
ful tracer in nuclear medicine. However, since glucose uptake is a
universal process, the use of FDG is hampered by its lack of selec-
tivity, making very difficult to discriminate tumoral from inflamed
regions, resulting in a non-negligible number of false positives
(up to 30% [40]) and requiring frequently complementary exam-
inations with additional tracers or modalities.
The first generation of contrast agents for MRI, involved
mainly Gd(III) chelates, able to enhance water relaxation rates in
those tissues where they accumulated [27]. Gd(III) is used because
it has seven unpaired, slow relaxing electrons, and depicts the
largest magnetic moment among the rare earth series. The ligands
most frequently used are either linear, derived from diethylendia-
mino pentaacetic acid (DTPA), or cyclic, derived from tetraazacy-
clododecane-1,4,7,10-tetraacetic acid (DOTA). In all these cases,
the ligand provides eight binding sites anchoring the Gd(III)
within the chelate, leaving free one the nine chelating sites of the
metal, for water contact. The contact between water in the solution
or tissue with the Gd(III), and the fast exchange of this water
molecule with the bulk solution, reduces very significantly the
relaxation times of tissue water, resulting in clearly enhanced
image intensity in those regions containing the chelate [27]. The
use of other lanthanides as Dy(III) may transform the same chelates
in T2 enhancing probes, due to the inherent T2 relaxing properties
of Dy(III) [41]. MRI contrast agents include additionally a large
variety of molecules able to enhance image intensity using other

mechanisms including mainly magnetization transfer methods
[42]. These agents are known as diamagnetic Chemical Exchange
Saturation Transfer (CEST) or Paramagnetic Chemical Exchange
Saturation Transfer (PARACEST) agents. These molecules can be
customized to reveal important aspects of the lesions including
properties of the microenvironment as pH [43], monovalent or
divalent ion concentration [44], or temperature [45], among
others.
Several probes have been proposed to investigate extracellular
pH (pHe) using Magnetic Resonance Spectroscopy (MRS) meth-
ods. Despite the reduced sensitivity of the spectroscopy approach,
these probes provide the advantage to detect directly pH changes,
through their effect on the chemical shifts of pH-sensitive protons,
mainly through in vivo31P NMR [46, 47] or1H Magnetic Reso-
nance Spectroscopic Imaging (MRSI) [48, 49]. More recently,
hyperpolarized13C strategies have been proposed to investigate
tumoral pH [50]. Together with the MRI probes, magnetic reso-
nance based contrast agents provide the widest array of probes for
the intra- and extracellular microenvironment of normal and dis-
eased tissues [51].
Finally, an ample collection of optical probes emitting fluores-
cence, luminescence or near-infrared radiation, have been described
[52–54]. These probes are limited by the reduced penetration in
tissues of the optical wavelengths, but are well endowed to image
small animals as mice or cells, using the new generation of CCD
(Charge Coupled Devices) cameras and optical scanners. The per-
formance of these probes is supported by their excellent selectivity
to respond to very specific molecular processes, including protease
activation [53], angiogenesis [55], or inflammation [56], among
others.
4 Nanometric Contrast Agents
The advent of Nanotechnology has provided a collection of new

tools and formulations that may overcome the limitations of the
first generation of sub-nanometric contrast agents. Basically, the
use of nanoparticles, decorated with vectorial molecules like anti-
bodies or peptides may increase significantly the selectivity of con-
trast agents for specific molecular targets within the tissue.
Moreover, nanoparticles may allow combining several
sub-nanometric probes for different imaging modalities over the
same nanometric platform, even adding a therapeutic molecule to
the assembly. These novel formulations are expected to decrease the
dose of administered contrast required for successful imaging, since
most of the nonselective contrast agents currently administered,
target undesired regions of the subject or become eliminated.
Nanotechnology has opened then, a new era in the contrast agents

field, optimizing the dose and detection efficacy of the imaging
agents as well as complementing the diagnostic and therapeutic
potentials, producing the new generations of theranostic agents.
Figure 2 introduces some of the most commonly used nano-
particle formulations. Solid nanoparticles (or powders) may be
produced of varying sizes, by the spray-drying method, allowing
for the encapsulation of drugs and contrast agents for oral delivery,
in most cases. To this end normally, poly(lactic-co-glycolic) acid
(PLGA) nanoparticles are used for oral or rectal administration
[57]. Superparamagnetic iron oxide nanoparticles were among
the first contrast agents with nanometric sizes proposed to increase
the relaxing capacity of the paramagnetic chelates [58–60]. These
nanoparticles contain a magnetite core (Fe3O4), covered most
frequently by a derivatized dextran, acrylic acid polymer or even
lipid coat. The particles are prepared by alkaline precipitation of
mixtures of Fe3+ and Fe2+ in the presence of stabilizing agents as
dextran or oleic acid. These depict enormous molecular relaxivity
values, as compared to paramagnetic Gd(III) chelates, allowing for
a significant increase in the sensitivity for MRI detection. This is
due to the fact that the cooperative alignment of the magnetic
moments from the iron ions in the superparamagnetic nanoparti-
cles results in significantly larger magnetic moments than the addi-
tive alignment of the paramagnetic Gd(III) moments.
Superparamagnetic behavior results mainly in T2 and T2* enhance-
ment, in contrast with the paramagnetic T1 enhancement, of the
Gd(III) chelates.
Lipid nanoparticles, liposomes [61, 62] and micelles [63],
provide excellent platforms to administer combinations of contrast
agents and therapeutic principles, in intravenous administrations.
Quantum dots depict improved sensitivity for fluorescence and
near-infrared imaging of single particle tracking and theranostic
combinations with drug or gene delivery systems, both in vitro
and in vivo [64, 65]. Finally, the use of graphene structures has
recently been added to the array of nanotechnology platforms, to
administer contrast agents, combining imaging probes and thera-
peutic agents. Gd(III) atoms have been included in fullerenes and
carbon nanotubes [66–70]. However, the magnetic properties of
these arrangements and their use as contrast agents remains, nowa-
days, insufficiently characterized.
The following paragraphs provide some illustrative examples
from advanced multimodal contrast agents formulated using nano-
technology methods, taken from our own laboratories.
4.1 Magnetoli- Liposomes have been proposed previously as novel nanotechnology

posomes formulations to improve drug delivery to a variety of inflammatory
diseases [71, 72]. Targeting of the inflammatory region may be
achieved using either active or passive approaches. Active targeting
Fig. 2 The development of recent nanotechnology approaches has driven the creation of new families of
imaging contrast agents. Classic contrast agents as Gd(III) DOTA are now competing with newer, more
sensitive and more selective, contrast agents based on PLGA nanoparticles (top left), SPIO nanoparticles (top
center), Lipid Nanoparticles (top right), Quantum Dots (bottom left), Dendrimers (bottom center), or Graphene
structures (bottom right)
involves the use of vectorial reagents embedded in the liposomal

membrane that recognize epitopes of the inflamed tissue [73]. Pas-
sive targeting refers to the passive accumulation of the liposomes in
the inflamed regions because of their increased capillary permeabil-
ity and limited clearance [74]. In both cases it becomes difficult to
visualize if the liposomal preparation has arrived to the target tissue
and many times, only indirect measurements of inflammation pro-
vide an index of the anti-inflammatory effect. We therefore pro-
posed to visualize directly and noninvasively the presence of the
drug loaded liposomes in the lesion by including suitable imaging
agents in the liposomal lumen. Briefly, we implemented a successful
protocol to encapsulate ω-3 poly-unsaturated fatty acid ethyl ester
(PUFA-EE) in liposomal preparations containing, in addition,
either the superparamagnetic nanoparticle Nanotex or the
rhodamine-100 dye [75]. These advanced theranostic preparations
maintained the therapeutic potential of free ω-3 PUFA-EE, poten-
tiated with important multimodal imaging capabilities. We demon-
strated then their anti-inflammatory effects in vivo, in animal
models of colonic and oncologic inflammation.
Figure 3a shows results obtained in an in vitro fluorescence
experiment, comparing the fluorescence of empty liposomes (left
track), liposomes containing rhodamine-100 (central track), and
liposome containing rhodamine-100 and Nanotex (right track).
Liposomes containing only rhodamine-100 present more fluores-
cence than liposomes containing both rhodamine-100 and nano-
particles (Fig. 3b), revealing that the acrylic acid coated
Fig. 3 Fluorescence imaging of liposomes. In vitro (a, b) and in vivo (c) fluorescence images of liposomes
loaded with nanoparticles and rhodamine-100 as acquired with the IVIS-Lumina camera. (a) Visible picture of
the 96 well plate used as phantom, showing the columns of wells loaded with decreasing (top to bottom)
concentrations of empty liposomes (left column), liposomes loaded with rhodamine-100 (central column) and
liposomes loaded with rhodamine-100 and Nanotex (right column). (b) Fluorescence image of the same
phantom. Note the larger fluorescence of the liposomes containing rhodamine-100 only as compared to those
containing rhodamine-100 and Nanotex. (c) In vivo images reveal the presence of the liposomal preparations
injected i.p., arrow left: liposomes containing rhodamine-100 and Nanotex, arrow right: empty liposomes.
Taken from [75]. Reproduced with permission of the publisher
nanoparticle may quench partially the rhodamine-100 fluores-

cence. Nevertheless, it became possible to image successfully
in vivo the fluorescence of subcutaneously injected magnetolipo-
somes (Fig. 3c, left arrow), or liposomes containing rhodamine-
100 only (Fig. 4c, right arrow).
We further investigated the anti-inflammatory effects of our
liposomal preparations in the DSS (dextran sulfate sodium salt)
model of colonic inflammation, using 18F FDG and PET-CT tech-
niques [76]. Figure 4 shows representative results of four different
animal groups including a control without any treatment (Fig. 4a),
an animal receiving a saline placebo treatment (Fig. 4b), and ani-
mals receiving magnetoliposomal preparations containing Nanotex
with (Fig. 4c) or without PUFA-EE (Fig. 4d). Colonic 18FDG
uptake was observed in all animals. The uptake appeared to be
higher in the animals without liposomal treatment (Fig. 4a and b,
full arrows), as expected from untreated colonic inflammation. In
contrast, animals receiving magnetoliposomes loaded (Fig. 4c) or
not (Fig. 4d) with PUFA-EE depicted less FDG uptake (Fig. 4c,
dotted arrow), as it would be expected from the efficient anti-
Fig. 4 PET-CT/FDG images of mice subjected to 5 days of DSS administration untreated (a), treated with
placebo (b) or liposomal suspensions containing Nanotex (c) with or without ω-3 PUFA-EE (d) [76]. Note that
FDG uptake ( full arrows) is higher in the untreated animals and those treated with placebo, than in those
treated with magnetoliposomes (dotted arrows), suggesting a higher inflammatory response in untreated
animals
Fig. 5 Effect of magnetoliposomal preparations containing (bottom) or not (top) ω-3 PUFA-EE on the time
course of glioma development after implantation of C6 cells in the mouse brain. Left panels (day 0), central
panels (day 3), right panels (day 6). Orange arrow shows accumulation of Nanotex nanoparticle. Taken from
[75]. Reproduced with permission of the publisher
inflammatory effect of PUFA-EE and a reactive oxygen species

scavenging effect of the magnetoliposome (Fig. 4d, dotted arrow).
We then examined the effects of magnetoliposomes containing
or not ω-3 PUFA-EE in the inflammation associated to glioma
growth, using the C6 glioma model. Figure 5 shows representative
MRI T2-weighted images of glioma evolution in a mouse treated
with magnetoliposomes containing (bottom panels) or not (upper
panels) ω-3 PUFA-EE. Mice receiving magnetoliposomes without
ω-3 PUFA-EE developed tumors to the same extent than controls
receiving no treatment. However, mice receiving magnetolipo-
somes containing ω-3 PUFA-EE decreased notably the rate of
glioma growth, and even increased the regression rate of the
implanted glioma, suggesting that the anti-inflammatory effects
markedly interfered with glioma growth.
4.2 pHe-Sensitive Tumors are known to present an acidic extracellular pH microenvi-

Liposomes ronment, which is not present in normal tissues, thus providing a
highly specific target for pH driven therapies [77–79]. In collabo-
ration with the University of Groningen (NL), we engineered
stealth-liposome preparations with a pH-sensitive ion channel
Fig. 6 Image guided drug release in vivo from pH-sensitive liposomes into a C6 glioma implanted in the mouse
brain. (a) The C6 tumor-bearing animal was anesthetized, positioned in the magnet isocenter, and 0.4 mL of
1 M ISUCA (blue solution) was injected i.p [49]. (b) Successive multivoxel spectroscopy grids acquired (c), the
Henderson-Hasselbalch calibration curve was generated (d), and the measured ISUCA chemical shift in every
voxel was transformed into an extracellular pH value generating a pHe map (e). After ISUCA was cleared from
the tumor and its resonances became no longer detectable, T1-weighted baseline images were acquired (f).
The pH-sensitive or pH-insensitive liposomal preparations containing Gd-DTPA (red solution) were injected
into the tail vein (g). T1-weighted images acquired for the next 60 min and relative changes of T1-weighted
images in signal intensity (SI) were calculated in a pixel-by-pixel manner using the baseline images as
reference (h). The maps representing the mean change in SI (ΔSImean) were calculated by averaging the
changes occurring in SI during 60 min after liposome administration and presented overlaid with the
anatomical T1 image. Taken from [80]. Reproduced with permission of the publisher
able to open under acidic conditions, thus releasing the liposomal

load only in the tumor microenvironment [80]. Briefly, we engi-
neered mechanosensitive channel protein with an imidazolic sen-
sor, to open the nanovalve under acidic conditions and loaded the
engineered liposomes with 0,3 M Gd(III)DTPA as imaging agent.
Using magnetic resonance spectroscopy and imaging, we showed
that these engineered liposomes could detect the mildly acidic pH
of the tumor microenvironment with 0.2 pH unit precision, releas-
ing their intraluminal content into C6 glioma tumors selectively,
in vivo.
Briefly, we obtained pHe maps of the glioma model before
administering the liposomal preparation [48, 49], an followed by
MRI the release of intraliposomal Gd(III)DTPA into every voxel of
the tumor, through the corresponding enhancement in T1-
weighted MRI signal intensities (Fig. 6). Voxel-by-voxel correla-
tions between the pHe values and the Gd(III)DTPA signal
enhancement profiles showed that liposomes functionalized with
the engineered ion-channel pH sensor can sense the mildly acidic

pHe (6.6 < pHe < 7.0) of implanted C6 glioma tumors occurring
in vivo. As a response, they released their Gd (III)DTPA content
locally and selectively in the extracellular space of the tumor,
increasing the intensity of the corresponding T1-weighted MRI
signals in vivo. MRI signal enhancement was, however, not
observed when the liposomes administered contained the same
amount of encapsulated Gd(III)DTPA and a pH-insensitive nano-
valve, revealing no appreciable release of the intraluminal imaging
probe to the in vivo tumor under these conditions. Together, these
results showed that the pH-sensitized nanovalve may provide a
suitable mechanism for pH-triggered drug release from stealth
liposomes into the extracellular space of tumors in vivo.
5 Conclusion
In summary, the present review provided an overview of the most

common multimodal imaging approaches implemented in biomed-
ical imaging, highlighting the role of sub-nanometric and nano-
metric contrast agents improving image sensitivity and selectivity.
In particular, we provided illustrative examples from our own
laboratories on the theranostic potential of magnetoliposomes
loaded PUFA-EE in animal models of inflammatory diseases, and
of image guided drug release to in vivo tumors from stealth lipo-
somes engineered with a pH-sensitive nanovalve.
Acknowledgements
Authors are indebted to Dr. Pilar López-Larrubia CSIC for the

careful reading of the manuscript and the valuable comments
provided, Mrs. Teresa Navarro CSIC for granting access to CSIC
small animal MRI facilities and expert technical assistance during
the MRI acquisitions, and to Mr. Javier Pérez CSIC for professional
drafting of the illustrations.
Financial statement: This work was supported in part by grants
SAF2014-53739-R and S2010/BMD-2349 to SC and grant
CTQ2013-47669-R to PB. DC held postdoctoral contracts from
Consejo Superior de Investigaciones Cientı́ficas CSIC. Funding
sources were not involved in the design of the study, in the collec-
tion, analysis and interpretation of data, in the writing of the report
nor in the decision to submit the article for publication.
References
1. Suetens P (2009) Fundamentals of medical 2. Medical imaging: technology ad applications
imaging, 2nd edn. Cambridge University (2013). CRC Press
Press, Cambridge
3. Textbook of contrast media (1999). CRC 17. Lauterbur PC (1973) Image formation by
Press, Boca Raton, FL induced local interactions: examples employing
4. Contrast agents I. Magnetic resonance imaging nuclear magnetic resonance. Nature 242
(2002) Topics in current chemistry, vol 221. (5394):190–191
Springer, Berlin 18. Farrell C, Wallace JD, Mansfield CM (1971)
5. Contrast agents II: optical, ultrasound, x-ray The use of thermography in detection of meta-
and radiopharmaceutical imaging (2002) static breast cancer. Am J Roentgenol 111
Topics in current chemistry, vol 222. Springer, (1):148–152
Berlin 19. Rontgen WC (1896) On a new kind of rays.
6. Contrast agents III: radiopharmaceuticals: Science 3(59):227–231. https://doi.org/10.
from diagnosis to therapeutics (2002) Topics 1126/science.3.59.227
in current chemistry, vol 252. Springer, Berlin 20. Brownell GL, Sweet WH (1953) Localization
7. Key J, Leary JF (2014) Nanoparticles for mul- of brain tumors with positron emitters. Nucle-
timodal in vivo imaging in nanomedicine. Int J onics 11(11):40–45
Nanomedicine 9:711–726. https://doi.org/ 21. Rudin M, Weissleder R (2003) Molecular
10.2147/IJN.S53717 imaging in drug discovery and development.
8. Heidt T, Nahrendorf M (2013) Multimodal Nat Rev Drug Discov 2(2):123–131. https://
iron oxide nanoparticles for hybrid biomedical doi.org/10.1038/nrd1007
imaging. NMR Biomed 26(7):756–765. 22. Bushberg JT, Boone JM (2011) The essential
https://doi.org/10.1002/nbm.2872 physics of medical imaging. Lippincott Wil-
9. Kim J, Piao Y, Hyeon T (2009) Multifunc- liams & Wilkins, Philadelphia, PA
tional nanostructured materials for multimodal 23. Weissleder R, Mahmood U (2001) Molecular
imaging, and simultaneous imaging and ther- imaging. Radiology 219(2):316–333. https://
apy. Chem Soc Rev 38(2):372–390. https:// doi.org/10.1148/radiology.219.2.
doi.org/10.1039/b709883a r01ma19316
10. Lee DE, Koo H, Sun IC, Ryu JH, Kim K, 24. Hounsfield GN (1973) Computerized trans-
Kwon IC (2012) Multifunctional nanoparticles verse axial scanning (tomography): Part
for multimodal imaging and theragnosis. 1. Description of system. Br J Radiol 46
Chem Soc Rev 41(7):2656–2672. https:// (552):1016–1022
doi.org/10.1039/c2cs15261d 25. Hutton BF (2011) Recent advances in iterative
11. Lee N, Yoo D, Ling D, Cho MH, Hyeon T, reconstruction for clinical SPECT/PET and
Cheon J (2015) Iron oxide based nanoparticles CT. Acta Oncol 50(6):851–858
for multimodal imaging and magnetorespon- 26. Antoch G, Freudenberg LS, Beyer T,
sive therapy. Chem Rev 115 Bockisch A, Debatin JF (2004) To enhance or
(19):10637–10689. https://doi.org/10. not to enhance? 18F-FDG and CT contrast
1021/acs.chemrev.5b00112 agents in dual-modality 18F-FDG PET/CT. J
12. Ding H, Wu F, Nair MP (2013) Image-guided Nucl Med 45(Suppl 1):56S–65S
drug delivery to the brain using nanotechnol- 27. Caravan P, Ellison JJ, McMurry TJ, Lauffer RB
ogy. Drug Discov Today 18 (1999) Gadolinium (III) chelates as MRI con-
(21–22):1074–1080. https://doi.org/10. trast agents: structure, dynamics, and applica-
1016/j.drudis.2013.06.010 tions. Chem Rev 99(9):2293–2352
13. Ding H, Wu F (2012) Image guided biodistri- 28. Pacheco-Torres J, Calle D, Lizarbe B, Negri V,
bution of drugs and drug delivery. Theranos- Ubide C, Fayos R, Lopez Larrubia P,
tics 2(11):1037–1039. https://doi.org/10. Ballesteros P, Cerdan S (2011) Environmen-
7150/thno.5321 tally sensitive paramagnetic and diamagnetic
14. Lammers T, Kiessling F, Hennink WE, Storm contrast agents for nuclear magnetic resonance
G (2010) Nanotheranostics and image-guided imaging and spectroscopy. Curr Top Med
drug delivery: current concepts and future Chem 11(1):115–130
directions. Mol Pharm 7(6):1899–1912 29. Elsinga PH, Dierckx RA (2014) Small mole-
15. Yu MK, Park J, Jon S (2012) Magnetic nano- cule PET-radiopharmaceuticals. Curr Pharm
particles and their applications in image-guided Des 20:2268–2274
drug delivery. Drug Deliv Transl Res 2 30. Cai J, Li F (2013) Single-photon emission
(1):3–21. https://doi.org/10.1007/s13346- computed tomography tracers for predicting
011-0049-8 and monitoring cancer therapy. Curr Pharm
16. Webb S (1988) The physics of medical imag- Biotechnol 14(7):693–707
ing. CRC Press, Boca Ratón, FL
31. Thomsen HS, Morcos SK (2000) Radio- 42. Hancu I, Dixon WT, Woods M, Vinogradov E,
graphic contrast media. BJU Int 86(Suppl Sherry AD, Lenkinski RE (2010) CEST and
1):1–10 PARACEST MR contrast agents. Acta Radiol
32. Pavel DG, Zimmer M, Patterson VN (1977) In 51(8):910–923. https://doi.org/10.3109/
vivo labeling of red blood cells with 99mTc: a 02841851.2010.502126
new approach to blood pool visualization. J 43. Sun PZ, Sorensen AG (2008) Imaging pH
Nucl Med 18(3):305–308 using the chemical exchange saturation transfer
33. Hahn MA, Singh AK, Sharma P, Brown SC, (CEST) MRI: correction of concomitant RF
Moudgil BM (2011) Nanoparticles as contrast irradiation effects to quantify CEST MRI for
agents for in-vivo bioimaging: current status chemical exchange rate and pH. Magn Reson
and future perspectives. Anal Bioanal Chem Med 60(2):390–397
399(1):3–27 44. McRae R, Bagchi P, Sumalekshmy S, Fahrni CJ
34. Pimlott SL, Sutherland A (2011) Molecular (2009) In situ imaging of metals in cells and
tracers for the PET and SPECT imaging of tissues. Chem Rev 109(10):4780–4827.
disease. Chem Soc Rev 40(1):149–162. https://doi.org/10.1021/cr900223a
https://doi.org/10.1039/b922628c 45. Zhang S, Malloy CR, Sherry AD (2005) MRI
35. Heinle SK, Noblin J, Goree-Best P, Mello A, thermometry based on PARACEST agents. J
Ravad G, Mull S, Mammen P, Grayburn PA Am Chem Soc 127(50):17572–17573.
(2000) Assessment of myocardial perfusion by https://doi.org/10.1021/ja053799t
harmonic power doppler imaging at rest and 46. Portman MA, Lassen NA, Cooper TG, Sills
during adenosine stress comparison with AM, Potchen EJ (1991) Intra- and extracellu-
99mTc-Sestamibi SPECT imaging. Circulation lar pH of the brain in vivo studied by
102(1):55–60 31P-NMR during hyper- and hypocapnia. J
36. Urtasun RC, Parliament MB, McEwan AJ, Appl Physiol 71(6):2168–2172
Mercer JR, Mannan RH, Wiebe LI, Morin C, 47. Fisher MJ, Dillon PF (1987) Phenylphospho-
Chapman JD (1996) Measurement of hypoxia nate: a 31P-NMR indicator of extracellular pH
in human tumours by non-invasive spect imag- and volume in the isolated perfused rabbit
ing of iodoazomycin arabinoside. Br J Cancer bladder. Circ Res 60(4):472–477
Suppl 27:S209–S212 48. Garcia-Martin ML, Herigault G, Remy C,
37. Leitha T, Glaser C, Pruckmayer M, Rasse M, Farion R, Ballesteros P, Coles JA, Cerdan S,
Millesi W, Lang S, Nasel C, Backfrieder W, Ziegler A (2001) Mapping extracellular pH in
Kainberger F (1998) Technetium-99m-MIBI rat brain gliomas in vivo by 1H magnetic reso-
in primary and recurrent head and neck nance spectroscopic imaging: comparison with
tumors: contribution of bone SPECT image maps of metabolites. Cancer Res 61
fusion. J Nucl Med 39(7):1166–1171 (17):6524–6531
38. Tharp K, Israel O, Hausmann J, Bettman L, 49. Provent P, Benito M, Hiba B, Farion R, Lopez-
Martin W, Daitzchman M, Sandler M, Delbeke Larrubia P, Ballesteros P, Remy C,
D (2004) Impact of 131I-SPECT/CT images Segebarth C, Cerdan S, Coles JA, Garcia-
obtained with an integrated system in the Martin ML (2007) Serial in vivo spectroscopic
follow-up of patients with thyroid carcinoma. nuclear magnetic resonance imaging of lactate
Eur J Nucl Med Mol Imaging 31 and extracellular pH in rat gliomas shows
(10):1435–1442 redistribution of protons away from sites of
39. Fukuyama H, Ouchi Y, Matsuzaki S, glycolysis. Cancer Res 67(16):7638–7645.
Nagahama Y, Yamauchi H, Ogawa M, https://doi.org/10.1158/0008-5472.CAN-
Kimura J, Shibasaki H (1997) Brain functional 06-3459
activity during gait in normal subjects: a 50. Gallagher FA, Kettunen MI, Day SE, DE H,
SPECT study. Neurosci Lett 228(3):183–186 Ardenkjaer-Larsen JH, Zandt R, Jensen PR,
40. Strauss LG (1996) Fluorine-18 deoxyglucose Karlsson M, Golman K, Lerche MH, Brindle
and false-positive results: a major problem in KM (2008) Magnetic resonance imaging of pH
the diagnostics of oncological patients. Eur J in vivo using hyperpolarized 13C-labelled
Nucl Med 23(10):1409–1415 bicarbonate. Nature 453(7197):940–943.
41. Aime S, Barge A, Delli Castelli D, Fedeli F, https://doi.org/10.1038/nature07017
Mortillaro A, Nielsen FU, Terreno E (2002) 51. Pacheco-Torres J, Calle D, Lizarbe B, Negri V,
Paramagnetic lanthanide (III) complexes as Ubide C, Fayos R, Lopez Larrubia P,
pH-sensitive chemical exchange saturation Ballesteros P, Cerdan S (2011) Environmen-
transfer (CEST) contrast agents for MRI appli- tally sensitive paramagnetic and diamagnetic
cations. Magn Reson Med 47(4):639–648 contrast agents for nuclear magnetic resonance
imaging and spectroscopy. Curr Top Med Roentgenol 186(2):300–307. https://doi.

Chem 11(1):115–130 org/10.2214/AJR.05.0523
52. Frangioni JV (2003) In vivo near-infrared fluo- 63. Torchilin VP (2002) PEG-based micelles as
rescence imaging. Curr Opin Chem Biol 7 carriers of contrast agents for different imaging
(5):626–634 modalities. Adv Drug Deliv Rev 54
53. Weissleder R, Tung CH, Mahmood U, Bogda- (2):235–252
nov A Jr (1999) In vivo imaging of tumors with 64. Wegner KD, Hildebrandt N (2015) Quantum
protease-activated near-infrared fluorescent dots: bright and versatile in vitro and in vivo
probes. Nat Biotechnol 17(4):375–378. fluorescence imaging biosensors. Chem Soc
https://doi.org/10.1038/7933 Rev 44(14):4792–4834. https://doi.org/10.
54. Chan WC, Maxwell DJ, Gao X, Bailey RE, 1039/c4cs00532e
Han M, Nie S (2002) Luminescent quantum 65. Michalet X, Pinaud FF, Bentolila LA, Tsay JM,
dots for multiplexed biological detection and Doose S, Li JJ, Sundaresan G, Wu AM, Gamb-
imaging. Curr Opin Biotechnol 13(1):40–46 hir SS, Weiss S (2005) Quantum dots for live
55. Yang M, Baranov E, Li XM, Wang JW, Jiang P, cells, in vivo imaging, and diagnostics. Science
Li L, Moossa AR, Penman S, Hoffman RM 307(5709):538–544. https://doi.org/10.
(2001) Whole-body and intravital optical 1126/science.1104274
imaging of angiogenesis in orthotopically 66. Wang L, Zhu X, Tang X, Wu C, Zhou Z,
implanted tumors. Proc Natl Acad Sci U S A Sun C, Deng SL, Ai H, Gao J (2015) A multi-
98(5):2616–2621. https://doi.org/10.1073/ ple gadolinium complex decorated fullerene as
pnas.051626698 a highly sensitive T(1) contrast agent. Chem
56. Sedghi S, Fields JZ, Klamut M, Urban G, Commun 51(21):4390–4393. https://doi.
Durkin M, Winship D, Fretland D, Olyaee M, org/10.1039/c5cc00285k
Keshavarzian A (1993) Increased production 67. Braun K, Dunsch L, Pipkorn R, Bock M,
of luminol enhanced chemiluminescence by Baeuerle T, Yang S, Waldeck W, Wiessler M
the inflamed colonic mucosa in patients with (2010) Gain of a 500-fold sensitivity on an
ulcerative colitis. Gut 34(9):1191–1197 intravital MR contrast agent based on an endo-
57. Barichello JM, Morishita M, Takayama K, hedral gadolinium-cluster-fullerene-conjugate:
Nagai T (1999) Encapsulation of hydrophilic a new chance in cancer diagnostics. Int J Med
and lipophilic drugs in PLGA nanoparticles by Sci 7(3):136–146
the nanoprecipitation method. Drug Dev Ind 68. Yang H, Lu C, Liu Z, Jin H, Che Y, Olmstead
Pharm 25(4):471–476. https://doi.org/10. MM, Balch AL (2008) Detection of a family of
1081/DDC-100102197 gadolinium-containing endohedral fullerenes
58. Na HB, Song IC, Hyeon T (2009) Inorganic and the isolation and crystallographic charac-
nanoparticles for MRI contrast agents. Adv terization of one member as a metal-carbide
Mater 21(21):2133–2148 encapsulated inside a large fullerene cage. J
59. Bonnemain B (1998) Superparamagnetic Am Chem Soc 130(51):17296–17300.
agents in magnetic resonance imaging: physi- https://doi.org/10.1021/ja8078303
cochemical characteristics and clinical applica- 69. Avti PK, Talukdar Y, Sirotkin MV, Shroyer KR,
tions. A review. J Drug Target 6(3):167–174. Sitharaman B (2013) Toward single-walled car-
https://doi.org/10.3109/ bon nanotube-gadolinium complex as
10611869808997890 advanced MRI contrast agents: pharmacody-
60. Liu F, Laurent S, Fattahi H, Vander Elst L, namics and global genomic response in small
Muller RN (2011) Superparamagnetic nano- animals. J Biomed Mater Res B Appl Biomater
systems based on iron oxide nanoparticles for 101(6):1039–1049. https://doi.org/10.
biomedical imaging. Nanomedicine 6 1002/jbm.b.32914
(3):519–528. https://doi.org/10.2217/nnm. 70. Negri V, Cerpa A, Lopez-Larrubia P, Nieto-
11.16 Charques L, Cerdan S, Ballesteros P (2010)
61. Grant CW, Karlik S, Florio E (1989) A liposo- Nanotubular paramagnetic probes as contrast
mal MRI contrast agent: phosphatid agents for magnetic resonance imaging based
ylethanolamine-DTPA. Magn Reson Med 11 on the diffusion tensor. Angew Chem 49
(2):236–243 (10):1813–1815. https://doi.org/10.1002/
anie.200906415
62. Mukundan S Jr, Ghaghada KB, Badea CT, Kao
CY, Hedlund LW, Provenzale JM, Johnson 71. Lian T, Ho RJ (2001) Trends and develop-
GA, Chen E, Bellamkonda RV, Annapragada ments in liposome drug delivery systems. J
A (2006) A liposomal nanoscale contrast agent Pharm Sci 90(6):667–680
for preclinical CT in mice. AJR Am J
72. Torchilin VP (2005) Recent advances with (2009) Bicarbonate increases tumor pH and
liposomes as pharmaceutical carriers. Nat Rev inhibits spontaneous metastases. Cancer Res
Drug Discov 4(2):145–160 69(6):2260–2268. https://doi.org/10.1158/
73. Hilgenbrink AR, Low PS (2005) Folate 0008-5472.CAN-07-5575
receptor-mediated drug targeting: from thera- 78. Raghunand N, Mahoney BP, Gillies RJ (2003)
peutics to diagnostics. J Pharm Sci 94 Tumor acidity, ion trapping and chemothera-
(10):2135–2146 peutics. II. pH-dependent partition coeffi-
74. Danhier F, Feron O, Preat V (2010) To exploit cients predict importance of ion trapping on
the tumor microenvironment: passive and pharmacokinetics of weakly basic chemothera-
active tumor targeting of nanocarriers for peutic agents. Biochem Pharmacol 66
anti-cancer drug delivery. J Control Release (7):1219–1229
148(2):135–146 79. Mahoney BP, Raghunand N, Baggett B, Gillies
75. Calle D, Negri V, Ballesteros P, Cerdan S RJ (2003) Tumor acidity, ion trapping and
(2015) Magnetoliposomes loaded with poly- chemotherapeutics. I. Acid pH affects the dis-
unsaturated fatty acids as novel theranostic tribution of chemotherapeutic agents in vitro.
anti-inflammatory formulations. Theranostics Biochem Pharmacol 66(7):1207–1218
5(5):489–503. https://doi.org/10.7150/ 80. Pacheco-Torres J, Mukherjee N, Walko M,
thno.10069 Lopez-Larrubia P, Ballesteros P, Cerdan S,
76. Calle D (2014) Novel contrast agents for mul- Kocer A (2015) Image guided drug release
timodal biomedical imaging based in nano- from pH-sensitive Ion channel-functionalized
technology. Dissertation, Universidad stealth liposomes into an in vivo glioblastoma
Autónoma de Madrid model. Nanomedicine 11(6):1345–1354.
77. Robey IF, Baggett BK, Kirkpatrick ND, Roe https://doi.org/10.1016/j.nano.2015.03.
DJ, Dosescu J, Sloane BF, Hashim AI, Morse 014
DL, Raghunand N, Gatenby RA, Gillies RJ
INDEX
A continuous arterial spin labeling (CASL) .........60, 63,

65, 67, 68
ActiveAx......................................................................... 360 labeling efficiency .................................. 60, 65, 67, 68
Active targeting ................. 236, 237, 245, 249, 410, 447 labeling plane.......................................................60, 65
Adiabatic full passage (AFP)................................ 182, 337
pseudo continuous arterial spin labeling
Adipose tissue imaging (pCASL)......................................................... 60
abdominal fat.................................................. 260, 267 pulsed arterial spin labeling (PASL)......................... 60
Dixon imaging ............................................... 262, 264
transit time...........................................................67, 68
fat-water imaging .................................................... 264 Atlas ...................................................................... 361–363
in-phase........................................................... 263, 264 AxCaliber ....................................................................... 360
intra-abdominal fat (IAT) ....................................... 267
Axial diffusivity ..................................................... 106, 112
out-of-phase ................................................... 263, 264
subcutaneous fat (SAT) ................262, 264, 265, 267 B
visceral adipose tissue (VAT) ......................... 262, 267
Alzheimer’s disease ......................................................... 36 B0
Analysis of functional neuroimages (AFNI) ................ 122 B0 inhomogeneities................................................... 32
Anesthesia B0 map ...........................................158, 163, 192, 195
alfaxalone ................................................................. 429 field inhomogeneities..........................................27, 32
barbiturate ............................................................... 429 fieldmap ................................................................... 366
bupivacaine .............................................................. 126 B1 homogeneity................................................... 108, 156
chloralhydrate.......................................................... 430 Bloch equations............................................................... 16
chloralose ................................................................. 430 Blood–brain barrier (BBB) ...................... 24, 42, 52, 352,
fentanyl/fluanisone ................................................. 429 355, 380, 395, 396, 404–406
halothane ................................................................. 221 Blood flow ........................ 42, 45, 59–69, 118, 145, 221,
isoflurane ........................ 47, 48, 60, 66, 68, 95, 108, 270, 302, 305, 309, 378, 379, 384, 385, 387,
109, 132, 139, 140, 144, 156, 176, 178, 184, 424, 434
192, 210, 212, 238, 241, 251–253, 260, 287, Blood oxygenation level dependent (BOLD) ....... 34, 42,
302, 319, 321, 333, 335, 336, 350, 352, 353, 118–121, 123–125, 131, 206, 221, 299, 300,
383, 411, 412, 434 302, 303, 306, 308, 424
ketamine ....................... 65, 221, 261, 301, 302, 333, Bolus
335, 429 arrival time ................................................................. 85
propofol ................................................. 301, 302, 429 tracking ......................................................... 42–45, 55
sevoflurane ............................................................... 434 Brownian
tribromoethanol ...................................................... 430 diffusion..................................................................... 89
urethane ................................................................... 123 motion ............................................................ 103, 135
Angiography ................................................ 378, 386, 388 B-value ....................... 96, 100, 104, 110, 113, 114, 137,
Animal monitoring............. 46, 47, 61, 91, 95, 108, 109, 140, 145, 146, 290, 354, 368
138, 140, 156, 157, 192, 209, 239, 271, 273,
C
274, 287, 319, 333, 350, 398–400, 411, 412
Anisotropy .....................90, 92, 106, 107, 114, 292, 354 13
C
Apparent diffusion coefficient (ADC) ...........3, 4, 90, 92, 13
C enriched substrates........................................... 152
95, 96, 104–106, 110, 112, 114, 136, 292, 13
C glucose........................... 170, 175, 177, 182, 184
354, 379, 381–383, 388 Cancer................................... 6, 104, 107, 155, 156, 189,
Arterial input function (AIF) ..........................44, 54, 396 259, 297, 316, 331–343, 442
Arterial spin labeling (ASL) Cardiac MRI
ASLtbx ....................................................................... 63 cine..........................................................273, 275–279
vol. 1718, https://doi.org/10.1007/978-1-4939-7531-0, © Springer Science+Business Media, LLC 2018
459
PRECLINICAL MRI: METHODS AND PROTOCOLS
460 Index
Cardiac MRI (cont.) coefficient ............................ 3, 90, 92, 104–106, 136,
CMR42........................................................... 272, 280 137, 141, 143, 144, 253, 354, 379, 383
ejection fraction (EF).............................................. 279 gradient..................91, 354, 360, 361, 368, 379, 386
end-diastolic volume (EDV) .................................. 279 time ................................................................... 91, 103
endo-cardiac border ....................................... 279, 280 Diffusion tensor imaging (DTI) ......................... 379, 388
end-systolic volume (ESV) ..................................... 279 DtiStudio ........................................................ 108, 111
epi-cardiac border .......................................... 279, 280 Diffusion weighted imaging (DWI) .................... 90, 136,
intragate (Ig) ........................................................... 271 348, 349, 379
long axes (LAX) .......... 271, 273, 275, 276, 278, 279 Diffusivity .................................................... 105, 106, 112
short axes (SAX)............................ 273, 275–278, 280 Directional-encoded-color (DEC) maps ............ 107, 294
Cardiovascular diseases (CVD) .................................... 269 Dynamic contrast enhanced (DCE)........... 301, 387, 396
Cardiovascular system ................................. 431, 433, 434 Tofts model .................................................. 73, 81–83
Carotid artery ................................................................ 384 two-compartments exchange model (2CXM) ....... 73,
Carr Purcell Meiboom Gill (CMPG)..............26, 28, 337 81, 83
Central nervous system (CNS).................... 35, 347, 356, Dynamic susceptibility contrast (DSC)....................42, 59
359, 360, 364, 365, 395, 396 Dynamic T2.......................................................... 412–415
Cerebral perfusion
cerebral blood flow (CBF)........................... 43, 45, 55 E
cerebral blood volume (CBV) ............................43, 55 Echo planar imaging (EPI)...................... 43, 63, 96, 110,
mean transit time (MTT) ............................ 43, 45, 55 122, 301, 353
Cerebrospinal fluid (CSF) ......................... 23, 24, 30, 31, Echo time (TE) ............................... 22, 63, 91, 130, 153,
54, 92, 200, 299, 356, 360, 380, 389, 390
158, 181, 261, 271, 290, 301, 337, 354, 398,
Chelate............................... 24, 25, 47, 55, 444, 445, 447 412, 413
Chemical exchange saturation transfer (CEST)3, 25, 378, Eddy-currents correction (ECC) ................................. 160
446
Edema ................................... 4, 23, 24, 31, 35, 269, 349,
Chemical shift. 7, 8, 114, 151, 158, 159, 163, 171, 189, 356, 359, 379–381, 383, 385, 388
237, 242, 243, 246, 247, 255, 260, 264, 337, Eigenvalues ........................................................... 105, 369
378, 446, 452
Eigenvectors ................................................ 107, 110, 292
Chemical shift displacement error (CSDE)158, 159, 163, Electric-stimulation....................................................... 117
171 Electrocardiogram (ECG) ......................... 108, 109, 114,
Choline (Cho) ............................................. 160, 331, 332 239, 281, 364, 427, 433, 437
Complexone .................................................................... 47
Electrodes ......................... 108, 109, 120–128, 131–133,
Connectivity ........................................117, 131, 349, 361 239, 298, 434
Contrast agent (CA) ..................................................... 321 Embryo .............................. 285, 288, 293, 294, 350, 357
diethylenetriaminepentaacetic acid (DTPA).....24, 47,
Encoding ............................ 9–13, 19, 22, 27, 28, 31, 34,
350, 357, 367, 445 44, 112, 114, 136, 141, 145, 255, 292, 309,
DOTA .......................................................24, 445, 448 353, 354, 357, 358, 361, 368
dysprosium (Dy) .................................................42, 47
EPI ghost....................................................................... 141
gadolinium (Gd) .............................. 24, 47, 350, 357, Ernst angle.............................................................. 34, 227
367, 445, 447, 448 Estimated glucose disposal rate (eGDR) ............ 183, 184
Coupling......................................66, 118, 120, 123, 151,
Ex vivo imaging ............................................................. 348
155, 170, 175, 182
Creatine (Cr) ..................... 155, 160, 162, 190, 191, 332 F
Cryoprobe ......................... 209, 214, 271, 348, 366, 397
19
F ..............................235–257, 298–300, 304, 308, 309
D FA (fractional anisotropy)................. 106, 107, 110, 112,
114, 285, 292, 294, 354, 355, 358–360, 363,
Decoupling ...................................................171–174, 182
364, 369
Demyelination ............................................. 349, 356, 359 Fast diffusion coefficient (Dfast) .......................... 137, 143
Deoxyhemoglobin ................ 34, 35, 299, 305, 322, 387 Fast diffusion phase (FDP).................137, 138, 141, 144
Dexmedetomidine......................................................... 123
Fast low angle shot (FLASH)...........................33, 34, 94,
Diffusion............................v, 3, 4, 41, 45, 135–146, 154, 139, 162, 207, 216, 219, 241, 271, 273, 281,
264, 286, 289–292, 294, 298, 299, 348, 349, 282, 323, 383, 386, 389, 398, 399, 412, 414
353, 354, 358–361, 364, 366, 368, 379, 383, Fastmap........................................... 51, 96, 113, 141, 163
386, 388
PRECLINICAL MRI : METHODS AND PROTOCOLS
Index 461
Fast spin echo (FSE) ........................ 30, 31, 35, 356, 389 Gradient recalled echo (GRE)............................ 206, 228,
Fe(iron)................................... 8, 25, 34, 35, 42, 47, 205, 389, 399, 401, 403, 404, 406
221, 237, 315–317, 325, 349, 356, 411, 447 GRASE ................................ 26, 290, 292, 294, 295, 348
Ferritin ........................................................................... 206 Gray matter (GM)............................. 29, 35, 68, 92, 294,
Fibre orientation distribution(FOD) .................. 358, 369 348, 359, 360
Fibre tracking ................................................................ 365 Gyromagnetic ratio ...................................... 6, 10, 32, 91,
Field of excitation (FOE) .................................... 290, 291 151, 152, 169–172, 264, 308
Field of view (FOV) ...................... 11, 13, 22, 51, 62–65,
110, 113, 130, 133, 157, 190, 192, 194, 197, H
215, 222, 225, 227, 242, 243, 250, 255, 1
H ........................................... 6, 10, 16, 47, 62, 95, 121,
261–263, 271, 273, 290, 301, 322–325, 336, 140, 151, 153, 155, 158, 160, 169–175, 179,
337, 353, 355–359, 365, 383, 386, 387, 398,
180, 182, 189, 207, 209, 210, 220, 222, 235,
399, 401, 412, 413 237, 240–243, 245–250, 252, 255, 299, 300,
Flip angle (FA) .......................... 7, 10, 22, 23, 32–34, 63, 304, 308, 331, 332, 336–339, 446
130, 172, 243, 261, 271, 288, 301, 323, 324,
Hemoglobin ...................................................34, 299, 305
336, 337, 339, 355, 357, 399 Hexafluorobenzene (HFB) .......299, 300, 305, 307, 309
Fluid-attenuated inversion recovery (FLAIR).30, 31, 35, High angular resolution diffusion imaging
380, 389, 390
(HARDI) .................................. 107, 353, 354,
Fluorine (19F) MRI..................................... 242, 244, 252 357–360, 362
Fomblin ......................................222, 350, 351, 357, 367 Huntington’s disease .................................................... 347
Fourier-transform (FT).......................... 7–9, 51, 99, 145, Hybrid echo .................................................................... 26
151, 158, 290, 291, 338, 354, 358
Hypercapnia.........................................427, 428, 433, 436
Fractional anisotropy (FA) ............... 106, 107, 112, 174, Hyperthermia ......................................215, 409, 433, 434
262, 263, 273, 292, 354, 355, 359, 360, 363, Hypothermia ..................................... 210, 239, 424, 427,
364, 369
431, 433, 434, 436
Fractional enrichment (FR) ................................. 172, 176 Hypoxia ............................................. 297, 298, 300, 427,
Free induction decay (FID)............................7, 8, 11, 13, 428, 433, 436, 445
158, 164, 196, 198, 366
Frequency encoding...............10–13, 19, 22, 27, 44, 255 I
FSL software.................................................................. 122
Full-width at half maximum (FWHM) ............... 163, 194 Initial area under the curve (IAUC) .............................. 71
Functional diffusion MRI (fDMRI) ................... 142, 143 Intracerebral microstimulation............................ 119, 120
Functional diffusion weighted imaging (fDWI) ........140, In utero MRI ........................................................ 285–295
142, 145, 146 Inversion recovery (IR) ............................. 16, 26, 29–31,
Functional MRI (fMRI) ............................ 117–133, 205, 264, 301, 356, 366, 380, 389, 399, 404
222, 299, 348, 380, 433, 434 Inversion time (TI) ...........................22, 29–31, 356, 399
Iron ....................................... 8, 34, 35, 42, 47, 206, 237,
G 315–317, 325, 349, 356, 411, 447
Iron oxide nanoparticle ......................................... 25, 447
Gamma-variate function ................................................. 43 Ischemia................................................45, 155, 189, 378,
Ghost artifacts ...................................................... 110, 114
381, 387, 388, 434
Glucose .............................................. 118, 146, 162, 170, Isotopic enrichment (IE).............................................. 176
175–177, 182–184, 259, 331, 395, 431, 445
Glutamate (Glu)................ 155, 160, 162, 181, 378, 380 K
Glutamine (Gln)........................155, 160, 162, 175, 181,
190, 254, 331 K-space .............................11, 13, 14, 27, 31, 56, 95, 96,
Glycerophosphocholine (GPC)................. 155, 160, 162, 99, 114, 140, 141, 145, 190, 194, 218, 225,
331, 332, 340 229, 290, 291, 337, 353, 358
Glycerophosphoethanolamine (GPE)................. 332, 340 Ktrans ................................................................... 73, 82, 83
Glycolysis .............................................................. 175, 445
L
Gradient echo (GE) ............................. 13–15, 18, 23, 25,
32–36, 43, 44, 50, 51, 63, 122, 132, 162, 207, Lactate (Lac)............................................... 155, 162, 175,
242, 252, 261, 288, 292, 294, 300, 306, 308, 177, 184, 190, 331
322, 355, 357, 358, 367, 389, 399, 412, 414 Larmor frequency ....................6, 7, 9, 10, 163, 182, 206
Gradient echo with flow compensation (GEFC) ......... 33, Liposomes............................................442, 447, 449–453
322, 323
462 Index
M Motion artifacts................................. 108, 110, 111, 114,
215, 288, 433
Magnetic field gradient................................. v, 6, 8, 9, 11, Multimodal imaging ...........................442–444, 449, 452
34, 90, 103, 137, 163, 220, 299, 424 Myelin .............................................................90, 107, 349
Magnetic moment.................................... 16, 24, 25, 152, Myocardial ................................................... 269, 279, 434
445, 447 Myo-inositol (myo-Ins) ........................155, 160, 162, 181
Magnetic nanoparticle (MNP) ............................ 409, 414
Magnetic resonance angiography (MRA) N
time-of-flight (TOF)............................................... 386
Magnetic resonance imaging (MRI)............................v, 3, N-acetyl aspartic acid (NAA)..................... 154, 160, 162,
21–36, 41–56, 118, 270–272, 285, 300, 301, 190, 332
315, 332, 347, 378, 382–386, 397, 398, 423, Nanoemulsion ................... 235, 236, 238–241, 251, 254
441, 444 Nanoparticle .................................. 25, 47, 239, 254, 409,
Magnetic resonance spectroscopy (MRS).......... 157, 158, 412, 414, 417, 446–449, 451
162, 164, 165, 171, 174, 191, 198 Neurite orientation dispersion and density imaging
chemical shift imaging (CSI).............................6, 189, (NODDI) ........................................... 354, 360
193, 237, 242, 243, 260, 264, 337, 378 Neurodegeneration ........................................45, 349, 359
13
C MRS Neurodegenerative disease ................................... 59, 297,
distorsionless enhancement by polarization 347–369, 444
transfer (DEPT) ................................. 171, 174 Neurodegenerative disorder ........................................... 35
echo planar spectroscopic imaging (EPSI) ................ 6 Neuronal activation.............................118, 119, 122, 144
heteronuclear MRS .............................. 169, 171, 174, Neurovascular coupling .............................. 118, 120, 123
176, 179, 182, 185 Nuclear Overhauser effect (NOE) ............................... 171
1
H MRS Null point ..................................................................29, 31
LCModel ........................ 157, 158, 162, 164, 165 Nyquist ................................................................. 110, 114
image selected in vivo spectroscopy (ISIS)............ 181
O
J-coupling ................................................................ 181
jMRUI Obesity......................................................... 259, 260, 267
QUEST..................................................... 191, 198 Orthotopic......................... 305, 307, 333, 335, 340–342
localized spectroscopy............................................. 158 Osirix ...................................................................... 75, 362
magnetic resonance spectroscopic imaging Outer volume suppression (OVS) ..................... 159, 164,
(MRSI).......................... 6, 189, 191, 331, 446 181, 194, 196
multivoxel ....................................................... 152, 452 Oximetry......................62, 298, 299, 301, 305–308, 437
31
P MRS .................................................................. 152 Oxygenation .................................. 34, 62, 176, 206, 212,
point resolved spectroscopy (PRESS) ................... 153, 214, 215, 297–309, 424, 428, 433, 434
189, 190 Oxyhemoglobin .......................................... 299, 308, 322
PRESS-MRSI .......................................................... 194
single voxel .............................................152–154, 178 P
spin echo full intensity acquired localized spectroscopy 31
P .............................................. 152, 169–171, 176, 332,
(SPECIAL) ................................ 154, 178, 181
337–340, 343, 446
stimulated echo acquisition mode (STEAM) ....... 154,
Parallel imaging .................................................... 145, 281
163, 181
Paramagnetic compounds............................................... 42
Magnetic susceptibility ............................ 32, 35, 42, 113,
Parametric imaging .................................... 43, 46, 50, 53,
205, 206, 219, 222, 322, 349
94–96, 98, 139, 142, 143
Magnetoliposomes .......................................442, 447–452
Parametric map........................................... 112, 140, 143,
Manganese ...........................................286, 287, 410, 411
144, 363, 364, 380
Manganese-enhanced MRI (MEMRI) ........................ 286
Parkinson’s disease ................................................. 36, 189
Mapshim ......................................... 51, 96, 113, 128, 141
Partial parallel imaging (PPI) ................................ 99, 145
Mean arterial pressure (MAP) ...................................... 434
pCO2 .............................................................................. 433
Mean diffusivity............................................................. 106
Perflurocarbon (PFC) ................................ 235–246, 248,
Medial cerebral artery occlusion (MCAO).................. 381
249, 251–255, 298–300, 303, 305, 309
Metabolic rate ...................................................... 169, 427
Perfusion..............................................v, 4, 34–36, 41–45,
Micro-sized particles of iron oxide (MPIO) ..... 316–320,
47, 50–53, 55, 145, 211, 222, 227, 269, 298,
322–325
302, 319, 350, 364, 365, 378, 379, 434, 445
Molecular MRI (mMRI) ..................................... 315–325
PRECLINICAL MRI : METHODS AND PROTOCOLS
Index 463
Perfusion-diffusion mismatch (PDM) ........................... 45 Receiver gain (RG)........................................................ 164
Perfusion weighted imaging (PWI) ....................... 41, 43, Region of interest (ROI) ............... 47, 98, 143, 303, 405
45, 378, 379 Relaxation ........................................... v, 4, 13–19, 23, 24,
Permeability........................... 4, 316, 395, 396, 406, 449 29–32, 35, 42, 43, 46, 89, 122, 157, 163, 165,
Permeability surface area product .................................. 83 171, 181, 196, 205, 227, 241, 242, 264, 290,
pH ....................................... 25, 189, 247, 318, 319, 323, 303, 305, 337, 340, 348, 349, 356, 366, 380,
325, 442, 446, 452 386, 389, 409, 413, 429, 430, 445
Pharmacodynamics........................................................ 298 longitudinal relaxation .................... 15–17, 23, 29, 65
Pharmacokinetics (PK) .......................381, 384, 409–417 longitudinal relaxation time, (see T1
Phase encoding.............................. 11, 13, 22, 27, 28, 31, relaxation rate, (see R1; R2; R2*
34, 96, 99, 114, 140, 145, 215, 227, 309, 358 transverse relaxation .................................... 17, 18, 23,
Phase image .............. 206, 207, 218, 220, 226, 229, 264 32, 205, 380
Phase mask............................................................ 206, 207 transverse relaxation time, (see T2; T2*
Phosphocholine (PC) ................................ 109, 155, 156, Relaxivity ................................................... 4, 25, 309, 447
160, 162, 190, 191, 237, 331, 332, 339, 340 Repetition time (TR) ................................. 11, 14, 22, 63,
Phosphocreatine (PCr) ..................... 155, 160, 162, 332, 91, 130, 157, 190, 218, 261, 271, 290, 300,
340, 343 337, 348, 354, 398, 412, 413
Phosphodiester (PDE)......................................... 332, 340 Respiratory system ........................................................ 433
Phosphoethanolamine (PE) ................................ 332, 340 Resting state ................................................ 119, 121, 144
Phospholipids .............................................. 236, 251, 340 RF channel..................................................................... 174
Phosphomonoester (PME) ................................. 332, 340 RF coil ........................................... 45, 50, 61, 62, 65, 91,
Physiological monitoring........................... 47, 62, 63, 95, 128, 138, 157, 174, 190, 207–209, 270, 271,
121, 126, 140, 260, 261, 301–303, 424, 426, 273, 278, 280, 282, 286, 303, 350, 352, 397,
427, 437 401, 412
Pixel-by-pixel analysis ................................................... 382 birdcage resonator.......................................... 350, 397
PO2 ............................298–303, 305, 307–309, 364, 433 phased array coil .....................................286–288, 293
pO2 mapping surface coil ......................................45, 47, 65, 67, 91,
DOCENT.................... 298, 300–303, 305, 306, 309 95, 132, 138, 140, 157, 162, 174, 185, 271,
FREDOM.............................................. 301, 307, 308 286–288, 351, 397
Poly-methyl methacrylate (PMM) ...........................46, 47 volume coil ........................................... 45, 47, 65, 91,
Post labeling delay ............................................. 63, 67, 68 95, 138, 140, 157, 286, 287, 351, 397
Proton density (PD) .................................. 18, 23, 24, 28,
33, 34, 63, 65, 67 S
Pulse sequence..................................4–15, 18, 21, 22, 29, Segmentation .................................... 260, 264, 265, 271,
43, 97, 142, 170, 207, 219, 242, 270, 281,
349, 353, 361, 366, 369, 399
369, 386 Segmented EPI ........................................ 96, 97, 99, 113,
Pyruvate ......................................................................... 175 141, 142, 145
Self-gating...................................................................... 281
Q
Short TI inversion recovery (STIR).................. 30, 31, 35
Quantitative imaging biomarkers (QIB) ....................... 14 Signal-to-noise ratio (SNR).................................... 55, 96,
Quantitative susceptibility mapping (QSM)...............219, 108, 121, 124, 141, 270, 300, 335, 348
226, 357 Single-shot EPI ............................................................... 96
Slice selection .........................................9, 10, 14, 22, 27,
R 44, 113, 243, 303
R1 ..................................................... 4, 17, 298, 299, 303, Slow diffusion coefficient (Dslow)............... 137, 141, 143
305, 308, 309 Slow diffusion phase (SDP)........................ 141, 143, 144
Spatial response function (SRF) ................................... 190
R2 ..................................................................................... 46
R2* ............................. 43, 206, 219, 227, 299, 303, 308 Specific absorption rate (SAR) .................................23, 30
Radial diffusivity.............................................13, 106, 354 Spin density (NH).................................................. 89, 241
Spin dephasing ....................................................... 89, 205
Radio-frequency (RF) .................................. 5, 21, 60, 61,
207, 270, 271, 290, 339, 350, 397 Spin-echo (SE) .....................................14, 23, 27, 90, 96,
Rapid acquisition using radiofrequency echoes 130, 136, 140
(RARE) ............ 122, 190, 194, 290, 337, 389 Spoiled gradient-echo ............................. 26, 66, 206, 300
464 Index
Statistical parametric mapping (SPM) ..........63, 122, 281 Tissue oxygen level dependent (TOLD) ....................299,
Steady state ...........................................34, 123, 126, 176, 300, 303, 306, 308
182, 184, 396 Transit time ................................................. 43, 44, 67, 68
Stejskal-Tanner (ST) ...................... 91, 95, 136, 353, 379 Tricarboxylic acid cycle (TCA) ..................................... 175
Stimulated echo (STE) ................................................. 154 Tridimensional (3D) ............................................. 6, 9, 51,
Stroke.................................. 36, 45, 54, 59, 67, 104, 154, 96, 141, 215, 219–221, 226, 227, 229, 230,
243, 297, 316, 347, 356, 359, 377–390, 395 243, 246, 247, 263, 285, 286, 288–292, 322,
Stroke volume (SV)....................................................... 279 323, 336–338, 343, 348, 357–359, 386, 424
Superparamagnetic nanoparticles.......................... 25, 447 Tumor ....................................................... 4, 5, 25, 36, 45,
Susceptibility artifacts ................113, 207, 222, 389, 416 59, 154, 155, 297–300, 303, 305–308, 316,
Susceptibility-weighted imaging (SWI) ........................ 34, 332, 333, 335, 337, 339–343, 355, 396, 445,
205–207, 209, 213, 218–221, 224–227, 230, 449, 451–453
357, 381, 386, 387 Tumor oxygenation ...................................................... 299
Turbo spin echo (TSE) ....................................30, 31, 262
T
U
T1 .....................................................................3, 4, 15–18,
23–25, 28, 30, 32–35, 42, 52, 60, 67, 68, 89, Ultrafast gradient echo ................................................... 26
90, 146, 165, 171, 200, 222, 242, 288, 289, Ultra-fast spin echo ......................................................... 26
298, 300, 301, 303, 307–309, 337, 340, 348, Ultra-short echo time (UTE).............270, 273, 281, 282
351, 356, 358, 359, 361, 363, 378, 380, 396,
404, 405, 407, 416, 447, 452 V
T1 relaxation .......................... 15, 30, 340, 348, 356, 366 Ventilation ...........................................251, 433, 435, 436
T1-weighted.................................4, 18, 23, 66, 163, 252, Ventricular ............................................................ 279, 356
288, 289, 294, 306, 337, 343, 349, 355, 356,
Volume of interest (VOI) ............ 96, 141, 190, 193, 243
363, 366, 396, 405, 412, 453 Voxel based analysis (VBA) ................................. 349, 363
T2 ................................................... 3, 23, 24, 42, 89, 122,
140, 157, 171, 193, 242, 262, 289, 305, 348, W
380, 412, 445
T2 relaxation .............................................. 16, 17, 32, 380 Water suppression
T2-weighted............................................4, 14, 17, 19, 64, chemical shift selective (CHESS) .................. 163, 181
122, 131, 157, 158, 178, 252, 288–292, 294, variable power and optimized relaxation delays
356, 366, 389, 412, 413, 415, 416, 451 (VAPOR) ......... 163, 164, 181, 196, 198, 343
T2* ...........................................23, 24, 32–35, 42, 43, 46, White matter (WM) ................................. 29, 35, 65, 106,
50, 52, 56, 89, 125, 225, 264, 288, 289, 107, 200, 226, 294, 348, 349, 355, 359, 360,
298–300, 303, 305, 308, 316, 356, 380, 381, 363, 364, 379, 380, 388
387, 447
X
T2* relaxation ..................................................43, 90, 367
T2*-weighted ............................................. 206, 219, 288, Xenograft .................. 305, 307, 308, 333, 335, 340–342
289, 299, 301, 303, 306, 316, 380, 389, 416 X-nuclei................................................................. 169–171
Target.........................................112, 119, 126, 175, 176, Xylazine................................ 65, 221, 301, 302, 348, 429
182, 183, 211, 236, 237, 239–242, 245, 248,
249, 254, 287, 290, 315, 316, 318–322, 325, Z
409, 410, 425, 431, 442, 444–447, 449
Zero crossing point ......................................................... 29
Theranostic ...........................................442, 447, 449, 452

Preclinical MRI - Methods and Protocols (PDFDrive)

Uploaded by

Copyright:

Available Formats

Preclinical MRI - Methods and Protocols (PDFDrive)

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Preclinical MRI - Methods and Protocols (PDFDrive)

Uploaded by

Copyright:

Available Formats

Methods in

Molecular Biology 1718

María Luisa García-Martín

For further volumes:

Methods and Protocols

María Luisa García-Martín

ISSN 1064-3745 ISSN 1940-6029 (electronic)

© Springer Science+Business Media, LLC 2018

Printed on acid-free paper

This Humana Press imprint is published by Springer Nature

Málaga, Spain Marı́a Luisa Garcı́a-Martı́n

PART I MRI BASICS

PART II PERFUSION, DIFFUSION AND FUNCTIONAL MRI

3 Dynamic Susceptibility Contrast MRI in Small Animals. . . . . . . . . . . . . . . . . . . . . . 41

PART III IN VIVO SPECTROSCOPY

PART IV SPECIAL MRI TECHNIQUES

13 Susceptibility Weighted MRI in Rodents at 9.4 T . . . . . . . . . . . . . . . . . . . . . . . . . . . 205

PART V MRI AND MRS IN ANIMAL MODELS OF DISEASE

20 Magnetic Resonance Spectroscopy Studies of Mouse Models of Cancer . . . . . . . 331

PART VI OTHER APPLICATIONS

PART VII ANESTHESIA AND ADVANCED CONTRAST AGENTS

25 Anesthesia and Monitoring of Animals During MRI Studies . . . . . . . . . . . . . . . . . 423

26 Advanced Contrast Agents for Multimodal Biomedical Imaging

NURIA ARIAS-RAMOS  Departament de Bioquı́mica i Biologia Molecular, Unitat de

KRISTINE GLUNDE  Division of Cancer Imaging Research, Russell H. Morgan Department

M. CARMEN MUÑOZ-HERNÁNDEZ  BIONAND, Andalusian Centre for Nanomedicine and

SEBASTIAN TEMME  Experimental Cardiovascular Imaging, Department of Molecular

Introduction to MRI Physics

Key words Pulse sequences, Encoding, k-Space, Relaxation, Contrast

Magnetic resonance imaging (MRI) is an enormously rich tech-

exchange saturation transfer (CEST), and others. MRI pulse-

of the art, and there exist many difficult challenges to overcome,

2 Basic MRI Pulse Sequences

MRI is a form of Nuclear Magnetic Resonance (NMR), which has

where ω0 is the angular frequency, γ H is the gyromagnetic ratio for

An RF pulse is a critical building block in magnetic resonance

Here τ is the duration of the pulse, and B1 is the amplitude of

more accessible representation. This operation is the fulcrum of

where t is time, f(t) is the sinusoidal signal, and ν is the frequency.

It is particularly useful for making a sum of multiple decaying

distinguished. This amounts to labeling and/or encoding a specific

where z is the axis in which Gz is applied. This equation is one of the

other laboratory axes, x, y, or in some combination of directions, to

2.1.7 Frequency When an RF pulse is applied at the Larmor frequency, it tips

common way to achieve this is through a frequency-encoding

dependent on the spatial position of a pixel and the strength of the

2.2 Reconstruction The k-space formalism is a convenient means of describing the

such that both sides of an echo can be obtained. However, in the k-

3 Relaxation and Image Contrast

In MR imaging, the underlying biophysical mechanisms that mod-

diseased state, or a response to therapeutic intervention. To

3.1 Relaxation Types Relaxation is a phenomenon inherent to MR that can be used to

relaxation is the rate constant that describes signal decay in that

Experimentally, there is a litany of methods available to measure

Here TR is used in place of t. Alternatively, linear regression of the

where R1 is the relaxation rate constant after injection of contrast

T2 relaxation describes the phase coherence of magnetization in the

array element (image) corresponds to an increase in TE. After

how T2 varies across an image, one can choose a value of TE that is

NURIA ARIAS-RAMOS Departament de Bioquı́mica i Biologia Molecular, Unitat de

KRISTINE GLUNDE Division of Cancer Imaging Research, Russell H. Morgan Department

M. CARMEN MUÑOZ-HERNÁNDEZ BIONAND, Andalusian Centre for Nanomedicine and

SEBASTIAN TEMME Experimental Cardiovascular Imaging, Department of Molecular

water (45 C) for 45 s. Also to make a small pressure in the tail