P1: SFK/UKS P2: SFK Color: 4C Cardiac CT, PET and MR · 2016-08-12 · Antti Saraste, Hossam Sherif...

P1: SFK/UKS P2: SFK Color: 4C

fm BLBK309-Dilsizian August 10, 2010 18:7 Trim: 254mm X 178mm Printer Name: Yet to Come

Cardiac CT, PETand MRSecond Edition

E D I T E D B Y

Vasken Dilsizian, MD, FACC, FAHAProfessor of Medicine and Radiology

Director, Cardiovascular Nuclear Medicine and PET Imaging

Chief, Division of Nuclear Medicine

University of Maryland School of Medicine

Baltimore, MD

USA

and

Gerald M. Pohost, MD, FACC, FAHAProfessor of Radiology, Keck School of Medicine

Professor of Electrical Engineering, Viterbi School of Engineering

University of Southern California

Los Angeles, CA;

Professor of Medicine, School of Medicine

Loma Linda University

Loma Linda, CA

USA;

Professor (Honorary), Xiamen University

Xiamen, Fujian

Peoples Republic of China

A John Wiley & Sons, Ltd., Publication

iii


fm BLBK309-Dilsizian July 21, 2010 19:33 Trim: 254mm X 178mm Printer Name: Yet to Come

ii



Cardiac CT, PET and MR

i



ii



Cardiac CT, PETand MRSecond Edition

E D I T E D B Y

Vasken Dilsizian, MD, FACC, FAHAProfessor of Medicine and Radiology

Director, Cardiovascular Nuclear Medicine and PET Imaging

Chief, Division of Nuclear Medicine

University of Maryland School of Medicine

Baltimore, MD

USA

and

Gerald M. Pohost, MD, FACC, FAHAProfessor of Radiology, Keck School of Medicine

Professor of Electrical Engineering, Viterbi School of Engineering

University of Southern California

Los Angeles, CA;

Professor of Medicine, School of Medicine

Loma Linda University

Loma Linda, CA

USA;

Professor (Honorary), Xiamen University

Xiamen, Fujian

Peoples Republic of China

A John Wiley & Sons, Ltd., Publication

iii



This edition first published 2010, C© 2010 by Blackwell Publishing Ltd

Blackwell Publishing was acquired by John Wiley & Sons in February 2007. Blackwell’s publishing program hasbeen merged with Wiley’s global Scientific, Technical and Medical business to form Wiley-Blackwell.

Registered office: John Wiley & Sons Ltd, The Atrium, Southern Gate, Chichester, West Sussex, PO19 8SQ, UK

Editorial offices: 9600 Garsington Road, Oxford, OX4 2DQ, UKThe Atrium, Southern Gate, Chichester, West Sussex, PO19 8SQ, UK111 River Street, Hoboken, NJ 07030-5774, USA

For details of our global editorial offices, for customer services and for information about how to apply forpermission to reuse the copyright material in this book please see our website at www.wiley.com/wiley-blackwell

The right of the author to be identified as the author of this work has been asserted in accordance with theCopyright, Designs and Patents Act 1988.

All rights reserved. No part of this publication may be reproduced, stored in a retrieval system, or transmitted,in any form or by any means, electronic, mechanical, photocopying, recording or otherwise, except aspermitted by the UK Copyright, Designs and Patents Act 1988, without the prior permission of the publisher.

Wiley also publishes its books in a variety of electronic formats. Some content that appears in print may not beavailable in electronic books.

Designations used by companies to distinguish their products are often claimed as trademarks. All brandnames and product names used in this book are trade names, service marks, trademarks or registeredtrademarks of their respective owners. The publisher is not associated with any product or vendor mentionedin this book. This publication is designed to provide accurate and authoritative information in regard to thesubject matter covered. It is sold on the understanding that the publisher is not engaged in renderingprofessional services. If professional advice or other expert assistance is required, the services of a competentprofessional should be sought.

The contents of this work are intended to further general scientific research, understanding, and discussiononly and are not intended and should not be relied upon as recommending or promoting a specific method,diagnosis, or treatment by physicians for any particular patient. The publisher and the author make norepresentations or warranties with respect to the accuracy or completeness of the contents of this work andspecifically disclaim all warranties, including without limitation any implied warranties of fitness for aparticular purpose. In view of ongoing research, equipment modifications, changes in governmentalregulations, and the constant flow of information relating to the use of medicines, equipment, and devices, thereader is urged to review and evaluate the information provided in the package insert or instructions for eachmedicine, equipment, or device for, among other things, any changes in the instructions or indication of usageand for added warnings and precautions. Readers should consult with a specialist where appropriate. The factthat an organization or Website is referred to in this work as a citation and/or a potential source of furtherinformation does not mean that the author or the publisher endorses the information the organization orWebsite may provide or recommendations it may make. Further, readers should be aware that Internet Websiteslisted in this work may have changed or disappeared between when this work was written and when it is read.No warranty may be created or extended by any promotional statements for this work. Neither the publishernor the author shall be liable for any damages arising herefrom.

Library of Congress Cataloging-in-Publication Data

Cardiac CT, PET, and MR / edited by Vasken Dilsizian and Gerry Pohost. – 2nd ed.p. ; cm.

Rev. ed. of: Cardiac CT, PET, and MRI / edited by Vasken Dilsizian and Gerald Pohost. 2006.Includes bibliographical references and index.ISBN 978-1-4051-8553-01. Cardiovascular system–Imaging. 2. Cardiovascular system–Tomography. 3. Cardiovascularsystem–Magnetic resonance imaging. 4. Tomography, Emission. I. Dilsizian, Vasken. II. Pohost,Gerald M. III. Cardiac CT, PET, and MRI.[DNLM: 1. Diagnostic Techniques, Cardiovascular. 2. Coronary Vessels–radionuclide imaging.3. Diagnostic Imaging–methods. 4. Heart–radionuclide imaging. WG 141 C26456 2010]RC683.5.I42C33 2010616.1′0757–dc22 2010010793

ISBN: 9781405185530

A catalogue record for this book is available from the British Library.

Set in 9.5/12 pt. Minion by Aptara R©, Inc., New Delhi, IndiaPrinted and bound in Malaysia

1 2010

iv

http://www.wiley.com/wiley-blackwell



Contents

List of Contributors, vi

Foreword, ix

Part I Instrumentation, ImagingTechniques, and Protocols

1 Positron Emission Tomography, 3Stephen L. Bacharach

2 Cardiovascular Magnetic Resonance: BasicPrinciples, Methods, and Techniques, 30Joseph Selvanayagam, Matthew Robson, JaneFrancis & Stefan Neubauer

3 Cardiac Computed Tomography, 72Ian S. Rogers, Quynh A. Truong, Subodh B.Joshi & Udo Hoffmann

Part II Clinical Applications

4 PET Assessment of Myocardial Perfusion, 95Thomas H. Schindler, Ines Valenta & VaskenDilsizian

5 Myocardial Metabolism in Health andDisease, 118Robert J. Gropler, Linda R. Peterson & VaskenDilsizian

6 PET Innervation and Receptors, 140Antti Saraste, Hossam Sherif & MarkusSchwaiger

7 MR Angiography: Coronaries and GreatVessels, 154Patricia Nguyen & Phillip Yang

8 Cardiovascular Magnetic Resonance:Evaluation of Myocardial Function, Perfusion,and Viability, 196Padmini Varadarajan, Ramdas G. Pai, KrishnaS. Nayak, Hee-Won Kim & Gerald M. Pohost

9 MSCT Coronary Imaging, 246Koen Nieman

10 Multislice Cardiac Tomography: MyocardialFunction, Perfusion, and Viability, 259Raymond T. Yan, Richard T. George & Joao A.C.Lima

11 Cardiac Computed Tomography and MagneticResonance for the Evaluation of Acute ChestPain in the Emergency Department, 278Eric M. Thorn & Charles S. White

Part III Concurrent NoninvasiveAssessment of CoronaryAnatomy, Physiology, andMyocellular Integrity

12 PET and MRI in Cardiac Imaging, 301Stephan G. Nekolla & Antti Saraste

13 PET and CT Imaging, 321Marcelo F. Di Carli

14 Image-Guided Electrophysiology Mapping andAblation, 334Timm-Michael L. Dickfeld

15 Structural and Molecular Imaging ofVulnerable Plaques, 354Farouc A. Jaffer & Jagat Narula

Index, 368

v



List of Contributors

Stephen L. Bacharach, PhDAdjunct Professor of Radiology

and Senior Tenured Research Scientist, NIH (ret)Center for Molecular and Functional ImagingUniversity of California at San FranciscoSan Francisco, CAUSA

Marcelo F. Di Carli, MDAssociate Professor of RadiologyDepartments of Radiology and MedicineBrigham and Women’s HospitalHarvard Medical SchoolChief of Nuclear Medicine, Co-Director of

Cardiovascular ImagingBoston, MAUSA

Timm-Michael L. Dickfeld, MD, PhDAssociate Professor of MedicineDivision of CardiologyUniversity of Maryland School of MedicineChief of Electrophysiology, VA BaltimoreBaltimore, MDUSA

Jane Francis, DCCRChief Cardiac MRI RadiographerDepartment of Cardiovascular MedicineUniversity of OxfordOxfordUK

Richard T. George, MDAssistant Professor of MedicineDepartment of Medicine, Division of CardiologyJohns Hopkins UniversityBaltimore, MD, USA

Robert J. Gropler, MD, FACCProfessor of Radiology, Medicine and Biomedical

EngineeringLab Chief, Cardiovascular Imaging LaboratoryMallinckrodt Institute of RadiologyWashington University School of MedicineSt. Louis, MOUSA

Udo Hoffmann, MD, MPHAssociate Professor of RadiologyDirector, Cardiac MR PET CT ProgramMassachusetts General HospitalHarvard Medical SchoolBoston, MAUSA

Farouc A. Jaffer, MD, PhDAssistant Professor of MedicineHarvard Medical School, Boston, MADirector of Vascular ImagingCenter for Molecular Imaging Research and Cardiovascular

Research Center, Attending Interventional CardiologistMassachusetts General HospitalBoston, MAUSA

Subodh B. Joshi, MD, MPHResearch FellowCardiac MR PET CT ProgramMassachusetts General HospitalHarvard Medical SchoolBoston, MAUSA

Hee-Won Kim, PhDAssistant Professor of RadiologyUSC/Keck School of MedicineUniversity of Southern CaliforniaLos Angeles, CAUSA

Joao A. C. Lima, MD, MBAProfessor of Medicine, Radiology and EpidemiologyDirector of Cardiovascular ImagingJohns Hopkins UniversityBaltimore, MDUSA

Jagat Narula, MD, PhD, FACC, FRCPProfessor of MedicineIrvine School of MedicineChief, Division of CardiologyDirector, Memorial Heart & Vascular InstituteMedical Director

vi



List of Contributors vii

Edwards Lifesciences Center for Advanced CardiovascularTechnology

University of CaliforniaIrvine, CAUSA

Krishna S. Nayak, PhDAssociate Professor of Electrical Engineering, Biomedical

Engineering, Medicine, and RadiologyDirector, Magnetic Resonance Engineering LaboratoryViterbi School of EngineeringKeck School of MedicineUniversity of Southern CaliforniaLos Angeles, CAUSA

Stephan G. Nekolla, PhDSenior PhysicistKlinik fur NuklearmedizinTechnisch Universitat MunchenGermany

Stefan Neubauer, MDProfessor of Cardiovascular MedicineDepartment of Cardiovascular MedicineUniversity of OxfordOxfordUK

Patricia Nguyen, MDInstructor, Department of MedicineDivision of Cardiovascular MedicineStanford University Medical CenterStanford, CAUSA

Koen Nieman, MD, PhD, FESCResident in CardiologyDepartment of Cardiology (Thoraxcenter) and Department

of RadiologyErasmus Medical CenterRotterdamThe Netherlands

Ramdas G. Pai, MD , FACC, FRCP (Edin)Professor of MedicineDivision of CardiologyLoma Linda University Medical CenterLoma Linda, CAUSA

Linda R. Peterson, MD, FACC, FAHAAssociate Professor of Medicine and RadiologyCardiovascular Division and Division of Geriatrics and

Nutritional SciencesDepartment of MedicineWashington University School of MedicineSt. Louis, MOUSA

Matthew Robson, PhDMRI PhysicistDepartment of Cardiovascular MedicineUniversity of OxfordOxfordUK

Ian S. Rogers, MD, MPHFirst Year Clinical FellowDivision of Cardiovascular MedicineStanford University Medical CenterStanford, CAUSA

Antti Saraste, MDResearch FellowNuklearmedizinische Klinik Technischen Universitat

MunchenMunichGermany

Thomas H. Schindler, MDAssistant Professor of Internal Medicine - CardiologyChief of Nuclear Cardiology and Cardiac PETUniversity Hospitals of GenevaGenevaSwitzerland

Markus Schwaiger, MDDirector, Department of Nuclear MedicineNuklearmedizinische Klinik Technischen Universitat

MunchenMunichGermany

Joseph Selvanayagam, MBBS (Hons),FRACP, DPhil, FCSANZ, FESCProfessor of Cardiovascular MedicineFlinders University of South AustraliaDirector Cardiac MR & CTFlinders Medical CentreAdelaide, SAAustralia

Hossam Sherif, MDResearch FellowNuklearmedizinische Klinik Technischen

Universitat MunchenMunichGermany

Eric M. Thorn, MD, MPHVirginia Cardiovascular Associates, P.C.Manassas, VAUSA



viii List of Contributors

Quynh A. Truong, MD, MPHInstructor in RadiologyCardiac MR PET CT ProgramMassachusetts General HospitalHarvard Medical SchoolBoston, MAUSA

Ines Valenta, MDResearch FellowNuclear Cardiology and Cardiac PETUniversity Hospitals of GenevaGenevaSwitzerland

Padmini Varadarajan, MD, FACCAssociate Professor of MedicineDivision of CardiologyLoma Linda University Medical CenterLoma Linda, CAUSA

Charles S. White, MDProfessor of Radiology and MedicineUniversity of Maryland School of MedicineChief of Thoracic RadiologyDepartment of Diagnostic RadiologyUniversity of Maryland Medical CenterBaltimore, MDUSA

Raymond T. Yan, MD, MAScClinical and Research FellowCardiac MR and CT ProgramDivision of CardiologyJohns Hopkins HospitalBaltimore, MDUSA

Phillip Yang, MDAssistant Professor, Department of MedicineDivision of Cardiovascular MedicineStanford University Medical CenterStanford, CAUSA



Foreword

Building upon several decades of technological de-velopment and clinical imaging experience, thefield of cardiovascular imaging has made spec-tacular advances in recent years. The advancedimaging techniques—positron emission tomog-raphy (PET), cardiovascular magnetic resonance(CMR), and cardiac computed tomography (CT)—now deliver noninvasive coronary angiograms, newexquisitely detailed insights into the coronary arterywall, characterization of cardiac structure and func-tion, and complementary functional data regardingperfusion, metabolism, and viability. These modal-ities challenge the existing paradigms for diagno-sis and risk stratification while presenting uncer-tainties for the practitioner regarding their optimalclinical application relative to diagnostic angiogra-phy and standard stress testing. Against this rapidlyevolving milieu, Drs Dilsizian and Pohost have pro-duced a remarkably definitive and balanced text thatcaptures the full scope, and the excitement, of cur-rent cardiovascular imaging knowledge. This newedition of Cardiac CT, PET, and MR is a timely, valu-able resource for a wide range of readers includingthe researchers who continue to enrich the field,the imaging subspecialists (both beginner and ex-pert) who refine the applications, and the clinicianswho refer patients for diagnostic imaging proce-dures. The editors have recruited an internationallyrecognized and talented panel of expert authors,who are the leading authorities in their respectivedisciplines.

Cardiac CT, PET, and MR provides in-depth,comprehensive discussion of the technical char-acteristics and clinical applications of each of theadvanced imaging modalities, including their com-parative strengths and weaknesses. Essential reviewsof imaging physics instrumentation and imagingprotocols underlying PET, CMR, and CT technol-

ogy in the initial chapters provide the founda-tion for the broader discussion that follows of themany present and future applications of each ofthe imaging technologies. The rapidly accruing ev-idence base supporting the current state of the artis presented in detail in a well-balanced dialogue.The final chapters focus on the exciting new direc-tions in fusion imaging with combined PET/CMRand PET/CT systems for concurrent assessment ofanatomy and physiology, mapping applications toguide advanced cardiac electrophysiologic proce-dures, and the frontiers of noninvasive characteri-zation of coronary plaque anatomy and pathobiol-ogy.

Noninvasive imaging methods are fundamentaltools for all physicians involved in the diagnosis andtreatment of patients with heart disease. PET, CMR,and CT have enormous potential to accelerate un-derstanding of basic pathophysiologic processes inanimals and humans while also providing the keysfor early diagnosis and assessment of efficacy ofnew therapies. There is also great need for clinicaltrials and comparative effectiveness research to de-termine the right test for the right patient at theright time. An understanding of the current andfuture capabilities of noninvasive imaging is essen-tial to fully achieve this potential. This new vol-ume of Cardiac CT, PET, and MR itself has arrivedat the right time to contribute importantly to thisprogress.

Robert O. Bonow, M.D.Goldberg Distinguished ProfessorNorthwestern University Feinberg School

of MedicineChief, Division of CardiologyNorthwestern Memorial HospitalChicago, Illinois

ix



x


c01 BLBK309-Dilsizian June 9, 2010 17:20 Trim: 254mm X 178mm Printer Name: Yet to Come

I PART I

Instrumentation,Imaging Techniques,and Protocols

1



2



1 CHAPTER 1

Positron Emission Tomography

Stephen L. BacharachUniversity of California, San Francisco, CA, USA

The goal of all cardiac nuclear imaging is to tracethe fate of radioactively labeled biochemical com-pounds (tracers) within the body, usually in the my-ocardium or blood pool. One usually either makesa static image of the distribution of the radiotracer(e.g., 18F-fluorodeoxyglucose (18FDG) or thallium-201 (201Tl)) or follows the uptake and clearance ofthe tracer with time. In the former case, static imag-ing is all that is required, while in the latter a seriesof images, acquired dynamically over time, is nec-essary. Positron emission tomography (PET) hasthese same goals. Although PET works in a mannervery similar to conventional tomographic nuclearimaging techniques (e.g., single photon emissioncomputed tomography or SPECT), there are somevery significant differences. It is these differencesthat make PET of great potential value in nuclearcardiology, and it is these differences we will em-phasize in this chapter.

Positron Decay

PET tracers, as their name implies, decay by emis-sion of a positron. Except for their opposite charge,positrons are nearly identical to ordinary nega-tively charged electrons (which in fact are oftencalled “negatrons”). They have the same mass andbehave similarly when passing through the body.Positrons, however, are the “antimatter” of elec-trons. When a positron and an electron are inclose proximity for more than the briefest inter-val, both will disappear (called “annihilation”),and their masses will be converted into energy

in the form of two gamma rays traveling in al-most exactly opposite directions. The energy ofeach photon is 0.511 meV (exactly the equivalentenergy corresponding to the mass of the electronor positron). These photons are sometimes called“annihilation” photons. The two photons travelin nearly exactly opposite directions in order toconserve momentum. The entire process is illus-trated in Figure 1.1. In this figure it is assumed thata positron emitter (in this case carbon-11 (11C)) isemitted by a tracer somewhere in the body (e.g., themyocardium). When the positron is emitted fromthe nucleus it is traveling at very high speed—nearlythe speed of light. It moves through the tissue justas an electron would, bouncing off many of theatoms and losing energy as it does so. Eventually(typically within a millimeter or so, depending onthe radionuclide) it slows down enough to spenda significant time near an electron. As soon as thishappens the two annihilate and the two gammarays (each with 0.511 meV) are emitted, as shownin Figure 1.1, each going in nearly the exact oppositedirection. Although in Figure 1.1 the annihilationphotons are shown traveling in exactly opposite di-rections, occasionally photons are emitted a fewtenths of a degree more or less than 180◦ apart.

PET scanners detect pairs of gamma rays result-ing from annihilation. By determining where thesetwo gamma rays (and all other pairs of gammarays) originated, the PET scanner can produce animage showing the location in the body where thepositrons have annihilated. However, if the positronhas traveled far from its parent atom, the imagewill be inaccurate—the locus of the annihilatingpositron will not correspond to the locus of theradioactive atom. For this reason the initial speed

Cardiac CT, PET and MR, 2nd edition. Edited by Vasken Dilsizian

and Gerry Pohost. c© 2010 Blackwell Publishing Ltd.

3



4 PART I Instrumentation, Imaging Techniques, and Protocols

Positron emission and annihilation

Gamma ray

β+

β−

Gamma rayOne to severalmm

11C

Figure 1.1 A positron is shown being emitted from thenucleus of 11C. It is assumed that the 11C atom is located intissue. The positron is initially emitted at a speed which is asignificant fraction of the speed of light. As it passesthrough the tissue, it gradually slows down, as it bouncesoff the atoms in the tissue. Eventually it slows downsufficiently so that it spends significant time near anatomic electron—its antimatter equivalent. When thishappens the electron and the positron both annihilate—their mass being converted to energy in the form of twophotons traveling in opposite directions, as shown.

(i.e., energy) of an emitted positron will affect thecapacity of the PET scanner to accurately definethe position of radioactive atoms within the my-ocardium. This in turn affects the ultimate spatialresolution of the images that can be obtained witha PET scanner.

There are many radioisotopes that emitpositrons, and so would be suitable for use witha PET scanner. Several of the most important onesare listed in Table 1.1, along with their half-lives andsome characteristics of the positron that is emitted

[1]. One of the reasons why PET has played suchan important role in basic research is that severalof the radioisotopes that are positron emitters (car-bon, nitrogen, and oxygen) are the basic buildingblocks of all physiologically important biochemicalcompounds. This has permitted researchers to labelamino acids, glucose, and a host of other biochem-ical compounds. Unlike the case with technetium-99m (99mTc), the labeling can often be done withoutmaking any alterations to the biochemical structureof the compound of interest. That is, a nonradioat-ive 12C atom can be replaced with a 11C atom, so thatthe resultant radiolabeled biochemical compoundbehaves just like the unlabeled one. The difficultywith 11C, nitrogen-13 (13N) and oxygen-15 (15O)is that their half-lives are very short. This meansthey must be produced locally with an on-site cy-clotron. It also means that the chemist in chargeof labeling the biochemical compound of interesthas very little time to do so. For these reasons (andothers discussed below), the two most clinically im-portant positron-emitting isotopes for cardiologyare the last two on the list, fluorine-18 (18F) andRubidium-82 (82Rb).

18F has a 2-hour half-life. This is long enough toallow production at a site up to a few 100 km away.The recent dramatic increase in the use of 18FDGfor tumor imaging has resulted in a large number ofsuch commercial production sites in the US (and toa lesser extent abroad). One can easily arrange fordelivery of daily unit doses of 18FDG. 18FDG hasproven very valuable in assessing myocardial via-bility [2]. Its use for this purpose has, in the past,been limited to large research institutions because

Table 1.1 Positron energies and ranges (in tissue).

Isotope

Maximum energy

(meV)

Average energy

(meV)

Average distance

positrons travel (mm)

Maximum distance

positrons travel (mm)

18F 0.635 0.250 0.35 2.311C 0.96 0.386 0.56 4.113N 1.19 0.492 0.72 5.215O 1.72 0.735 1.1 8.188Ga 1.90 0.836 1.1 9.482Rba 3.35 (83%) 1.52 2.4 16.7

aRb emits two different positrons. Eighty-three percent of the time it emits a 3.35-meV maximum energy positron and

12% of the time a 2.57-meV positron.



CHAPTER 1 Positron Emission Tomography 5

of the lack of availability of 18FDG and a PET scan-ner. As mentioned, 18FDG is now widely availablecommercially, and there are a huge number of newPET scanners which have been installed, the ma-jority in nonresearch hospitals. Although most ofthese scanners were installed for oncology imaging,the machines are often suitable for cardiac imagingas well.

The other clinically important radiopharmaceu-tical in Table 1.1 is 82Rb. This is a potassium ana-log and can be used to measure myocardial perfu-sion [3]. No labeling is required. Although it has avery short half-life (76 seconds), it can be producedfrom a longer lived 82Sr generator, with a half-life of25 days. At the moment such generators are fairlyexpensive, but the cost is expected to drop substan-tially if demand increases.

Aside from half-life, two other factors mustbe considered when determining the utility of apositron-emitting isotope. First, it is important thatnearly all the decays are by positron emission, ratherthan by other forms of decay whose emissions can-not be imaged with a PET scanner. 11C, 13N, and15O all decay nearly 100% of the time by positronemission, and 82Rb decays about 95% of the timeby positron emission [4]. The remaining fractionof the decays is by electron capture—a process thatproduces radiation that cannot be imaged with aPET scanner. In addition, for 82Rb a small frac-tion (∼12%) of the positrons are accompanied byan additional high-energy gamma ray (0.778 meV)which can produce some interference with imagingthe 0.511-meV annihilation photons and which in-creases radiation exposure slightly. There are otherpositron emitters (e.g., 94Tc, 124I, several isotopes ofCu, and many others) that have an even larger num-ber of other emissions and other significant modesof decay. This often results in poorer dosimetry forthe patient because these emissions may increasethe patient’s radiation exposure, but do not produceuseful imaging information. Nonetheless, many ofthese isotopes have been used successfully in PETimaging.

The second factor one must consider when eval-uating a radioisotope is the energy of the positronthat is emitted. As mentioned above, this is im-portant because what one images with a PET scan-ner is not the distribution of the radiotracer, butrather the distribution of the annihilation photons.

Positrons are not emitted with a single characteris-tic energy as are gamma rays. Instead they have arange of possible energies from 0 up to a charac-teristic maximum energy. Each positron-emittingradionuclide has its own characteristic maximumand average energy of positron emission, as shownin Table 1.1. Because of this, and because the pathof the positron as it slows down is quite tortuous(e.g., Figure 1.1), not all the positrons emitted by agiven type of atom travel the same distance—sometravel quite far and others do not. Table 1.1 alsoshows the average distance from the parent atomeach positron travels in tissue. The positrons emit-ted by 18F have a very low energy. Thus, on averagethey travel only a very small distance away fromthe parent atom (about 0.35 mm). In contrast, 15Oemits positrons that are considerably more ener-getic and travel an average of 1.1 mm. Positronsfrom 82Rb travel an average of 2.4 mm. Becausethe spatial resolution in a cardiac PET image canbe as good as ∼5–7 mm, the extra blurring causedby the range of travel in tissue can be significantfor isotopes such as 82Rb, and to a lesser extent,15O.

Before we can further discuss the characteristicsof PET scanners, it is necessary to understand howtomographic images are made and how they are“reconstructed” from the radioactivity seen by thering of detectors surrounding the patient. Manytreatises have been written dealing with the math-ematical steps necessary to produce cross-sectionalimages with emission tomographs [5]. Here we willdescribe the reconstruction process in a physical,rather than in a mathematical, way.

To define the three-dimensional (3D) shape ofan object, one must first be able to look at the ob-ject from all sides. This may be an evolutionaryadvantage of binocular vision (two eyes, not one).Each eye’s slightly different view of the same ob-ject, when processed by the brain, allows formationof a 3D image of the object’s surface. Because oureyes are not placed very far apart we cannot seeall sides of an object at once, and so we must ex-trapolate (often incorrectly) using the informationfrom the two angles we can see in order to visualizethe object’s full appearance. In a similar manner, aphysician may wish to examine several planar 201Tlscans, each taken at a different angle, in an effort tomentally reconstruct the 3D distribution of 201Tl in




Problem: Collimators block about 999/1000photons. Very low sensitivity device.

Collimator tells us where gamma ray came from

Extrinsic collimation

Need to know Where photons came from

Figure 1.2 In single photon tomography (SPECT) acollimator is needed to tell the direction from which thegamma ray came. The camera must then rotate around thepatient (the dashed line shows the rotation) in order tomeasure the projection images at every angle.

the myocardium. The situation in this case is morecomplex because nuclear medicine images portraynot just the surface of an object, but its interior aswell. That is, the object is transparent (except forattenuation) to its radiation.

X-ray tube

(a)

(b)

Arc of detectors

One planar projection at angle shown (anterior)

CT Scanner (one projection)

Width of the 16 rows

16 detector rows Figure 1.3 In X-ray computedtomography (CT) the X-ray tube mustmove around the body (a) to acquireprojection images just as the gammacamera must move around the body inSPECT. Image (b) shows one planarprojection. The tomographic image canbe reconstructed from a set of theseprojections at all angles.

Just as all sides of an object must be seen by theeye and brain to appreciate its 3D surface, manytwo-dimensional (2D) planar views, each taken ata different angle, are necessary to allow determi-nation of the 3D interior activity concentration ofan object. Each of these 2D views at a particularangle is referred to as a “projection.” The recon-struction process (i.e., the method for producingtomographic slices) is based on acquiring these pro-jections. PET, SPECT, and even computed tomog-raphy (CT) must all acquire projection images, andin fact all use a similar method for reconstructingthe 2D projection images into tomographic slices.The only difference is in how each modality obtainsits projection images. In SPECT these “projection”images are obtained by rotating a gamma cameraaround the object being imaged, as in Figure 1.2.In CT the views are obtained by rotating an X-raytube around the patient and measuring how manyphotons are able to get through the body (so eachprojection is just a planar X-ray image—see Figure1.3). We will see shortly how PET accomplishes thesame thing—creating a planar image of the positronannihilation radiation at each angle.

In theory, an infinite number of projections arenecessary to define the 3D distribution of activityin an object. In practice, cardiac SPECT images are




usually reconstructed from fewer than 100 angles,while several hundred different views, each at a dif-ferent angle, are usually acquired for PET.

Once the PET (or SPECT or CT) scanner hascollected data from all these projections, or views,several steps are necessary to create a tomographicslice. The details of these steps [5] are unimpor-tant for understanding the rest of this chapter. Theymay be considered simply as mathematical opera-tions that convert the many projection images intoa single tomographic section or slice.

PET scanners can simultaneously obtain all theviews necessary to reconstruct a tomographic imagewith the use of a ring (or multiple rings) containinghundreds or thousands of detectors that encircle thepatient. The mechanical assembly holding all thesedetectors is called the “gantry.” The means by whichthe ring of detectors acquires data for the manyviews required can be explained by first remember-ing the basic information that is needed to performthe reconstruction, i.e., the projection images. Aprojection image is made up of all the photons thatcame from a certain direction (projection angle).In the SPECT example of Figure 1.2 the camerais able to show from which direction the photonshave come by using a collimator [6]. All photonsthat do not strike the camera perpendicular to itsface are blocked by the collimator. So in Figure 1.2the number of gamma rays detected at each pointon the gamma-camera face must have come onlyfrom 270◦ (numbering angles clockwise, with 0 atthe top). Unfortunately, blocking all the photonsarriving from other angles is a very inefficient wayto make a projection image. Many collimators blockmore than 999 out of every 1000 photons emitted bythe radioactive atoms in the patient. Such a SPECTdevice would therefore waste over 99.9% of all thephotons emitted by the patient. That is the priceone pays for using a collimator to determine whatdirection the photons came from. PET can get thesame information—how many photons were emit-ted and what direction they came from—without acollimator, potentially making PET far more sensi-tive than SPECT. How is this done? Consider Figure1.4a, showing a ring of detectors surrounding thepatient. Only four of the hundreds of detectors areshown (and those four are shown greatly enlargedfor clarity). Imagine that a 511-keV photon has juststruck detector 3, as in Figure 1.4a. If this were the

only piece of information that the PET scanner had,it would not be of any use. We would know that anannihilation had occurred, but we would not knowfrom which direction it had come. It could havecome from almost any direction. However, recallthat for annihilation photons, there is always an-other photon traveling in the opposite direction.Therefore, if a 511-keV photon struck detector 3and simultaneously (i.e., in “coincidence”) another511-keV photon struck detector 2, then the com-puter would realize that this pair of photons mustcome from a positron that annihilated somewherealong the line B connecting the two detectors (seeFigure 1.4a). This is useful information—we knowan annihilation occurred and we know from whichdirection the two photons came. This method of de-termining where the photons came from, withouta collimator, is called “coincidence imaging” [7]. Inreality the two photons may not be detected trulysimultaneously. For example, the annihilation mayoccur closer to one of the detectors than the other,and so may reach that detector first (although thedifference in time is usually on the order of a bil-lionth of a second). One therefore usually acceptsany pair of photons that occur within a narrow timeinterval as being in “coincidence.” This window iscalled the “coincidence window” or “timing win-dow” or the “resolving time” of the scanner, usuallydesignated by the symbol τ. It is typically 5–20 nswide, depending on the scanner.

The pairs of detectors in Figure 1.4b connected bysolid lines (A-1, B-2, etc.) provide one “view” of theobject at a given angle. Coincidences between A-2,B-3, C-4, etc. (dashed lines on Figure 1.4b) provideanother view of the object, at a slightly differentangle. The PET camera has electronic circuits thatcan distinguish coincidences from every possiblepair of detectors in the field of view of the camera.

In Figure 1.4b, the solid lines comprising one“view” or projection are spaced rather far apart. Toallow the PET scanner to distinguish small objectsfrom one another, it is desirable that these lines beas close together as possible. This is accomplishedby making the width of each detector small andplacing the detectors as close together as possible.This decreases the spacing between lines andincreases the number of possible angles (and there-fore the number of views). Of course, increasingthe total number of possible coincidences in this




B

D

(c)

A

C

1

2

A

12

3 4 5

Detectorwidth

BC D E

(b)(a)

1

C

B

A

PET does not need a collimatorCollimation by coincidence detection

2

3

4

Figure 1.4 (a) Showing how the direction from which aphoton came can be determined in positron emissiontomography (PET) by use of coincidence detection. Whendetectors 2 and 3 both detect a photon at the same time,the computer deduces that the pair of photons must havecome from an annihilation along the line connectingdetectors 2 and 3, as shown. (b) Showing how groups of

detector pairs can form a projection image at a particularangle. Two projection angles are shown—the solid lineshows the anterior–posterior projection, while the dashedlines show a projection about 10◦ shifted. (c) Showing aschematic diagram of coincidence detection among thecrystals of one ring of detectors.

way increases the number of crystals, coincidencecircuits, and other electronic components required,making the PET scanner more costly.

A factor that limits the number of crystals em-ployed in a PET scanner is the number of photomul-tiplier tubes required. When a detector “detects” agamma ray, it produces a small flash of light thatis converted to an electronic pulse by a photomul-tiplier tube. Ideally each crystal would be attachedto one photomultiplier tube, but the tubes cannotbe made arbitrarily small and are quite expensive.Thus, manufacturers have devised schemes to allowone photomultiplier tube to share many crystals. Aschematic diagram of coincidence detection among

the crystals of one ring of detectors is shown in Fig-ure 1.4c.

Most scanners for cardiac imaging use sev-eral rings of detectors, often separated by highatomic number shielding (e.g., lead, tungsten)called “septa,” to acquire data for multiple slices.When a PET scanner is operated with septa betweenrings it is said to be operating in “2D” mode. Thisis a bit of a misnomer, since of course such a scan-ner still acquires 3D data. As will be discussed later,some scanners operate without the septa. Thosescanners are said to be operating in “3D” mode[8–11]. To increase the number of slices, coinci-dences are often recorded between one detector in




one ring, and an opposing detector in an adjacentring. Such a slice would be called a “cross” slice.With three rings of detectors (numbered I, II, andIII) five slices could be produced. The first wouldconsist of all coincident events from opposing pairsof detectors in ring I (a direct slice); the secondwould be a cross slice consisting of all coincidentevents between one detector in ring I and an op-posing detector in ring II (or vice versa); the thirdwould be formed from events only in ring II, and soon. Some PET scanners have completely separaterings of detectors. With this design, what consti-tutes a cross slice and what constitutes a direct sliceis obvious. Other scanners have crystals so close to-gether in the Z-axis that the concept of physicallyseparate rings no longer applies. What is importantin any case is the final spatial resolution obtained(in all three directions) and the number of, andspacing between, slices.

Cardiac PET scanners reconstruct transaxialslices. The number and spacing of the slices is usu-ally such that at least a 15-cm axial distance is en-compassed by the slices—a quite adequate size forcardiac imaging—large enough to include the en-tire left ventricle in nearly all subjects. Dependingon the scanner anywhere between 30 and 70 slicesor more cover this ∼15-cm axial field of view. It isoften desirable to include some of the left atrium inthe image also (even though it is not usually visu-alized well) to allow arterial blood concentrationsof tracer to be measured. Some scanners permit aslight rotation and tilt of the gantry, but no scannerpresently available can be positioned to yield truecardiac short-axis slices directly. Rather, one refor-mats the transaxial slices into short- or long-axisviews.

It is important to understand the quantity be-ing measured in the reconstructed image obtainedfrom a PET scan. Each of the projections describedpreviously measures simply the total number of co-incidences seen by each detector pair at a given angleduring a specific time period (the scan time). Forexample, in Figure 1.4b, one projection is formedby the solid lines A-1, B-2, etc. The quantity mea-sured by each detector pair in this projection isthe number of coincidences/second seen along theline, for example that formed by A-1. This “line”is not an infinitesimally thin line, but has a width,because the detector pair A and 1 both have fi-

nite width. The number of coincidences seen bythe pair A and 1 are those produced by all theradioactive material lying in the volume betweenthem. The units of the measurement are therefore“coincidences/second/volume.” These projectiondata are reconstructed to determine the numberof coincidences arising from each point in the finalreconstructed image. Since each point in the imagealso represents a small volume in the object beingimaged, the units are again coincidences/s/volume.Finally, it is assumed that the number of coinci-dences/second measured in a volume is directlyproportional to the amount of radioactive material(usually measured in Bq (Bequerels) or Ci (Curies))in that same volume. Providing all the correctionsdescribed below are made, this assumption is cor-rect. The units of the PET scan can therefore be anyof the following: coincidences/s/cc, Bq (or nCi)/cc,or grams of radiolabeled material/cc. Use of the lastunit is possible because Bq can be easily convertedto number of atoms or grams as long as the half-lifeis known.

Accidental Coincidences

Unfortunately, it is possible for two photons thatdid not come from the same annihilation eventto be erroneously identified, quite by accident, ashaving occurred “simultaneously,” that is, withinthe resolving time τ of the PET scanner.

Figure 1.5 illustrates such a case. Only one of thephotons from annihilation A has reached a detec-tor; the other missed the ring. At nearly the sametime atom B decayed. Only one of its photons wasdetected, the other also missing the ring. If thesetwo separate events happen to occur at nearly thesame time within the resolving time of the PETscanner, they will be considered to be in coinci-dence. The PET scanner then will falsely treat thedetection of the two photons as if they resultedfrom a single annihilation that took place alongthe line between the two detectors (the dashed linein Figure 1.5). Such false coincidence is called anaccidental or random coincidence. Random coin-cidences produce background activity in the recon-structed image that varies slowly in magnitude atdifferent positions over the image, depending onthe radioisotope distribution.




BA

B

21False coinc

detector1

A

detector2

Randoms

• Two single, unrelated photons are accidentally detected at same time• # Randoms α (activity)2

Figure 1.5 Illustrating how if byaccident, two separate annihilationevents (A and B) are detected at nearlythe same time a false or “accidental”coincidence can occur. This accidental(or “random”) coincidence causes thePET camera to erroneously think theannihilation occurred along the dashedline indicated.

Accidental coincidences between unrelated pho-tons must be distinguished from “true” coinci-dences between pairs of annihilation photons. Theprobability that an accidental coincidence will oc-cur depends on the duration of the resolving timeinterval, τ: if it is very long, it becomes much morelikely for two unrelated photons to be accidentallyin coincidence. The resolving time of a PET scanneris therefore an important parameter, defining howwell the scanner will distinguish true coincidencesfrom accidental ones.

A second factor influencing the number of acci-dental events recorded is the amount and locationof activity detected by the PET scanner. If the activ-ity within the patient is doubled, the number of truecoincidences will of course double also. The numberof accidental coincidences, however, will increase bya factor of 4, i.e., as the square of activity. This hasimportant ramifications. At sufficiently high levelsof activity (e.g., for 82Rb scans [9,11,12]), the num-ber of accidental coincidences may equal or evenexceed true coincidences. With administration ofexcessive amounts of tracer, the patient may, there-fore, be exposed to a higher radiation risk withouta comparable increase in the amount of useful in-formation obtained. The amount of activity con-stituting an excess varies with the machine used; itmay be only a few millicuries in the field of viewor as much as 50 or more mCi. Many manufactur-ers specify the concentration of activity that whenplaced in a specified phantom will produce equalnumbers of true and accidental coincidences. This

is useful to help evaluate the maximum activity thatone might inject in a patient from the point of viewof excessive randoms.

The reason that accidental events increase as thesquare of activity can be discerned from consid-eration of Figure 1.5. Suppose that detector onemeasures S1 (S refers to “singles”) counts/second,independent of whether these counts were in co-incidence with any other detector. The count rateobserved by a single detector, as opposed to a coin-cident pair of detectors, is called the singles countrate of that detector. Suppose also that detector 2measures a singles rate of S2/second. Consider thatone photon has just struck detector 1. If an unre-lated photon were to hit detector 2 within the next τ

seconds or has already hit detector 2 within the pre-vious τ seconds, it will be in accidental coincidencewith the event recognized by detector 1. Becausethere are S2 events detected by detector 2 each sec-ond, the number of these that will occur during theτ seconds before or the τ seconds after the eventin S1 is S2 × 2 × τ. For every photon that strikesdetector 1, there are therefore 2 × τ × S2 accidentalcoincidences/second. However, there are S1 pho-tons striking detector 1 every second. Therefore thetotal number of accidental coincidences/second is:

Accidental coincidences/second = 2 × τ × S1 × S2

If the activity in the patient is doubled, the sin-gles rate for every detector is also doubled, so bothS1 and S2 double, giving a factor of 4 increase inaccidental coincidences.




Consideration of the above equation suggests away to correct for accidental coincidences. If thesingles rate is measured at every detector, the num-ber of accidental coincidences can be computed forevery detector pair, and this number can be sub-tracted from the measured true events. Althoughmeasured singles rates include some counts fromtrue coincidences, singles rates usually greatly ex-ceed true coincidence rates. Thus, the error intro-duced by such a correction scheme is usually quitesmall.

Another approach to correction for random co-incidences is the delayed coincidence method. Con-sider a single pair of opposing detectors in Figure1.4. The output of detectors is split. One of thesignals goes to the coincidence circuitry as usual.The other goes to a special circuit (or even a longlength of wire) that causes a prolongation of traveltime for the signal, perhaps of several 100 ns. Thissecond wire is connected to the usual circuit, whichdetermines whether the two pulses (one from thedelayed signal from the second wire of one detector,and the second from the undelayed, first wire of theopposing detector) occurred within time τ of eachother. If a true coincidence event occurs, the de-layed signal traveling down the second wire will notregister as a coincidence with the undelayed signalof the opposing detector. The signal traveling downthe long wire would reach the coincidence electron-ics much later than the undelayed signal from theopposing detector. Any coincidences measured bythis long wire would, therefore, only be acciden-tal coincidences and not true coincidences. Theycould, therefore, be subtracted from the total num-ber of coincidences measured with the undelayedstandard short wires of both detectors to yield thenumber of true coincidences. The delayed coinci-dence method is quite accurate. However, it is lim-ited by low signal-to-noise ratios because the num-ber of randoms measured by the second delayedwire is often quite small, which may introduce ad-ditional noise into the final corrected image. Onthe other hand, use of the singles method discussedpreviously adds little noise to the image because thenumber of singles recorded by each detector is sohigh. However, the singles method has its own dif-ficulties, and requires measurement of τ, which issubject to inaccuracies.

Attenuation Correction

If 511-keV annihilation gamma rays were made totravel through a substance with a very high atomicnumber, such as a lead brick, only a few of the pho-tons would pass completely through the brick unal-tered [13]. Most of the photons would interact withthe atoms of lead. Of those that interacted, somewould do so by a process called the photoelectric ef-fect, which involves both an atomic electron and thenucleus of the lead atom. In this process, the pho-ton completely disappears. It is totally absorbed or“stopped” by the lead, its energy transferred to thenucleus and a fast-moving atomic electron. Othergamma rays passing through the lead brick wouldinteract by a process called scattering (or moreproperly Compton scattering, after A H Compton,its discoverer). In this process, the photon strikesone of the atomic electrons surrounding the atom,the gamma ray is deflected from its original di-rection and continues in a new direction with re-duced energy. The bigger the angle of deflection, themore energy the gamma ray will have lost. A pho-ton undergoing such a collision is said to have scat-tered. In lead, the two processes—complete absorp-tion or stopping, and scattering—are both likely. Insoft tissue, complete absorption almost never oc-curs. Instead, essentially all interactions result in thephoton scattering. Even in bone, 511-keV photonsare absorbed only rarely. Instead they most simplyscatter.

Now consider photons emitted by a small regionof myocardium. Some small fraction of the anni-hilation photons will be headed in a direction suchthat both photons would strike a detector in thering. As these photons travel toward the detector,they must pass through the tissue of the body. Ifeither of the photons scatters, it will no longer beheaded toward the detector. In all probability it willmiss the ring entirely, or on those occasions thatit does not, its energy may be too reduced to bedetected. A coincident event that would have oc-curred in the absence of intervening tissue, nowdoes not occur. The photons emanating from thissmall section of the myocardium are said to havebeen attenuated, and the loss of detected events dueto interactions with atoms of the intervening tissueis called attenuation. The number of photons thatmake it through unscathed decreases exponentially




with the thickness (d) of interposed tissue:

Number of photons reaching the detector

= Number of photons headed for detector × e−µd

The constant µ is the attenuation coefficient andhas a value of ∼0.096 cm−1 for 511-keV photonsin soft tissue. As can be seen by applying the equa-tion above, only half of the photons will make itthrough 7.2 cm of tissue. Lower energy protonssuch as those emitted by 99mTc (140 keV) are at-tenuated more easily, because µ is higher at lowerenergies. It takes only 4.6 cm of tissue to stop halfthe photons of 99mTc from reaching their originaldestination. It would, therefore, seem that attenu-ation would be much more significant for SPECTscans than for PET scans, because the lower energy99mTc photons used for SPECT scintigraphy are somuch more easily attenuated. This presumption is,however, incorrect. In a PET scan, both photonsin a pair must reach their respective detectors. Asillustrated in Figure 1.6a, a photon headed towarddetector D1 must travel through a thickness of tis-sue X1 without interaction, and the photon goingin the opposite direction must travel through thick-ness X2 to reach detector D2. The total attenuationis then:

Number of coincidences

= (Number of photons headed for D1 and D2)

× (probability photons gets to D1)

× (probability photon gets to D2)


× e−µ(X 1) × e−µ(X 2)


× e−µ(X 1+X 2)

= (Number of photons headed for detector)

D1 and D2 × e−µD

where D is the total distance, X1 + X2, through thebody.

Attenuation

(a)

D1 α e – µX1

D2 α e – µX2

Coinc α e – µX1 . e – µX2

α e – µ(X1 + X2)

X1 X2

D2

(X1 + X2) = thickness of patient. So it does notmatter how deep the tracer is in body!

D1

Attenuation correction

(b)

D1 α SD2 α Se – µ(X1 + X2)

# Coinc = ( )e – µ(X1 + X2)

D2D1S

X1 + X2

Figure 1.6 (a) The attenuation suffered by the pair ofphotons does not depend on where in the body that pairof photons originated. No matter where they originate,together they have to traverse a thickness X1 + X2 oftissue. (b) Because of this, the same attenuation isexperienced by a radioactive source placed outside thebody, permitting the attenuation to be measured. Onesimply measures the number of pairs of photons detectedwithout the patient and compares this to the numberdetected when the patient is present.

Therefore the attenuation of the pair of photonsdepends only on D, the total amount of tissue thepair of photons has to traverse. It does not dependon where in the body the annihilation occurred.One does not need to know X1 and X2, only theattenuation resulting from its sum, D. This is notthe situation for SPECT, in which only one photonis detected. In the SPECT case one needs to knowwhat depth (e.g., X1) the photon came from. Thisis a piece of information that is usually unknownand unmeasurable.

For typical chest thicknesses most of the photonswill be attenuated. Often only ∼10–30% willmake it through the body. Photons traveling in




other directions toward other detectors and thoseoriginating from other sections of the myocardiummay be more or less extensively attenuated. In a70-kg subject, attenuation by factors of 5–20 is notuncommon. Attenuation can be even greater inobese subjects. Obviously, attenuation has signif-icant effects on the results of cardiac PET scans.Although this problem is serious with PET becauseboth photons in a pair must survive intact, accurateattenuation correction is possible. In contrast, withmethods such as SPECT in which only one photonis involved, such correction is not possible becausethe attenuation correction factor, e−µ(X1), dependson measurement of the depth at which the isotopeis located in tissue. This measurement cannot bemade before imaging. The value necessary forattenuation correction of PET images, however, is(X1 + X2). This quantity is independent of howdeep the isotope is located in the body and dependsonly upon the attenuation through the total bodythickness, which can be measured accurately. Themost common method for making this measure-ment involves performance of a “blank” scan andan “attenuation” scan (often called a transmissionscan). Figure 1.6b illustrates this approach for onedetector pair [7,14]. Before the patient to be imagedis placed in the ring, a small positron-emittingsource is placed at one side (as in Figure 1.6b)and the number of photons detected is recorded(just as in Figure 1.6b but without the patient).The position of the source is maintained, and thepatient is positioned in the ring (before injection ofthe isotope) as in Figure 1.6b. Again, the numberof photons are recorded. The difference in the

Rotatingrod

source

(a) (b)

Rotatingrod

source

“Blank” scan Attenuated image= blank * e–µx

Figure 1.7 Illustrating [7] how theprocess described in Figure 1.6 can beimplemented. In panel (a) the “blank”scan is taken with a source and nopatient. In panel (b) the same source isused with the patient in place. For everypossible pair of detectors the ratio ofdetected events with and without thepatient is compared to compute theattenuation correction factor.

counts detected in the blank and transmissionscans is of course caused by attenuation throughthe patient. For example, if S coincident countswere recorded by the detector pair shown in Figure1.6b before the patient was in place, then S × e−µD

counts would be recorded by the same detectorpair when the patient was interposed. The ratio ofthe counts without the patient (called the “blank”scan counts) to the counts with the patient in place(called the “transmission” scan counts) gives thefactor e+µD, which is the factor needed to correctfor attenuation for this particular detector pair.Making the same measurement for all detectorpairs permits complete attenuation correction.

In order to make the measurement of attenuationfor all detector pairs, often a rod of activity is usedwith its long dimension oriented along the Z-axis(Figure 1.7). Such a rod is attached to a mechanismthat rotates it at a fixed speed around the gantry.The rod is usually filled with a relatively long livedpositron emitter (68Ge which decays to 68Ga). Therod is first made to rotate without the patient, giv-ing the “blank” scan counts (Figure 1.7a). Thenthe patient is positioned and the measurement re-peated (Figure 1.7b) giving the transmission scan.The ratio of the counts in the blank to the countsin the transmission for every detector pair gives theattenuation correction factor for that detector pair.

As the rod rotates, only those detectors that lieon the line formed by the detector, the rod, and theopposing detector can possibly be in coincidence.By turning on only the appropriate detectors as therod rotates around the gantry, most accidental andscatter coincidences can be eliminated. With proper




correction software, this also permits the transmis-sion measurement to be made even after activityhas been injected [15,16]. Some manufacturers useinstead an isotope which emits only a single pho-ton (e.g., 137Cs). This makes it more difficult toremove scattered events, but in general both meth-ods have been shown to work satisfactorily. Manymodern PET scanners have been combined witha CT scanner, and in this case the CT scan (aftersuitable processing) can be used to perform attenu-ation correction. This will be discussed further laterin this chapter.

A typical scanning sequence is: (1) obtain a blankscan (usually only done once per day or week); (2)for static FDG imaging, inject the patient, wait foruptake (typically 1 hour) then position the patientin the gantry and obtain a transmission scan ei-ther immediately prior or immediately followingthe emission scan. This minimizes motion betweenthe transmission and emission scan. For dynamicscanning, the transmission scan must be taken priorto injection (or it can be done following the scan).As mentioned above, to perform the transmissionscan after injection requires that hardware and soft-ware be available for correcting for emission activitypresent during the transmission scan. This is fairlystraightforward when positron emitters (and co-incidence detection) are used for the transmissionsource. It is often more difficult when single pho-ton emitters are used for the transmission source.No matter what scheme is used, it is important toprevent patient motion between the transmissionand emission scans. Such motion can produce ap-preciable errors in the uptake image [17].

Scatter

When annihilation photons pass through tissue,they frequently collide with electrons and scatter.The photon is deflected from its original directionand loses some fraction of its energy. The higherthe angle of deflection, the greater the energy loss.The great majority of scattered photons never reacha detector, as illustrated in Figure 1.8. A small per-centage of scattered photons, however, may still hita detector in the ring and register coincidences, asshown in Figure 1.8. When this occurs, the PETcamera erroneously computes the position of theradioactive atom (dotted line in Figure 1.8). Such

Scatteredcoincidence

Scatter

*

Figure 1.8 Illustrating the effect of scattered radiation. Ifone of the pair of photons originating in the heart (shownas an asterisk in the free wall of the myocardium) scattersand is detected (as shown at about the 1-pm position inthe detector ring), the PET scanner will erroneously thinkthe positron was emitted along the dotted line. A“scattered coincidence” will have occurred.

mispositioning of events can cause false counts toappear in cold areas of an image when a hot re-gion is nearby. In general, the phenomenon slightlyblurs sources of radioactivity from hot regions intocold regions (even those a few cm away). This isof particular importance in cardiac imaging, sincethe observer is frequently trying to detect defectsof uptake in segments of myocardium adjacent tonormal regions (and perhaps adjacent to a hotliver).

Most PET scanners are designed to reduce theeffects of scattered photons by rejecting thosephotons whose energy is below a certain thresholdvalue. In most older, and some newer generation,scanners bismuth germinate (BGO) crystals areused. The energy resolution of these detectors is notvery good, making it more difficult to reject scatter.Other crystal types (lutetium oxyorthosilicate(LSO) or gadolinium oxyorthosilicate (GSO)) intheory have better energy resolution, but in practicehave yet to achieve much better scatter energy rejec-tion than the new generation BGO scanners. In PETscanners with septa (2D scanners), scatter is usuallyfairly small anyway. However, as will be seen later,




scanners without septa (3D mode) have severaltimes higher scatter, and so the problem is more se-vere. If a scanner operates with an energy thresholdof 360 keV, it can reject all photons that have beenscattered by more than about 57◦, but not thosescattered less than this. Because photons are morelikely to scatter at small angles, a large number ofscattered photons will still be detected. Attempts toraise the energy threshold to, for example, 400 keV(as is being done in some of the scanners withLSO or especially GSO crystals) would result in therejection of photons that had scattered by morethan about 44◦, but of course the higher one raisesthe threshold, the larger the fraction of unscatteredphotons that are rejected as well. Energy rejectioncan, therefore, be used to reduce large-anglescattering, but can only eliminate smaller anglescattering at the expense of eliminating unscatteredphotons as well. This situation will improve if theenergy resolution of the scanner can be improved.Meanwhile, more empirical methods must beapplied to correct for the remaining scatter. Mostmodern scanners have relatively sophisticatedalgorithms for correcting for this residual scatter[25]. However, especially for cardiac imaging(with its mixture of lungs and adjacent soft tissue)the algorithms are not perfect. The situation forsepta-out (3D) imaging is more problematic, andfor this and other reasons (as discussed furtherbelow), there is still discussion as to whether septaout imaging might be a poorer choice than 2D(septa-in) imaging for cardiac [9–12].

Deadtime

Quantitatively accurate PET studies require that thenumber of true coincidences be directly propor-tional to the concentration of radioactivity. In ad-dition to physical phenomena such as scatter andaccidental coincidences, a significant electronic ef-fect in PET cameras can alter this relationship. Everytime a photon produces a scintillation in a detec-tor, a complex series of electronic events must oc-cur: The light must be converted into an electronicpulse; the exact time of occurrence of the electronicpulse must be determined for use in timing co-incidence; and the magnitude of the pulse must becomputed to allow rejection of scattered events, etc.All of this takes time. If a second photon should ar-

rive before the processing of the previous pulse iscomplete, the second pulse may be lost. There is,therefore, a time interval after a photon has inter-acted with a crystal during which the PET scannerelectronics may be unable to process further pulses.Pulses that occur during this interval, termed the“deadtime,” are lost. The higher the count rate, thelarger will be the fraction of lost pulses. The num-ber of coincidences/second at first increases linearlywith activity, but at high activities it deviates fromlinearity due to this deadtime. Successive increasesin activity produce successively smaller increases incoincidence rate.

The principal source of deadtime is often notthe number of coincident events the machine mustprocess per se, but rather the rate at which the sys-tem must process single photons (each one of whichmust be analyzed to see if it meets the energy re-quirements and to see if it is in temporal “coinci-dence” with another photon, etc.). The singles raterecorded by a detector is often one or more or-ders of magnitude greater than the coincident rate.Often the deadtime loss of a detector can be pre-dicted quite accurately as a function of the singlesrate measured by the detector. This relationship isthe basis for one effective method for correctingfor deadtime. The corrections can be quite large,especially with imaging techniques that require bo-lus injections of isotope (e.g., 82Rb). It is probablybest to limit the amount of activity injected so thatthe required deadtime correction during imagingwill be less than a factor of 2. Activity levels greaterthan this will result in increased radiation expo-sure to the patient without a comparable increasein true coincidences. In addition, the accuracy oflarger correction factors may be suspect.

In many circumstances cardiac PET studies areespecially susceptible to the effects of deadtime,particularly with septa-out (3D) scanners. This isparticularly true with dynamic cardiac studies thatattempt to measure the wash-in or wash-out of ac-tivity from the myocardium, or to measure arterialactivity as a function of time by monitoring theactivity in the atrial or ventricular cavities. During abolus injection or even during a 1-minute infusionof isotope, the PET camera field of view maycontain a large fraction of the entire injected dose.This is in marked contrast to the 60 minutes postin-jection static cardiac scans in which only a small




percentage of the injected dose is in the field of view.The PET camera’s deadtime characteristics (as wellas random coincidences) may sometimes limit theamount of activity that can be administered. Again,the problem is far more severe when operating PETscanners in septa-out (3D) mode than in septa-in(2D) mode.

Resolution

The term “resolution” is one of many parametersused to characterize PET scanners. The term re-quires more careful definition. The spatial resolu-tion of a PET scanner is a measure of how well thescanner can distinguish two small objects placedclosely together. Certain standard measurements ofresolution have been adopted. With one, a verysmall spot of radioactivity is placed in the scan-ner’s field of view and is imaged. If the range of thepositron is very small (e.g., that of a positron froman isotope such as 18F embedded in plastic or alu-minum), then comparison of the apparent size ofthe object in the image and the actual size of the ob-ject allows calculation of the scanner’s resolution.However, very small point sources of radioactivematerial are hard to construct. Instead, often a thinrod of radioactive material, for example a long thinneedle or capillary tube filled with 18F, may be used.Steel prevents the positrons from leaving the nee-dle. The needle or rod is placed in the scanner, withits long axis perpendicular to the plane of the ringas shown in Figure 1.9a. Data are acquired and theimage is reconstructed as shown in Figure 1.9b. Thetop right of the figure shows a plot of the number ofcoincident events as a function of distance across theimage. Typically such a plot follows a bell-shaped,approximately Gaussian curve. By convention, thewidth of this curve at half its maximum height (fullwidth at half maximum, or FWHM) is used as ameasure of spatial resolution. Since the initial mea-surement is obtained within one slice, or plane, itis called the “in-plane” resolution of the scanner.The in-plane resolution will usually be somewhatlarger (perhaps a few millimeters or so, dependingon the scanner) when the measurement is made atthe edge of the field of view, rather than at the cen-ter. Because the free wall or apex of the myocardiummay be 10 cm or more from the center of the fieldin a cardiac PET study, it is useful to know the PET

scanner’s resolution not just at the center, but also10 or 15 cm from the center. In addition, at a givendistance from the center of the scanner’s field ofview, the resolution in the anterior–posterior or Ydirection may not be the same as that in the lateral,or X direction.

The scanner shown in Figure 1.9a is made up ofcrystals with a width, W length, L, in the axial orZ direction, and a depth, D. As pointed out previ-ously the width, W, of the detectors influences thein-plane resolution. Similarly the length, L, of thedetector in the Z direction determines, in part,the resolution of the scanner in the axial direction.To measure the resolution in this direction, a small“dot” of radioactive material, placed on the bed ofthe gantry, might be used. An image could be madeof this dot of activity, the source could be movedthrough the gantry by 1 or 2 mm, and a second im-age made. Progressively moving the source of activ-ity through the scanner in 1 or 2 mm steps, makingan image at each location, would result in a seriesof images as a function of the Z-axis position of thesource. Plotting the number of coincident eventsin each image as a function of the Z-axis position(clinically, the bed position), would produce a plotsimilar to those in Figure 1.9b. The FWHM reso-lution in the Z-axis direction is sometimes calledthe “slice thickness” because it is a measure of howfar into the Z-axis the slice extends. A PET scanner,then, has at least two (possibly very different) spa-tial resolutions: The in-plane resolution (made upof the resolution in the anterior–posterior directionand the lateral directions,) and the Z-axis, or axialresolution.

The axial resolution, or slice thickness, shouldnot be confused with the separation between slices.The spacing between slices may be greater orless than the “thickness” (i.e., FWHM in the ax-ial direction) of each slice. If the spacing betweenslices is less than the thickness of the slice, then theslices may be considered to partially overlap. Evenif the spacing between slices is greater than the slicethickness, some overlap will be present because the“edges” of a slice are not sharp but are Gaussianshaped.

The resolution of a PET image is determined by:The design of the machine (including crystal sizeand spacing, ring diameter, among other factors);physical factors such as the finite range of positrons




Length

(a)

y

x

Radioactiveneedle

Depth

z

Width w

l

d

# C

oin

cid

ence

s

FWHM(b)

FWHM

x

# C

oin

cid

ence

s

y

max

max2

Figure 1.9 (a) Showing placement of rod source to measure in-plane resolution [7]. (b) Upper left: a transaxial image ofthe rod source. Upper right and lower left: profiles through the transaxial image.




in tissue and the deviation of annihilation photonsfrom exact colinearity, and processing, includingwhatever smoothing is performed during or afterreconstruction [19–23]. The effects of image pro-cessing are to some extent controllable. Positronrange is of course a function of the isotope used.Its effects can be estimated as follows. If the num-ber of positrons detected is plotted as a functionof distance in tissue from the source, the numberdecreases almost exponentially with distance [20].Some of the positrons therefore travel relatively far,altering the resolution curve from its usual Gaus-sian shape. The resolution curve produced by ra-dioisotopes emitting very energetic positrons is acombination of the typical Gaussian curve illus-trated in Figure 1.9b and the approximately expo-nential curve associated with positron penetration[21]. Therefore, the curve is roughly Gaussian inshape near the center, but exhibits a long, roughlyexponential, tail. The amount of degradation in res-olution that would occur with use of a positron witha relatively long range in tissue, such as 82Rb, canbe estimated as follows:

Final resolution = (R2 + 1.89 × (2 × D)2)0.5

(1.1)

where R is the resolution of the scanner (includingany smoothing) measured with a nearly zero-rangepositron source (e.g., when 18F in a thin steel needleis used) and D is the average distance the positrontravels, as shown in Table 1.1.

For example, with a scanner having 7-mm use-able resolution as measured with 18F, the resolu-tion expected with the use of 82Rb is based onthe average distance an 82Rb positron travels (D),2.3 mm. The resolution of 82Rb scan can be cal-culated from the equation above to yield a finalresolution of approximately 9.4 mm FWHM—asignificant increase compared with that of a lowerenergy positron emitter. The factor of 1.89 is en-tered into Equation (1.1) in consideration of thefact that the number of positrons decreases withdistance in an exponential rather than a Gaussianmanner. Because the resolution curve is not Gaus-sian with an isotope such as 82Rb, specifying theFWHM does not tell the full story. The numberof positrons decreases exponentially with distancefrom the source, so many positrons will travel muchfarther than the average. Some 82Rb positrons willtravel more than a centimeter before annihilating.

This produces an exponential tail on the resolutioncurve, in turn causing a small fraction of the countsin one part of an image to blur into other, distantparts of the image. To describe this effect, the fullwidth at tenth maximum (FWTM) is measured inaddition to FWHM.

To reduce the point-to-point random statisti-cal fluctuations (called “noise”) that are invari-ably present in a PET image, an image is often“smoothed” by averaging adjacent picture element(pixel) values together. Although this reduces imagenoise, it degrades resolution.

Various filters can be used at the time of recon-struction to facilitate smoothing. “Filtering” is thename given to the process of averaging neighbor-ing pixels together [22] by replacing a pixel valuewith a weighted average of itself and its neighbors.For example, one commonly used filter replaces apixel value with one-half times its own value plusone-eighth times each of its four nearest neigh-bors’ values, so that the weighting factors for thisfilter would be 1/2 and 1/8. Such a filter will pro-duce a less noisy image, but one with poorer spatialresolution. Filters are often given names (e.g., the“Hanning” and “Butterworth” filters). Despite theirspecialized names, all filters do nothing more thanaverage neighboring pixels together; they differ onlyin their weighting factors, which may be positive ornegative.

In addition to filters that reduce noise but worsenresolution, filters exist that improve resolution andexaggerate noise. Unfortunately, it is a consequenceof the basic laws of physics that it is impossible to si-multaneously reduce noise and improve resolution,and because of statistical fluctuations caused by thelimited numbers of coincident events, PET imagesalmost always must be filtered with a smoothing,rather than a resolution-improving, filter. Some 3Dsmall animal PET scanners are the exceptions tothis rule. In these scanners there are often plenty ofcoincident events, because the dose/gm injected ishigh (dosimetry is frequently not a limiting factorin animal imaging). Resolution recovery can thenbe used, at the expense of slightly worsening statis-tical fluctuations, to achieve the higher resolutionoften required for small animal imaging. Some res-olution recovery techniques are most easily built-into so-called “iterative reconstruction” programs.

Clinical PET scanner software usually gives theinvestigator a choice of which smoothing filter to

P1: SFK/UKS P2: SFK Color: 4C Cardiac CT, PET and MR · 2016-08-12 · Antti Saraste, Hossam Sherif...

Documents

Transcript of P1: SFK/UKS P2: SFK Color: 4C Cardiac CT, PET and MR · 2016-08-12 · Antti Saraste, Hossam Sherif...