__ r-.....::"'; . .,.-...- __ ..;:.;..:.,..:...;:..:--z-- __

Introduction 1O. Ratib and M. Schwaiger

-t>

Contents PET/MRI in Clinical Practice

PETfMRl in Clinieal Practiee .. Following the path of elinical applications of hybrid PET/CT it would see m logical that PETlMR can provide an inno-vative and attractive alterrative taking full advantage of thesuperiority of MR over CT in differentiating soft tissue char-acteristics with a reduction in radiation exposure by replac-ing CT with MR imaging. The challenges of PET/MR arestili numerous both on the tcchnical side as on the practicaland elinical side.

From PET/CT to PETfMR ...

Potentials and Challenges of PETfMR ..

The Challenge of Hyhrid Irnaging Protocols ...

Domains of Clinical Applicatios of PETfMR .

Future of PETfMR in Oncology .

4

4

4

Furthcr Reading ...

From PET/CTto PET/MR

The development of PET/MR started even before the firstprototypes of PET/CT were developed. As deseribed indetails in the next chapter, the technical challenges were con-siderable to overcome the interference and cross-talk effectsbetween magnetic field and the photomultipliers of PETdetectors. Several alternatives have emerged and lead to thefirst hybrid devices to appear on the market for elinical appli-catiors. Whether it is through a combination of separatecoplanar systems or by integrating solid-state PET detectorsinside an MR! they provide the means to explore the poten-tial elinical applications of whole-body PET/MR in elinical

practice.Given the broad ran ge of applications of PET imaging,

it is oneology that remains today the elinical domain wherePET is mostly used. PET imaging was shown to be supe-rior to other imaging techniques in staging and follow-upof numerous specific tumors. The advent of PET/CT hasreinforced the elinical utilization of PET by a!lowing com-bined PET and CT studies to be acquired quasi-simultane-ously with perfect alignment of anatomical and metabolicimaging data. While most elinical studies showed rela-tively modest improvement in diagnostic accuracy throughthe sensitivity and specificity of hybrid PET/CT over PET

O. Ratib(tSl)Department of Medical Imaging and Information Sciences.Divisior of Nuclear Medicine and Molecular lmaging,Geneva University Hospitals, Geneva. Switzerlande-mail: osman.ratib@hcuge.ch

M. SchwaigerKlinikum rechts der IsarNuklearmedizinische Klinik u. Poliklinik, Techrische UniversitatMrchen, Munich, Germany

::'!l!r~~O. Ratib et aL.(eds.). Arlas of PET/MR lnaging in Oncology,DO! 10.1007/978-3-642-31292-2_1, Springer-Verlag Berlin Heidelberg 2013

4O. Ratib and M. 5chwaiger

and CT performed separately, several studies demonstratedhowever a significant improvement in diagrostic confidencewhen both studies were acquired together and images wereinterpreted with fusion of both modalities. Uncertainties ofdiagnostic findings that could occur when PET image s areinterpreted without accurate anatomical localizatior canbe avoided when CT images are co-registered with PETimages and provide the necessary anatemical references.Conversely, diagnostic criteria of CT based on structural andmorphological observations can be of ten misleading andsubject to difficult interpretation and can be significantlyimproved when additional metabolic information of PET isprovided with combined PET/CT images. While MRI hasreplaced CT in many elinical applications by providing abil-ity for higher tissue characterization and funetional irnag-ing, the added value of PETlMR over PET/CT stili remainsto be dernonstrated in elinical practice. The most immediateapplication might be in patients that require both PET/CTand MRI in their elinical workup and could benefit froma single study combining PET and MR when CT is of nosignificant additicnal value.

Patentials and Challenges of PET/MR

Technical designs of hybrid PETlMR devices differ corsid-erably between differert vendors as shown in the next chap-ter of this book. While integrated system my represent themost logical solution they retnain technically more challeng-ing and copianar systems combining two scanners adjacentto each other provide an alternative that resembles in itsdesign current PET/CT systems where the patient is rnovedfrom one to the other modality sequentially. In oneologyapplications both sequential and simultaneous aeqisition ofirnages may provide similar results they differ only on theirnaging protoeols used,

The rernaining challenge of all hybrid PETlMR systemsis the calculation of tissue auenuation correetion maps sirni-lar to those calculated from whole body CT seans in hybridPET/CT devices. Several techniques of attenuation correc-tion were explored, but the emphasis has been mainyon MRimage segrnentaion for dervation of an attenuation map,Besides image segmentalion, other technical challenges foreffective attenuation correetion in a whole-body PET/MR,including compensation for MR image truncation and cor-reetion for RF eoils and aeeessories also need to be irnple-mented. Accurate caleuIation of tissue attenuation airnsmainly toward providing quantitative measurement of PETracer uptake in tissue. Semi-quantitative ealeulation of stan-dard tracer upake (SUV) is the most comman techriqueused today and provides an attractive simple method to

estimate the amount of tracer uptake in different tissues.While it can allow for differentiating between benign physi-ological and potentially pathologieal tissue tracers uptake, ithas only marginal added value in routine elinical practicewhen PET findings are combineo with observations fromother imagirg modalities such as CT and MR and confrontedwith multiple other criteria in elinical decision making. Itrernains however necessary to achieve the best and mostaccurate correction of ussue auenuution to use PET imagingtechnology at its full potential and berefit from the addedvalue of quantitative imaging over visual interpretation of

PET findings.

The Challenge of Hybrid Imaging Protocols

Depending on the scanner design, imaging protocols can dif-fer significantly depending on the ability to perforn sequer-tial or simultaneous acquisitiors of both modalities. ltremains however that one of the most challenging aspect ofhybrid PETlMR is the complexity and heterogeneity of MRprotocols that are being used in elinical practice. The wealthof different types of imaging parameters that MR! providesand the diversity and lack of standardizetion of differentimaging protoeols have lead to significant differences in pro-tocols used in elinical practice. The general trend is that MRimaging protocols have considerably been extended toinclude several sequences for better tissue characterizationand extraction of functional and physiological pararneters ofdifferent issues and organs. The combination for such com-plex protocols together with additicnal whole-body imagingsequences of PET and MRI can lead to significantly longerexaminatior time. Therefore, additional efforts and researchis needed to opimize hybrid imaging protocols to benefitfrom the best potential capabilities of euch modality whilemaintaining reasonable imaging time that is cornpatible wihroutine elinical pracice.

Domains of Clinical Applications of PET/MR

The prirnary elinical applications of PET/MR that we electedto cover in this book are in oncology. But there are otheremerging applications in cardiovascular, n inflammatoryand infectious disease as well as in neurodegenerative dis-eases. MRI has gained a wide adeption in cardiology for thedetection of ischemic disease, myocardial viability ad car-diac function. The added value and complementarity of PETin providing a better sensitivity and quantitative analysts ofmyocardial perfusion and myocardial viability can becornegood elinical justifications for combineo PETlMR studies.

1 Introduction

Further ReadingBrain imaging for acute as well as chronic disease of ten relyon combination of multiple imaging modalities such as PETand MR. In brain imaging however, the rigidity of the headand the easily identifiable skull structures, allow software-based registration technique to provide adequate fusion ofdifferent rnaging modalities. A wider availability of hybridPETlMR can however facilitate the use of hybrid imag-ing for brain studies with more eonvenience to the patientthat does not have to undergo separate studies on different

scanners.Arother factor that favors MRI over CT for hybrid imag-

ing is the reduction -in radiation exposure. Although forelderly patient and patients under palliative treatment forcancer, radiotion exposure may be of minor impact, redueedradiation exposure of PETIMRI compared with PET/CT maybe relevart in non-oneological patients and in youngerpatients with poterially curable disease.

Future of PET/MR in Oncology

This book highlights the potertial applications of whole-body hybrid PETlMR in oncology. While it is stili a collec-tion of convincing aneedotal cases, it only reflects earlyobservations of two acadernic centers that were first in adopt-ing this new technique n elinical praetice. The diversity ofcases and broad scope of elinical domains covered in the dif-ferent chapters of the book underline the potential applica-tions in oneology but alsa in other elinical dornains, The useof different radiolabeled tracers such as 18F-fluorocholineand 'F-ftuorotyrosine show the potential of PET beyondthe conventional 18F_FDG tracer n elinical applieations ofhybrid PET/MR. However this new imaging rnodality hasonly been recently introduced for elinical use and will facethe same challenges and skepticism that PET/CT techniqueeneountered when it was first introduced. The lack of tangi-ble added value of corrbired PET/MR over the two exami-nations acquired separately is the first issue that needs to beaddressed both from a elinical perspective and from a med-ico-ecoromic point of view. it is however foreseeable thatinerensing demand for objective criteria in determination ofadequacy and efficacy of new treatments, in particnlar inoneology. will drive the development of new tracers andinnovative hybrid imaging protocols that take full advantageof complementarity of PET and MR modalities.

Antoch G, Bockisch A (2009) Combined PETIMRl: a new dirnensionin whole-body oneology imaging? Eur J Nucl Med Mol Imaging 36Suppll:SI13-St20

Aruoch G. Saoudi N. Kuehl H et al (2004) Accuracy of whole-bodydual-modality ftuorine-18-2-ftuoro-2-deoxy-O-glucose positronernissiou tomography and computed tomography (FDG-PET/CT)

~ for tumor staging in solid tumors: comparison with cr and PET. J Clin Oncol 22:4357-4368

Bar-Shalom R. Yefrernov N. Guralnik L et al (2003) Clinicalperfonnanee of PET/cr in evaluntion of cancer: additicnal valuefor diagnostic imaging and parient management. J Nud Med 44:1200-1209

Beyer T. Pichler B (2009) A decade of conbined imaging: from a PETattached to a CT to a PET inside an MR. Eur 1 Nucl Med Mollmaging 36 Suppl I:S I-S2

Beyer T. Weigert M. Quick HH et al (2008) Mk-based aueruauon cor-rectior for torso-PETI1vIR imaging: pitfalls in mapping MR to CTdata. Eur J Nucl Med Mol Imaging 35: 1142-1146

Che n W. Jian W, Li HT e al (2010) Whole-body diffusion-weighedimaging vs. FDG4PET for the detection of non-small-cell lung can-cer. How do they measure up? Magn Reson Imaging 28:613-620

Collins CD (2007) PET/cr n oncology: for which tumours is it thereferenee standard?Cancer Imaging 7 Spec No A:S77-S87

Czemin J. Allen-Auerbach M. Schelbert HR (2007) Imprcvements incancer staging with PET/CT: literaure-based evidence as ofSeptember 2006. J Nucl Med 48 Suppl I :78S-88S

De1so G. Ziegler S (2009) PET/MRI system design. Eur J Nucl MedMol Imaging 36 Suppl t:S86-S92

Hcusncr TA, Kuemmel S, Umutlu L et al (2008) Breas canccr stagingin a single session: whole-body PET/cr mammography. J NuclMed 49: 1215-1222

Heusner TA. Kuemmel S. Kocninger A el al (2010) Diagnostic value ofdiffusion-weighted rragnetic resonance imuging (DWl) comparedto FDG PET/cr for whole-body breast cancer staging. Eur J NuclMed Mol Imaging 37(6): 1017-1086

Hofmann M, Pichler B. Seholkopf S. Beyer T (2009) Towards quantita-tive PET/MRI: a review of Mp-based attenuation correction tech-niques. Eur J Nucl Med Mol Imaging 36 Suppl I:S93-S104

Hu Z. Ojha N. Renisch S et al (2009) Mfc-based attenuarion correc-tion for a whole-body sequential PET/MR system. In: lEEEnuelear science symposium conferenee record. Ortande. 2009. pp3508-3512

Lonsdale MN. BeyerT (20 W) Dual-rrodality PET/cr irstrurnentatior-today and tomorrow. Eur J Radio! 73:452-6O

Punwani S, Taylor SA. Bainbridge A el al (2010) Pediatrie und adoles-cent Iymphoma: comparison of whole-body STIR half-FourierRARE MR imaging with un enhaneed PET/cr reference for initialstaging. Radiology 255: 182-190

Schmidt GP. Reiser MF. Baur-Melnyk A (2009) Whole-body MRI forthe staging and foJlow-up of patients with metastusis. Eur J Rudiol70:393-400

Shao Y, Cherry SR. Farahani K et al (1997) Simultaneous PET and MRirnaging. Phys Med Bio142: 1965-1970

2PET/MRInstrumentationT. Beyer, O. Mawlawi, and H.H. Quick

ContentsIntroduction ...

PET/MR Design Concepts ....

MR-Compatiblc PET Detcctors

PET/MR Methodological Pilfallsand Tcchnological Challengcs ..

PET/MR Safety

~

Introduction

14Anato-metabolic Imaging

Summary and Conclusion ..

T. Beyer(BI)Center for Medical Physics and Biornedlcal Engineering,General Hospital Vienna. Medical University Vienna.4L Waehringer Guertel 18-20. 1090 Vienna. Austriae-mail: thomas.beyer@meduniwien.ac.at

O. MawlawiDepartment of Imaging Physics.MD Anderson Cancer Center. Uuit 1352. Houston. TX 77030. USA

H.H. QuickInstitute of Medical Physics (IMP),Friedrich-Alexnnder-University (FAU) Erlangen-Nrnberg,Henkestr, 91. 91052 Erlangen. Germany

18 Most people require diagrostic tests during their lifetime inorder to detect a suspected malignancy, plan a therapy andfollow-up on a treatment. In almosr all of these cases diag-nostic tests ertail a single imaging exarnination or a series ofcomplementary imaging exarns, Nor-invasive irnaging iscentral to personalized disease managernert and includesimaging technologies such as Computcd Tomography (CT),Single Photon Emission Computed Tomography (SPECT),Magnetic Resonance !maging (MR!). Ultrasound (US) orPositron Emissior Tomography (PET).

Each of the above imaging tests yields a wealth of infor-mation that can be separated generally into anatemical andrnetabolic information. Anatemical information, such asobtained from CT or US. is represented by a set of sub-mmresolution images that depiet gross anatamy for organ andtissue delineation. Malignart disease is typically detected onthese images by rneans of locally altered image contrast orby ab normal deviatiors from standard human aratorny. lt isimportant to note, that anatomical changes do not necessarilyrelate LO the orset of rnalignan diseases. In other words,maligrunt diseases are expressed as abnormal alterations ofsignaling or metabolic pathways that may lead to detectableanatomical changes. Therefore, anatomical imaging aloneOlay rniss diseases frequently or diagnose diseases at anadvanced stage only.

PET. as a representative of nuclear medicine imagingmethods. has been shown to support accurate diagnosis ofmalignant disease [] as well as providing essertial informa-tion for early diagnosis of neurodegenerative diseases [2]and malfunctions of the cardiovascular system [3]. However,over 90 % of all PET examinations are performed for oncol-ogy indications. PET is based on the lise of race amounts ofradioactively labeled biomolecules. such as [18F]-FDG, afluorine-l S labeled analogue to the glucose molecule, thatare injected into the patient whereby the distribution of the

21

24

25

27

..~~ _ o. Ratib e al. (cds.). Atlas of PET/MR lnaging i Ocoiogy,____ DOI 10.1007/978-3-641-31292-1_2. Springcr- Verlag Berlin Heidelberg 2013

T. Beyer et aL.

p.n+e+v.~ e + range~"", -,;'-/ ~/"',/ ',~i N~: Ccincidence 0i!!i "i r:l _~-----t:.J

np p

p P8nJpC!0

Sinograms

Reconstructed image

Fig. 2. Schernattcs of PET imaging: u biornoleeule is labeled with apositron emilter (e.g .. IlIF, Tn - 109.8 min) and injccted into the patients.The radioactive isoope labei decays by emining a positron. which arni-hilutes with an electron from the surrounding tissue. hus creating two

racer is followed by detecting the annihilation photonsresulring from the ernission and annihilation of the positrons(Fig.2.1).

in most cases of malignant diseases early diagnosis is keyand, therefore, imaging the anatamy of a patient may notsuffice in rendering a correct and imely diagnosis. Thus, med-ical doctors rypically employ a combination of imaging tech-niques during the course ofdiagnosis and subsequent treatmentto monitor their patients. Henceforth, both functional and ana-tomical information are essential in state-of-the-art patientmanagement. An appreciation for this type of combined infor-mation is best illustraed with the introduction of the term"anato-rnetabolic imaging' [4], in reference to an ideal imag-ing rnodality that gathers both anatomical and functionalinformation, preferably within the same examination.

Historically, medical devices to image either anatomicalstructure or functional processes have developed along same-what independent paths, The recognition that combiningimages from different modalities can offer significant diagros-tic advantages gaye rise to sophisticated software techriquesto co-register (aka align, fuse, superimpose) structure andfunction retrospectively (Fig. 2.2). The usefulness of combi n-ing anaornical and functional planar images was evident tophysicians as early as in the 1960s [5]. Sophisticated imagefusion software was developed from the Iate 1980s onwards.

HO-CH,

Neutron delicientisotope

18F-FDG glucose

~/--;./. ,,",-~

..,,/7" /-'/~/

PET tomograph

annihilation photous ha are ernitted hack-to-back and detected by aring of PET detectcrs. Image reconstruction then follows the same prir-ciples as in CT (Courtesy of David W Townsend. Singapore)

For relauvely rigid objects such as the brair, software can suc-cessfully align images from MR, CT and PET, whereas inmore flexible enviranments. such as the rest of the body, accu-rate spatial alignment is difficult owing to the large number ofpossible degrees of freedoru. Alternatives to software-besedfusion have now becorne available through instrumentationthat combines two complementary imaging modalities withina single system. an approach that has since been termed hard-ware fusion. A cornbined, or hybrid, tomograph such as PET/CT can acquire co-registered structural and functional infor-mation within a single study. The data are complementaryallowing cr to accurately localize functional abnonnalitiesand PET to highlight areas of ab normal metabolism.

The advantages of integrated, anato-rnetabolic imagingare manifold [6]. A single imaging exarnination providescomprehensive information on the state of a disease.Consequently. funetional information is gathered and di s-played in an anatornical context. Patients are invited for onlyone, instead of multiple exams. As shown by several groups,the combination of complementary imaging modalities canyield syergy effects for the acquisition and processing ofimage data [7. 8]. And, firally, expers in radology andnuclear medicine are forced to diseuss and integrate theirknowledge in one report, which will perhaps be more appre-ciared and considered a benefit in the years to come.

2 PET/MR Instrumentation

1960

1990

Semi-automated SW co-registration

2Oj~i.QI~ -~~C~.'1' _~. _ ...:;~'-'t .- . 1~i ~ ~_~

Ji.

Hardware-based fusionPET/MR

1980

Lightbox viewing and film

e.

2000

SPECT/CT

.~~~ . ~!~'~,r-~

;... - '-'':rr- ~.:.~,~~- ~~-~ ... '-..\_~~.

-r~'Hardware-based fusion: PET/CT, SPECT/CT

Fig. 2.2 The history of fusion imaging: from the 19605 to the 19905complementary image information was aligned rnanually and later withthe srpport of computer-based algorithms.With the introduction of Pr'>

PET/CT Imaging

PET imaging has been in elinical practice since the Iate1980s, thus providing valuable information in addition to CTimaging in cases where complementary diagnostic informa-tion was elinically indicated, However, the lack of fine ana-tornical detail in PET images may limit the Iocalization oflesions and permit only a poor definition of lesion boundar-ies. This challenge was overcome by combining high-resolution anatemical cr imaging with PET, thus, providinga hardware combination for "anato-rretabolic" imaging L9].The first proposal to combine PET with CT was made in theearly 1990s by Townserd, Nun and co-workers. The fore-most benefit of a PET/CT hardware combination was theintrinsic alignment of complementary image information,further supported by a elinical need at the time. A secondaryberefit of this combination came with the ability to use thecr irnages to derive the required PET attenuation correctionfactors, one of the pre-requisites for quantitative PET imag-

totype SPECTICT and PET/cr imaging in the 1990s and PET/MRimaging systcms in the mid 2000s the field of hardware image fusionwas changed drarratically

ing [IOJ. CT-based attenuation correction has now becornethe standard in all PET/CT ornographs [ll] despite the factthat same assumptions have to be made in order to transformthe attcruation vaIues of human tissues at cr energies (e.g.effective CT energies are on the order of 60-90 keV) toattenuation coefficients at the PET energy of 511 keV [12,13]. Figure 2.3 illustrates the main drivers for PET/CT:anato-rnetabolic alignment and CT-based attenuationcorrection,

Following the introduction and validatior of the firstwhole-body PET/CT prototype in 1998 [14] first cornmercialPET/CT concepts were proposed as of 2001 leading to abreadth of 25 different elinical PET/CT systems offered bysix vendors worldwide in 2006. Today, four major vendorsoffer a range of whole-body PET/CT syserns with greatlyimproved functionalities forboh, PETand CT [151. Table 2.1sumrnarizes the state-of-the-art PET/CT technology. In brief,all PET/CT systems permit total-body imaging within a sin-gle exarnination while using the available CT image

o

Functioral anatamyHigh functionl resolutionEarly datectlon possible

PET = Emission

Topogram

Fig.2.3 PET and CT can be operared in closc spatial proximity with-out cross-talk degradation of their respective perforrnarce pararneters.(a) PET and CT images provide complementary diagnostic irforma-

information for routine atteruarion and seatter correcrion ofthe PET data [6]. Major technical advances include the incor-porntion of ime-of-flight (TOF) PET acquisition mode [161,the extersion of the axial field-of-view (FOV) of the PET[17] and the ineorporation of system information, such as thevariability of the point spread function across the field-of-view, into the reconstruction process [17].

Time-of-flight-PET was first suggested in the Iate 1960sin order to improve the signal-to-noise ratio (SNR) of thePET data [18]. In essence. TOF-PET requires the measure-ment of the arrival time of tWQ arnihilation photors arisingfrom a given annihilaton; which helps localize the origin ofthe annihilation (i.e. the traeer) beer. TOF-PET requiresrast scintillation detectors and advanced detector electronics(see also seetion "MR-Compatible PET Detectors").In human studies TOF-PET can help increase the SNR by ataeter of 2. Today. still few studies are available that demon-strate a significant diagnostic benefit in routine elinical appli-cations [19, 20], but the options for trading a gain in SNRinto reduced injected activities or into shorter ernission seantimes are available taday.

Extending the axial FOV of a PET system comes at theexperse of more PET detectors to be added in the axialdirecion. However. for a given injected aciviy, more

T. Beyer et a'

~---

CT ;: Transmission

\1'S- -

CT Emission Attn-corr Emission

tion. (b) The use of the cr transmission images for the purpose ofnciseless ntenuation correction of the emission data comes as a second-ary benefit of PET/cr

annihilation photon can be detected. thus. increasing the sys-tem sensitivity by 80 % for an additional 25 % axial cover-age. This gain n sensitivity can be used for redueed emissionsean times or activities injeeted. Despite the required increasein axial bed position overlap, the number of contiguous bedpositiors required to cover a given eo-axiaI imaging range isredueed in case of PET imaging sysems with an extendedaxial FOV.

Parallax errors arising from depth-of-interaction effeetscause the spatial resolution of the PET to be a variart of thespatial location of the annihilation. if the spatial variation ofthe poin-spread-function (PSF) is known apriori. for exam-ple, by means of standardized measurements, it can beincluded in the reconstruction algorithm [17, 21]. The recon-struction process becomes computationally demanding buthelps improve the spatial resolution and renders the varia-tions of the PSF in the images uniform across the field-of-

view,Over the years, the above advances have helped improve

the qualiy and reproducibility of PET and PET/CT data(Fig. 2.4) and suppo1 a routine examination time for a stan-dard whole-body FDG-PET/CT study of 15 min. or less, asignificant advantage when eompared to PET/CT irnaging

from a deeade ago.

2 PET/MR Instrumentation11

Table 2.1 State-of-the-art PET/CT imaging systems GE Healthcare. Philips Healthcare. Mediso and Siemens Healthcare (from left LO riglt). Thefigure shows key parameters and performance measures of the PET/Cl series

Discovery VCT Ingenuity TF AnyScan Biograph mCT

;;;

~

CT: 16-128 slices CT: 16-128 slices 16-sliceCT CT: 20-128

70 cm patient port 70 cm (85 cm) patient port 70 cm diameter patient port 78 cm patient port

250 kg table weight Hmit 215 kg tabla weight limit 250 kg table weight limit 250 kg tabe weight limit

170 cm co-sean range 190 cm co-sean range 360 cm co-sean range 170 cm co-sean range

24 rings of LYSO(Ce) 44 rings of LYSO(Ce) 24 rings of LYSO(Ce) 52 rings of LSO (Ce) crystals

4.2 x 6.3 x 25 mm' 4.0 x 4.0 x 22 mm' 3.9 x 3.9 x 20 mm' 4.0 x 4.0 x 20 mm3

Time-of-flight Time-offlight Time-of-flight

15.1 cm axial FOV 18 cm axial eoverage 23 em axial eoverage 21.6 cm axial eoverage

70 cm transaxial FOV 67 cm transextat FOV 55 cm transaxial FOV 70 cm trarsaxial FOV

PET resolution model PET resolution model PET resolution model

In-plane resolution: 4.9 mm In-plane resolution. 4.7 mm In-plane resolution: 4.1 mm In-plane resolution: 4.4 mm

Axial resolution: 5.6 mm Axial resotution. 4.7 mm Axial resolution: 4.2 mm Axial resolution: 4.4 mm

3D Sensitivity: 7.0 cps/kBq 3D Sensitivity: 7.0 cpsll

T. Beyer et al.12

Fig.2.4 Corenal (top) and transaxial (bottom) view of an whole-body[ISFI-FDG-PET image of a putient with n BMI of 35 acquired in30-node with septa retracted and reconsructed using: (~L)30 filteredback-projection algorithm with reprojecticn (30-FBRP, 7 mm Gauss).(h) elinical reconsrruction using FORE rebinning+20 OSEM (8 sub-

sets. 3 uerations: 5 mm filter), (c) 30 Ordinary Poisson (QP)-OSEMwith PSF reconstruction (14 subsets.2 irerarions: no smoothing), and(d) 3D OP-OSEM wih both PSF and Tine-of-Flight (TOF) reconsruc-lion (14 subsets. 2 iterations. no smocthing) (Case courtesy of DWTowrsend, Singapore)

Fig.2.5 Expectations for PET/MR in the context of the existingexperiences with PET/CT forpauent imaging

C'~~~(Ul~~~U

PET/CT PETlMR

pro

High-resolution anatamy Best passible, intrinsic ca-registratian Quasi-simultaneaus Jmaging Noiseless/fast attenuation carreetion Fast whole body imaging Integrated report

High saft tlssue contrast through MR Simulataneous imaging possible Less ionizing radiation (MR=O) MR upgrade=New MA sequences

con

MRcompatible PET detector MRbased attenuatien ccrrectlon PET/MR design restrictions Patient acceptance Clinical and research apoucattons

Patient exposure from CT Local, motion-induced misalignment Only quasl-simultaneous scanning Hardware Upgrades via fork-Iift Reimbursement for PETonclear

imaging togetler, with the added potential of MR-based motioncorrection of the PET data, significantly reduced patiert expo-sure and a increased soft issue contrast through the use of MRinstead of cr, wherever c1inically indicated.

Soft tissue enhaneement in MR (versus cr) may benefitthe imaging of pediatric patients where normally little fattytissues are present (Fig. 2.6). as well as for studying patientsfor indieations related to the brain, parenchymal organs or the

2 PET/MR Instrumentation

Fig. 2.6 Slde-by-side comparison of cr. MR and PET images of apatient with previously irradiated fibrosarcoma. The tumcur is poorlyvisualised on cr but the MRI shows a residual mass.The PET sbowsresidal moderate FDG-avidity, and resection confirmed residual viable

rnusculoskeletal system. In addition to mueh improved softtissue contrast MR is a versatile imaging rnodality sine e it pro-vides addirional measures of physiologic and metabolic char-aeteristies of human tissue [251. MRI goes beyond plainanatonucal imaging by offering a multitude of endogenouscontrasts and a high capability of differentiating soft tissues, aswell as many exogenous contrast rnedia ranging from gadolin-ium-based agents to highly specified cellular markers [261

MR specroscopy (MRS), for exarnple, can be used to dis-sect the rnolecular composition of tissues by applying seIee-tive radiofrequency exeiration pulses [271. Functionalprocesses in living subjects can also be studied via diffusion-weighted (DWJ) MR! [28]. Here, a spatially and tenporallyvanant magnetic field, generated by different magnetic fieldgradients in allthree spatial directions. is used to rnap phasedifferenccs in the MR! signal that are caused by diffusirgnolecules. DW!-MRI has potential elinical applicationsranging from diagrosing ischemia in early stroke diagnos-ties, cancer, multiple selerosis, or Alzheirner's disease togeneral fiber traeking via diffusion tensor imaging (DT!)[26,29,30], and it is not restricted to the brain [3IJ. In addi-tion, funetional MR! (fMRI) studies can be performed dur-ing the same exarnination. Functional MRI (fMRl) studiesare frequently based on the BOLD (blood oxygen leveldependent) effect [32]. This effect deseribes the fact that thernagneic properties of oxygenated and deoxygenatedhemoglobin in the blood are different and. therefore, produce

13

,

tumour. Lack of soft tissue contrnst. particularly lack of fa in childrencompromises anatemical evaluation on cr compared to MRI (Courtesyof Rod Hicks. Peter MacCal1um Caneer Cenre, Melbourne Australia)

different signals when imaged with T2~-sensitive MRlsequences. The BOLD effeet alsa has certain applications incancer irnaging, such as to study tumor angiogenesis, turnoraxygenation and brain activation in eloqueru areas prior tosurgical resection.

Any of the image information above can be acquired andpresented in any directian in space. thus rendering re-orier-tatior of image information in MR sirrilar to a "virtual tilt",that is available in CT-only in directions perpendicular LO themain scanner axis, and that are not available in PET/CTimaging.

Similar LO CT and PET, MRI has become a whole-bodyimaging rnodality tharks, for example, to the advent of par-allel imaging techniques and alitheir derivalives [33-35] andrhanks LO new whole-body imaging strategies [36, 37]. Imageacquisition times have been shortened, thus allowing fastsingle-contrast MR whole-body eoverage from 30 s [36]ranging to muli-contrast. multi-station whole-body MRIexarrinations to be acquired with high spatial resolution inless than i h. lnitial results show that whole-body MRI is apromising modality in oneology. especiaIly for the deteetionof metastases and hematologic malignancies.

Therefore, MR! holds a great potential in replacing CT asthe complementary rnodality to PET in dual-modality tomo-graphs for selected indications where MR outperforms CTalready. In theory. MRI seems a perfeet anatomical comple-ment to PET.

14

PET/MR Design Concepts

Following the successful adopion of PET/CT in elinicalrouine and the ongoing efforts toward s combining PETand MR! for pre-clinical research applications [241,industry has quickly adopted the idea of combining PETand MR! for human studies. Figure 2.7 summarizes themain approaches towards PET/MR hardware fusion. Inessence. three different design concepts have been pro-posed: separate PET/CT and MR! systems operared inadjacert rooms (a), PET and MR! sysems arranged in thedirectian of the main scanner axis with a patient handlingsystem mounted in between (b) and a fully integratedPET/MRI system (c).

PET/CT-MR Shuttle System

GE Healthcare proposed a straightforward design in Iate2010. This design is based on a combination of a dual-modality, whole-body TOF-PET/CT and a 3 T MR systemthat are operared in adjacent rooms; patients are shuttledfrom one system to the other without getting off the bed[38]. This approach substitutes the challenges of hardwareintegration for immense logisticn\ challerges n timingaccess to the two systems while minimizing patient rnotionin between examinations. The advantage of this rather sim-plistic approach to PET/MR is that it is based on existingimaging technologies without significnnt changes to heirhardware components. Patients urdergo a PET/CT studyleveraging the bercfits of time-of-flight PET as discussedbefore. Following the PET/CT examination patients arethen Iifted on a mobile docking-table system and shuttledto the MR system where a loco-regional or whole-bodyMR study is performed depending on the elinical indica-tion. Figure 2.8 illusttates a elinical ease from the com-bined use of PET/CT and MRI Llsing the PET/CT/MR

system.While this design is stili available as prootype technol-

ogy only, it has been argued alsa as the most cost-effcctivecompared to fully integrated PET/MR based on workflowaspects and machine utilization [39], both of which aresite- and operatiors dependent. Therefore, in pracucethe elinical and cost efficacy of the separate PET/CT/MR design opion (Fig. 2.7a) wou ld be affected by vari-ous workflow and installation requirements. For exarnple,both systerns need to be installed next to each other andoperared within a combined scheduling system. Anydeviation from standard protocols would entail exterdedwaiting times with the patients Iying on the shuttle sys-tem until the next exam can cornmcnce. Also, two or eventhree shule systerns are required to facilitate a seamless,high-throughput workflow. On the upside this approachdoes ensure proper attenuation and seatter correction ofthe PET data based on the available CT information. In

T. Beyer et aL.

Fig.2.7 Design concepts for PETlrvIR: PET/CT and MRI tornographsare operated in adjacen rooms and intertirked with a mobile shuttlesystem Ca). a co-planar PET/MR with a whole-body PET and MR oper-ated in close proximity and a combined table platform (b). and a fuUyintegruted PET/MR wirh MR-compatiblc PET deteetion system slip-fitinto the MR (c) i'

turn, the examination time is likely to be the longest ofall PET/MR designs and patient convenience is limitedby the repositioning in MR or PET/CT using the shule

system.

Co-planar PET/MR

Philips Healthcare proposed a slightly more integratedapproach to PETlMR in 2010 [40J. They also presented thefirst comrnercially available PETlMR system for elinical usecalled the Philips Ingenuity TF PET/MR!. The system(Fig. 2.7b) is based on a co-planar design concept that inte-grates a whole-body time-of-ftight (TOF) PET system and anAchieva 3 T X-series MR system. Both components arejoined by a rotating table platform mounted in between [41].

The PET detector and electronics system is based on avail-able Philips PET/CT technology. However, given the proxim-ity of the PETand MR system (about4 m) some modificationswere required to ensure MR-compatibility of the PET system.These modifications include the addition of bulk magneticshielding of the PET to reduce fringe magnetic fields, the useof higher permeability shields of the photomultiplier tubes

2 PET/MR Instrumentation 15

UniversityHospitalZurichii-

'..:1.. \

o.. ...,

FDGPET/CTFDG-PET

Fig. 2.8 Patiert with large left lung lesion undergoing wholc-bodyFDG.PET/CT and whole-body MR! on the separate PETeTIMR sys-tem (Fig. 2.7a). From left lo right; FDG~PET following CT-based atten-uanon correction (CT~AC). PET af ter TAC fused with whole-body

(PMT) inside the PET gantry and the rotation of the cathodesof the PMTs. Further, power and signal cables penetratingthe room walls need to be filtered through specially designedradiofrequency (RF) penetratian panels to preven extraneouselectrom:.gnetic radiation to enter the scanner room and PETacquisition electronics are enclosed in an RF tight cabinet.These and other modificalions are diseussed in more deail in[411. The authors show that despite the rnodificatiorsto the PET/M RI system components the performance of nei-ther the PET nor of the MR is degraded, and that both systemscan be operared in close spaial proxirniy.

Figure 2.9 illustrates a total-body imaging exarninationfrom the co-planar PETIlVIR system. While this design con-ce pt nay be regarded as a step closer towards integraredPETlMR (compared to sequential imagirg, Fig. 2.7a) itoffers sequential PET and MR! imaging with delays that areon the order of those in PET/CT and in sequential PET/CT-MR imaging [42]. It could be argued that co-registrationof PET and MR information is slightly better and. perhapsmore reproducible, in both modalities compared to the shut-tIe system in Fig. 2.7a. since patierts are not relocatedbetween rooms and repositioned using a mobile patiert han-dling system. However. no study to date has been able toverify this. Unlike with the separate PET/CT-MR system.one modality is idiing during co-planar PET/MR imaging,which may be argued to be less cost-effective. However, oneshould keep in mind that today few indications are c1earlydefined as key indications for PET/MR, and, therefore.throughput is likely not an issue for the time being. The co-planar PET/MR system offers full MR-flexibility and TOF-PET functionality. Unlike with the separate design. notransmission source is available, hus requiring MR-bascd

___.- - attenuation correction methods (see below).

.,

MRI FDG-PET-MRI

cr. complementary WB-MR and retrospectively aligned and fuscdPETUCf)-MR (Data courtcsy of Patdek veit-Haibach. MD, UniversityHospital Zurich)

Integrated PET/MR

The first PET/MR design for human use was presented asearly as in 2006, representing also the most challengingdesign concept (Fig. 2.7c) [43]. This PETlMR prototype sys-tem (BrainPET, Siemens Healthcare) was intended for brainimaging onlyand considered a proof-of-concept for a fullyirtegrated PET/MR. The BrainPET system was based on aPET detector ring designed as an insert to a 3T whole-bodyMR scanner (Magentom Trio, Siemens Healthcare Sector.Erlangen. Germary) with the novely being theMR-compatible PET detection system that was integratedinto the MR system. Here. the PMT were replaced byAvalanche photodiodes (APO), which have been shown tooperate in magnetic fields of up to 7 T [44J (see also seetion"MR-Compatible PET Detectors"). Therefore, in this designLSO (lutetium oxyorthosilicate)-based derector blocks, com-prising of a 12 x 12 matrix of2.5 x2.5 x20 mm' crystals weredirectly coupled to a compact 3 x 3 APO array. With this sys-tem PET and MRI cover an active co-axial FOY of 19.3 cmsirnultaneously, The point source sensitivity of the PET sys-tem measured with aline source in air w as 5.6 % and thespatial resolution was 2. I mm at the cerre of the FOY. Nodegradation of the MR images was observed due to the presence of the PET detectors and no detrimental effeet on theperformanee of the PET detectors was observed for a nurn-ber of standard MR pulse sequenees [45]. Since 2006 theBrainPET was installed at 4 sites worldw ide, with one siteoperating the PET insert inside a 9.4 T MR! as welL. Sornepreliminary elinical research data are deseribed in [46-481.Looked upon retrospectively, the elinical test phase of theBrainPET helped pave the road towards whole-body PET/MR, the advanced development of MR-based attenuation

16T. Beyer et a'

Fig.2.9 29-y/o Iernale paticnt with Maffucci syndrome diagnosed inher childhood, This disease is sporadic wilh multiple enchondromasand hemangiomas. An [18FI-FOG-PET/CT total body study was per-formed for staging. Subsequent total-body PET/NIR. using the same

correction and, perhaps most importantly, an improved COI11-municatior and closer collaboration of radiologists, nuclearmedicine physicians and physicists.

Based on the aforementioned positive BrainPET experi-ences a further step towards the integrared design corcept(Fig. 2.7c) was suggested in Iate 2010. Then, the first who le-body, integrated PETfMRI system (Biograph mMR, SiemensHealhcare) was proposed. Each PET detector block consistsof an 8 x 8 matrix of LSO crystals coupled to a 3 x 3 APD-array. The transaxial FOV of the MR is SO cm. whereas theaxial FOV is 45 cm. The PET subsystem consists of 8 ringsof 56 blocks with tm axial FOV of 25.8 cm and ring diameterof 65.6 cm. Both. the extended axial FOV of the PET and theredueed ring diameter help inerease the sensitivity of thePET insert. which in tum could be leveraged. for exarnple.for shorter emission sean times or reduced injected PETactivities. Thus, tle lack of TOF-capability in APD-bascdPET systerns (see Chap. 3) can be compensated for. in the-ory, by bringing the PET deectors eloser to the centre of theFOV and by extending the axial coverage. A detailed descrip-tion of the system together with a performance characeristic

FDG injection. rrore elearly presents bone involverrcnt and is preferredbecause of the need of multiple follow-up exarninations (Data courtesyof Osman Ranb. MD, University Hospital Geneva)

can be found in [491. On the downside of the eloser integra-tion the integrated PET/MR system, the bore drameter isreduced to 60 cm. which - for the moment - is the reversetrend of PET/CT and MR-only instrumentation with ganryand borc diameters of up to 80 and 70 cm. respectively,Inerensed ganry and bore diarneters help improve patiencomfort and compliance and, in addition. leave room forimage-guided interventions, if needed.

Perhaps most irnportantly, the integrated PET~IR designcorcept allows for simultaneous data acquisition. However,simultaneity of complementary volumetric data acquisitionis assured only for a selected MR sequence and emissiondata that are acquired for the duratior of that specific MRsequence. Nonetheless, simultaneous PETfMR is argued toimprove the diagnostic accuracy of combined PET/MR oversequential imaging (Fig. 2.7a, b). Figure 2.10 illustrates acase from the Biograph mMR system with very good spatialalignment of PET and MR images in the abdomen.

While the benefit from improved spauo-ternporal align-ment is imrnanent to the PETIMR images from integratedPETfMR it is not elem as to how rnuch it is required for elinical

2 PET/MR Instrumentation

Fig.2.10 61-y/o ferrale with known squamous cell carcinoma of thelung undergoing [18F}-FDG-PET/MR imaging on an integratedBiograph mMR PETINIR system. (a) lrcreased PET tracer activity syn-onymous of disseminated disease is depicted in the bronchia! carci-

rouine. Further, PET and MRdata are simultaneously acquiredonly for a limited period of time or for a selected region. orvoxel in Is extreme. Without a doubt, the closer alignment ofPET and MR data in both, an anatemical framework and overvarious imaging times will help in elinical researeh, such aswhen comparing perfusion studies with II SAl-water and ane-rial spin labeling (ASL) in MR [24J. Also, using integratedPETINlR imaging for shortening combined examinauon timesover those n sequertial and co-planar designs is preferred forthe well-being of patients with acute diseases, pediatric patientsrequiring sedation and patierts with neurodegenerative dis-eases. Finally, since MR-based rnotion detectior is conceiv-able during simultaneous PET/MR irnaging, such MR-derivedmotion vector can potentially be used to eorrect for motion-indueed blurring of the PET emission data [50, SI J.

As with the co-planar design. the integrated PET/MRdesign does not allow for separate transmission imaging and,therefore, PET-based attenuation data must be derived fromthe available MR information. Thus, anormal workftowstarts with the simulaneous acquisition of emission data anda dedicated MR sequence for the purpose of deriving attenu-ation data. As soon as the short MR-attenuation sequenee iscomplete, addirional diagnostic MR sequenees can be

17

'"

noma. frenral lobe metastasis. pancreas and in secondary. metastaticcolcrectal eaneer. Coronal whole-body Tl-weighted MR image (b).attenuation corrected PET (d), and PET/MR image (i) and correspond-ing axiul images through the bronehial carcinoma are shown (c. e. g)

acquired for the remainder of the pre-defined emission sean.Alternatively, the PET emission data can be acquired in list-mode format and reframed af ter finishing the MR sean,

Table 2.2 provides an overview of currently availablePETiMR sysems, All systems support the acquisition ofwhole-body, ifnot total-body, examinatiors. These first PET/MR design concepts vary more widely than the first conceptsfor PET/CT. This variation can be explained by the cornplexphysical and more dernanding echnical requirements for afull integration of PET and MR imaging systems, comparedto those from a PET/CT integration.

The foremos diffcrence betwcen the PETiMR systems isthe type of PET detector. Integrated PETiMR imagingrequires a novel PET-based detection system, which will beexplained in more detail in Chap. 3. APO-PET does not sup-port TOF-based acquisitiors due to the insufficient timingresolution of the APO, thus, only two PETfMR designs offerTOF-capabilities (Table 2.2). Major differences are alsa seenin the patient table design. which has subsequenteffects on thehandling of the patients and worktlow. Both, the co-planarand the integrared PETiMR require the use of the MR imagesfor human soft tissue attenuation correction, which today isperhaps the biggest challenge for combined PET/MR in the

18T. Beyer et aL.

Table 2.2 State-of-the-art PET/MR imaging systems by GE Healthcare. Philips Healthcare and Siemens Healthcare (from left to righr). Note. asof 2012 the Philips and Siemens system were FDA-approvcd and ccmmercially available. The figure shows key parameters and performunce

measures of the various PETlMR series

Discovery PET/CT-MR Ingenuily TF Biograph mMR

2 PET/MR Instrumentation 19

-. , .._,"';:ik. ..- .~\ PMT

Scintiltator

aB> OT

b Avalanche photadiedes

APD-based

PET detector

ot

MROiscovery MR 750w

70 cm bore diameter50 x 50 x 45 cm FOV

t 6 (32) receive channels0.5 ppm field homogeneily

PETDiscovery PET/CT 690

4.2 x 6.3 x 25 mm3 LYSO(Ce)81 cm detector ring diameter

Time-of-flightPET15.7 cm axial coverage

70 cm bore diameter

Patient handling system

Shuttle and docking system159 kg patienlload

170 cm co-sean range

CT-based AC

MRAchieva 3T TX

60 cm bore diameter50 x 50 x 45 cm FOV

16 (32) receive channels

0.5 ppm field homogeneityPET

MR3T (nol Verio-based)60 cm bore diameter50 x 50 x 45 cm FOV

18/32 reeeive ehannels

0.1 ppm field homogeneityPET

4.0 x 4.0 x 20 mm' LSO66 em ring diameter

No Time-of-flightoption25.8 em axial eoverage

60 em bore diameterPatient handling system

Inlegrated table platform200 kg patient load

140 cm co-sean range

MR-based AC

In-plan e resolution: 4.9 mm

Axial resolution: 5.6 mm30 Sensitivity: 7.0 eps/kBqPeak NECR: 130 kcpsCoinedenee 4.9 ns

TOF resolution: 549 ps

In-plane resolution: 4.9 mmAxial resolution: 4.9 mm

3D Sensitivity: 6.4 cpslkBq (NEMA)Peak NECR: >91kcps @ 16kBqiml

Comeldence : 3.8 nsTOF resolution. 535 ps

In-plane resolution: 4.5 mm

Axial resolution. 4.5 mm3D Sensitivily: 13.2 cpslkBq

Pea k NECR: 175 kcpsCoincidenee : 5.9 ns

NoTOFFiq, 2.11 (a) Convennonat PET detectcrs using photorrultiplier tubes(PM7) do not work inside a magnetic field. This is illustrated by thescintillatcr position profile that is skcwed already from fringe fieldsfrom a horseshoe magnet pluced next to the PET deteetorIPMR(Courtesy Bernd Pichler. University of Tbingen). New PET detectors

Compaet

High quantum effideney

Law bias voltage

Magnetic field nsensitiveLower gain

Limited time resolution

c SiPM or G-APD

SiPM-basedPET deteclor

4.0 x 4.0 x 22 mm' LYSO(Ce)90 cm deteetor ring diameter

Time-of-flightPET18 cm axial coverage

70 cm bore diameter

Patient handling system

Turning table platform200 kg patient load

190 cm co-sean range

MR-based AC

Highgain

Low bias voltage

Compact

Magnetic field insensitive

Very fas and TOF- compatible

Low cost CMOS process

Low quantum efflciency

Low photon detection efficieney

(b) based on avalanche photodicdes (APD) can be rnade more compactand have been shown to perform in rnagnetic fields up to 9.4T. Recemdevelopments indicare further improvcments for MR-compatiblc PETdetectors based on SiPM. a type of Geiger-APOs (c)

light of c1inieally adopted PET/CT imaging when absolutequantification of PET data is eonsidered.

MR-Compatible PET Detectors

Cross-talk effeets between PET and MRI may oceur wheninserting a eonventional PET derector and associated elec-troric components into an existing MR system. This rnayrelare to disruptions of the PET sigral cascade as well as todegraded MR maging. The possible interactions betweenPET and MR signal generation are manifold. A perfeet tech-nical integration of bat h modalities requires the MR with itselectronagnetic environment not to disturb the sensitive PETsignals. This encornpasses the strong static and honoge-neo us magnetic field for spin a!ignment (in the range of sev-

eral Tesla), the strong spatially (mT/m) and temporally(mT/m/ms) varying electrornagneic gradient fields for

spatial signnl localization, as well as the pulsed rudiofre-queney (RF) transmit and receive fields for spin exeirationand RF signal reception (MHz range). On the other hand,PET system components must not intertere with any of theabove listed eleetromagnetic fields of the MR system.

Consequently, for a fully-integrated PET/MR system. allPET eleetronics must be RF shielded in order not to disturbthe highly sensitive RF sigrals detected by the MR compo-nents. When shielding the PET components that are locatedclose to the MR gradient coils. the RF shielding has to bedesigned such that the strong time variaut gradient pulses donot produce unwanted Eddy currents in the shielding, whichrnay have a negaive effect on the gradient linearity, poten-tially Icading to image distortions [521

Given the design of standard PET detectors based on pho-tomultiplier tubes (PMTs). a PET/MR corfiguraion is obvi-ously technically more challenging than the combination ofPET and CT becausc phototubes are sensitiye even to low

magnetie fields (Fig. 2.11). Therefore, early developers ofPET/lvIR co nce pts. such as Hammer and co-workers, pro-posed to place only the PET scintillator inside the MR and touse light guides to channel the scintillation light from thedetector to the PMT situated outside the prirnary magnericfield of the MR system [53]. This idea was advanced furtherby other groups. as discussed by Wehrl in arecent review ofthe origins of PET/MR [24].

In order to provide PET perfornanee in PET/MR that issimilar to PET performance in PET/CT. any MR-compatiblePET detector mu st support accurate 30 positioning and veryfast timing information at no cost of volurne sersitivity. This.n turn, calls for combinations of scintillators with novel,

MR-compatible photodeteetors of high granularity such asAvalanche Phoodiodes (APO) and Silicon Photornultipliers(SiPM) [54].

Avalanche Photodiodes (APO]

Avalanche photodiodes (APO) are semiconduetor devicesthat transform deteeted light into an electrical signal follow-ing the same principles as ordinary photodiodes, However.nlike ordirary phoodiodes, APD's operate exclusively athigh electric fields. When an electron-hole pair is generatedby photon absorption, the electron (or the hole) accelerateand gain sufficient energy from the high electric field beforeit collides with the erystallattice and generates another elec-tran-hale pa ir while losing so me of its kineric energy in theprocess (Fig. 2.11b). This process is known as inpaet ioriza-tion. The original as well as the secondary electron (or hele)can the n accelerate again under the inftuenee of the highelectric field and create more electron hole pairs. Thisprocess ereares an avalanche of electron hole pairs - hence

201. Beyer et aL.

the name avalanche photodiodes. The rate at which electron-hale pairs are generated by impact ionization is balanced bythe rate at which they exit the high-field region and are eel-lected. If the magnitude of the electric field (reverse-biasvoltage) is below a value known as the breakdown voltage,the rate of collection exceeds that of electror hale erentionand causes the population of electrors and holes to deelineand eventually stops.

The number of created electron-hole pairs, referred to asinternal gain, is typically in the ran ge of 10'_10' and deper-dent on the electric field strength (reverse bias voltage).Because the average number of created electron-hole pair isstrictly proportional to the incident Iight photons, this modeof operation is known as linear mode.

Unlike amplification in PMTs. the internal gain of APDsis characterized by fluctuations due to the statistical nature ofimpact ionization. These gain flctuations produce excessnoise, which increases as the internal gain increases by rais-ing the reverse bias. Other factors that affect the performanceof APO include temperature. doping. as well as diode rnate-rial properties. In addition. APDs are characterized by a rela-tively long timing resolution (FWHM> 1,000 ps), whichlimits their use in TOF PET systems. Bccause of the se fac-ters, it is desirable to use APDs at moderate reverse bias volt-age and temperature to ensure their stable operation.

On the other hand, APDs are characrerized by high quantumefficiency (QE - number of elecrron or holes created per num-ber of incidem scintillation photons) particularly at the wave-lengths of PET scintillaion detectors. APDs are alsa immline(o after-pulsing, which are spurious pulses gererated fromelectron-holes being rapped by crysal defects and releasedatter a certain delay time, thus, confounding the detection pro-cess. Most impotantly and conrary to PMTs, APDs areirnmune to stationary and varying magnetic fields, thus, render-ing them suiable for PET/MR systems. APDs typically have amaximum size limited to about i cm'. due to the difficulty ofmanufccruring large area scmicorductor devices. however, theeost of rnanufacturing APDs is relatively law.

SiIicone Photomultipliers (SiPM)

A promising development in photodetectior for PET im:.g-ing is the introduction of Geiger mode avaianche photodiodes(G-APD. Fig. 2.llc), commonly referred to as silicon photo-multiplier (SiPM). This is a novel type of photodetector thatis about to reuch a performance level that offers significantimprovement over APD-based PET.

A SiPM is an APD operated with areverse bias voltageabove the breakdown voltage (-50-60 V above breakdownvoltage). In this case, the electron hale pairs generated byphoton absorption will multiply by impact ionization fasterthan they can be extracted, hus, resulring in an exponential

growth of electron-hole pairs and their associated photocur-rert. This process is known as Geiger discharge. The currentflow produced by the Geiger discharge is large and results ina large signal gain (more than 10'). Following a Geiger dis-charge, the SiPM is reset by dropping (quenching) the volt-age across the photodiode below the breakdown voltage.This will reduce the number of created electron hale pairsand eventually stop the Geiger discharge. The discharge-and-reset cycle is known as the Geiger mode of operation ofthe photodiode. The turn-on transient of the current dischargeis cornparatively fast, with several picoseconds while thetum-off transient through quenching is mostly dependent onthe SiPM size and is on the order of 100 ns. Quenching canbe achieved using active or passive techniques although forhigh counting capabilities, active quenching is preferred.

One important application of SiPMs is the ir abiliy tocount photons, which could be used to determine the energyof the incident annihilation photon on a scintillator in a PETsystem. However, a single SiPM eell has a linnitation in thatit is essentially either on or off. It eannot distinguish betweera single and multiple photons that arrive simultaneously. Oneonly knows that the APO was triggered. This limitation,however, is overcorne when us ing an array of SiPM cells thatare connected in parallel. In this case the output of the SiPMarray is the sum of the output of each SiPM cell (pixel) in thearray. For exarnple, when the photon ftux is lawand photonsarrive at a time interval longer than the recovery time of apixel, the array will output pulses that equate lo a single pho-toelectron. The gererated pulses are then converted to digialpulses and courted. However. when the photon flux is highor the photons arrive in short pulses (pulse width less thanthe recovery time of the SiPM), the pixel outputs will add upto the equivalen number of incident photons. In this ense,the SiPM array behaves in a pseudo-analog marner, becauseit can measure the incident number of photons per pulse,which is not possible with single photon counting SiPMs.

An imprtnnt feature of SiPMs is their immunity to excessnoise. This is primarily due to the fxed number of electron-hale pairs produeed in Geiger mode, which is not defined bythe statistics of the impact ionization process as in APDs.Anather irnportant feature of SiPMs is their relatively fastrise time, and short time jitter (FWHM = 0.1 ns) defined asthe statistical variation of time interval betweer the photonarrival and the resulting electrical sigral from the SiPM -thus. supporting their use in TOF-PET tomographs.Furtherrrore, the performance of SiPMs (like APDs) isimmune to the effects of stationary and temporally varyingmagnetic fields which allows their use in PETlMR systems.

On the other hard, SiPMs have a relatively law photondetection efficiency (POE). due to their lAWQE for scintilla-tion light from PET detectors (40 % at 420 nn). In addition.SiPMs are characterized by high dark count rae, high crosstalk and af ter pulsing as well as a strong ternperature and

2 PET/MR Instrumentation 21

PMT . -- APD ---- iPMActive area (mm 2) t-2,ooocm' 1-100 mm' I-LO mm'

Gain HJ'-IO' lO' 10'_10'

Rise time

22T. Beyer et a'

~.?"If;;~ '":~......@ ,~-@.~ ~~ ~ '"X), re; "'. i XPET emission (30 min) PET transmission (15 min) Attenuation-corrected

PET emission imagea

AUenuation correction tactar

i; i e -oL ~(X, E PET)dxo i .

! Attenuaton

i True PET Signal

Tissue attenuatloruevdensity) in (HUIb

Aif -1000

Lungs -850 ... -250

Water OSaft tissue 20 ... 300

Bone 300 .. 2000 CT transmission (1 min)

,?.-i

Proton density in tissueT1 ,T2,T2w relaxation parameters cMR,AC-sequence (2 min)

Fig. 2. 2 Cballenges of MR-based auenuation correction. (a) in PETattenuancn correction facors can be calculated from separate PETtransmission measurements. which take a relatively long time but pro-vide attenuation values at s keV (b) in PET/cr Cl-based attenuationvalues. representing a measure of the electron-density. can be used to

estimate PET attenuaticn coefficients. (c) In PETlMR no measure ofeleetten density is available and tissue appearance on MR and crimages is markedly different for air and bone. Therefore. no direct rnea-surement is available for MR-based attenuation coefficients

for reconstruction of fat-only. water-only and fat-waterimages, and results in tissue segrnentation of air, fat, muscle,and lungs [6IJ. Bone is not accounted for in this approach.Initial results in elinical pilot studies have shown that thisapproach works reliably and provides results that are cornpa-rable to corrected images form PET/CT in the same irdivid- _ual, However, further studies are needed to assess the impactof ignoring bone and the overall aceuraey of MR-based ACmethods on PET quantification.

This relates to the patient table, transmit and receive radio-frequency (RF) coils as well as positioning aids, The fact thatthe RF eoils are located inside the FOV of the PET system(Fig. 2.7b, c) is a ch alien ge and has only started to beaddressed. For brain seans, the head coil is rigid and its atter-uatior values can be estinated from a refererce CT-bascdauenuation map, Subsequently for any PET/MR studyonlythe relative position of the head coil inside the PETIlvIR sys-tem would be required, Additicnal work has been directedtowards reducing the amount of attenuating materials in N1Rcoils used in PET/CT as exernplified in a rnodified brain coilfor integrared PETIlvIR irnaging [62, 63].

For extru-cranial exarninations the situation is moredemanding. MR surface coils are required to achieve optimalsignal-to-noise-raio (SNR) and high quality MR irnages,Surface RF coils may contain elastie components and hencetheir individual position on the patient cannot easily be

The Effect of MR Radiofrequency (RF)Coils on MR-AC

In addition to the general transformatian of suitable rvlRimage information of paien tissues. other. hardware-relatedattenuators mu st be considered during the transformution.

2 PET/MR lnstrumentation 23

Fig.2.13 Ccntributors toattenuation and image distortionsin PET/MR. Topics marked asgree are resclved and addressed.those marked in yeltow areknown but solutions are work-ir-progress Patient

Tssue

-Motion

predicted. The effect of flexible body coils on overall PETattenuation was recently estimated by Tellmann and col-leagues [64]. The authors report a maximum bias of 4 % inattenuation-corrected torso PET if surface coils are notaccounted for during AC. This bi as is negligible compared tothe respective bias in head studies when ignoring dedicated,rigid head RF coils (up to 20 %). MacOonald and colleaguesreport sirnilar results [65].

All PETlMR vendors today offer CT-based attenuationtemplates for rigid coils as well as for the patient bed that areseamlessly integrared during the attenuation correction.Nonetheless, elinical studies are required to further study theeffect of misaligred RF coil ternplates and missing ternplatesfor aecurate represertation of flexible coils on PETquantification.

The Presence of MR Contrast Agents

MR-based AC could potentially be biased from the presenceof MR centrast materials. which are ypically made up ofiron oxide and Gd-chelates for oral and intravenous (IV)applicaion, respectively, It is known from the developmentof CT-based AC that the presenee of contrast materials withatomic nurnbers higher than those of water may lead tobiased attenuation maps for PET crnissior data. The sameeffects may oecur with MRCA that are applied during PET/MR irnaging, Furthermore, the presence of MR contrastagents may produce changes in the vR signal intensity thatyield biased attenuation maps. First studies indicate no nega-tive effect from MR contrast on PET quantification follow-ing MR-based atteruatior correction [66, 67J.

/' Transverse FOV

MR non-uniformities

-t

J *~i iiiiiii\: 7 / iiiiiii~~

Understccd and WIP

Limited FOV and Truncation Effects

Given the reduced bore diameter and the relatively longexarninaion times in PETlMR cornpared to elinical PETICT, most patierts are positioned in the more cornfortablepositior with their arms dowr. Thus, the patient anatomymay well extend beyond the transverse FOV of the MR(typically 50 cm). whereby the arms or the trunk of thepatient are not fully eovered by the MR irnages used forMR-AC. This may yield an underestimation of the recon-structed, attenuation-corrected ernissior activity eoncentra-tion. Truncation artifacts were deseribed for PET/CTimaging [68J and have been reviewed for PET/MR [69]. ltwas shown that with the arrns extending beyand the FOV ofthe MR the PET activity following MR-AC was biased byup to 14 % in the area of truneation. The underestirnatedactivity eoncentration could be recovercd to within 2 % ofthe norniral concentratior following sirnple, manual exten-sion of the atteruatior map.

An alternative solution would be to use the uncorreetedPET image to estirnate the patient cross-seetion in thoseareas outside the measured FOV where no - or only geo-metrically distorted - MR information is available [70, 71].The elinical feasibility of this approach still needs to be vali-dated. Particularly, in irnagirg scenarios with highly specifictracers the arms may be difficult to segment auornatically inthe uneorrected PET irnages.

Figure 2. 13 surnmarizes the challenges and the status ofMR-based attenuation correctiors in PET/MR. Most chal-lenges are understood with so me being addressed sufficientlyand some awaiting further optimization, val idation and elini-cal adeption.

24T. Beyer et aL.

MR-Based Partial Volume Correction corrected

As early as 1991, Leahy et aL.suggested that PET reconstruc-tion could be improved by using anatomical MR irnages fromthe same patient as prior information [721 lt is common elini-cal practice today for neurology patients with a PET-indicationto alsa undergo an MR examination. However, MR-guidedPET reconstruction has not yet made the transition fromre search into elinical routine. Aside from logistical problemsassociated with the retrieval of the complementary image sets,sub-optimal rerospective image aligrment would significartlydeteriorate the qualiy of the PET data [731. However, in com-bined PETlMR imaging systems, the sparial (and ternporal)alignment accuracy could be improved, thus, helping to pro-mote the concept of MR-guided PET image recorstruction.

Even if the PET image is reconsructed independertly ofthe MR image, it is stili possible to use the MR image of thepatient as an aid for improved quantifieation. In particnlarMR-guided partial volume correction (PVC) was suggestedas early as in 1990 [74, 75]. Again PET and MR images fromcombined PETlMR examinations may facilitate irnprove-ments in MR-based PET quantification through the use ofMR-based PVc.

Uncorrected(18F)-FDG PET

.,.alP'

MR-Based Motion Correction

Fig.2.14 FDG-PET irragesfollowing auenuation correction (left) andmotion + atenuation correction (riglt) clearly demonsrae the poten-tially improved quality of the data from lengthy examinations (Courtesyof J Scheins and H Herzog, Research Centre Jlich)

Patient motior. from involnt::ry movements, and cardiac aswell as respiratory cycles, is a major contributior to degradedPET image quality. In addition. patient motior will lead tolocal or extended mis-alignment of complementary anato-metabolic image information. In PET/CT, for example, thePET image is acquired over several minutes, while the CTsean is amalter of seconds and frequenly acquired during asingle breath hold. As a result, patiert moion typicallycauses local rnisaligrment between the PET and CT imagesand may lead to seriolis artifacts for AC, for example nearthe diaphragrn. Dedicated breahing instructions have beenshown to help reduce misalignment in the horax and upperabdemen [76. 77]. Other authors have reeommended 4DPET/CT acquisition and AC, however, this involves a sub-stantially higher patient radiation dose [78-80J.

Respiration is expected to generate misalignment andblurring in PETlMR irnages, too. As MR seans generallytake nch longer than CT seans, patients spend an even lon-ger time in the PETlMR compared to PET/CT, and can se-quently patient motion is likely to cause eve n more severeartifacts. However, integrated PET/MR system technologyoffers a promising solution LO the problem (Fig. 2.14).Various NIRl motior-tracking techniqes are available inelinical settings, including but not limited to cloverleaf navi-gators [81J. Such echniques have been tested with theBrainPET system with promising results from estimating

and correctirg involuntary head motior as a result of relax-atior of neck muscles. Usiug 3D Hoffman brain phantomand human volunteer studies. Catana et ol. reported thathigh- temporal-resolution MRI-deived ma tion estimatesacquired simultaneously on the hybrid BroinPET system(Siemens Healthcare) can be used to improve PET imagequality, thus increasing its reliability, reproducibility. andquantitative accuracy [SOL.

Likewisc. novel 30 cine sequences are under develop-ment to track spatio-temporal deformation of organs such asthe heart and the thorax, Subsequently, deforrnation fieldsare generacd and incorporated ino the PET reconstnction

[51.82-85].Thus, the use of periodic MR navigator signaIs in co n-

junction with a 4D model of the human torso may help tocorrect for motion-induced image degeneration in PETlMRdata following 4D-MR-AC. which would be o major advar-age over CT-AC.

PET/MR Safety

Combined PET/CT has been clinically very successful andmay well serve as a benchmark for the development of PET/MR. However, despite the success of PET/CT the re are alsa

2 PET/MR lnstrumentation 25

shortcomings in the use of CT as the anatornical corrple-mert to PET. As such, CT uses a souree of iorizing radia-tion for imaging and. therefore, adds significart radiatiordose to the overall exarnination. Brix et al, have shown thatthe diagnostic CT contributes up to 75 % of the effectivedose in patiens undergoing whole-body FDG-PET/CTexarninations for oneology indications leading to a total ofabout 25 mSv effective dose [86]. These dose levels mayraise concem in selected population like adolescens andfemales. Figure 2. 15a illusrates the relative cortribution topatient exposure from the individual steps in a combinedPET/CT examinatioi.

In PETlMR examinations, overall patient exposure isreduced significantly by replacing the CT imaging step withan MR imaging sequence (Fig, 2.1Sb). In addition, MR pro-vides advaneed functioral imaging information, such asDWI or MRS, wihout adding to the overall radiatior expo-sure burden. Nonetheless, staff exposure is expected toinerense slightly in PETlMR, given the complexity of thepatient set-up when employing a range of surfaee RF bodycoils. However, no valid data are available as of yel.

Long-tem experience and hundreds of millions of rou-tinely and safely performed MR exarninations confinn thatMRI is a safe imaging modaliy. Nevertheless, a number ofsafety concems do apply to PETlMR as discussed by Brixet al. [871, of which all are all associated with the generalsafety issues known from MR-only imaging. The strong staticmagnetic field associated with MR systems potentially canattract ferrornagneic equipment as well as same patientirnplants. and accelerate these towards the strongest magneticfield in the isocertre of the PETlMR system. In same patientsthe strong and fast switching gradient fields may lead toperipheral nerve stimulation that are harrnless but neverthelessdisturbing. Finally, the strong-pulsed RF fields for MR signalexeiration can cause tissue henring. As with all other RF trans-mitting devices, the RF power n MR imaging is limited loharmless values of the specific absorption rate (SAR) not lead-ing to critical tissue heating. Some electric condeting metalimplants, however, potentially may increase the local SARvalues during an MR examination above the allowed SAR!imits. To reduce all associated poential risks of MR imaging,patient questionnaires and patient screening and selection pro-cedures have to be established and used in daily routine.

Accordingly, MR and PETlMR examinations of patienswith passive implans (e.g., vascular elips and clarnps, intra-vascular stents and filters, vascular access ports and catheters,heart valve prestheses. orthopedic prostheses, sheets andscrews, intrauterine contraceptive devices), active irnplants(e.g., cardiae pace-makers and defibrillators, cochlearirnplants, electronic drug infusion pumps) or other objectsof ferromagnetic or unknown material (pellets, bullets) arealways associated with a potential risk. Careful pre-exarni-nation interviews of the patients regarding the presence or

absence of passive implants, which may interfere with theMR imaging protocol, or deter the patient from this examina-tion all together is mandaory [87J.

Summary and Conclusion

Multi-modality imaging insrumentation has evolved dra-rnatically during the past decade. Combined SPECT/CT,PET/CT and, lately, PETlMR have revolutionized imagingand medical diagnosis. In these times of limited resources inhealthcare and rapidly increasing radiotion awareness, anypredietions for fuure developments of PETlMR technologymust take into account a variety of aspects, ranging fromcost-effectiveness to overall radiation dose. While techno-logical innovation, such as PETlMR. always pairs withenthusiasm and public interest. subsequen commercial sys-tems must be affordable and srategies for the ir elinicalimplementation must be assessed for their health benefit tojustify their pursuit within alocal or global healthcare sys-tem [88]. The impressive advances in imaging technology ofthe past decade came at a cost. but at what point do theseadvarces become cost-effective? Whole-body PET exami-nations that took i h at the start of the last decade now take5 min on PET/CT; the actual imaging takes only o fraction ofthe time needed for patient preparation and positioning orreporting the study. Does the increased wealth of availableinfannation from the MR rnake up for the increased exami-nation time?

The radiation dose to the patiert incurred by PET/CT iselearlyan issue. Although the ALARA (as low os reasonablyachievable) principle is sond advice, there are elearly groupsof cancer-sufferers such as those in children and young adultswhere the probability of inducing a second. radiaion-associared cancer exceeds the benefits that can be accruedfrom the study. Different imaging stratcgies should then beadopted, such as MR!, optical imagirg or ultrasound.

The cornrnendable drive to reduce radiation exposure topatierts has fostered an interest in a combination of PETwith MR!. However. it is fair to assume that as long as dis-eases such as cancer and dementia rernain primarily diseasesof the elderly, the benefis of nuclear and X-ray imaging willlargely outweigh the risks.

Will the coming decade witness the replacement of PET/CT by PETIMRI? Same believe it wilL.just as in the 1980sthe re wcre those who predicted that MRI wold replace CTwithin 5 years. Of course that never happered, as both tech-niques have strengths and weaknesses and they have eachfound their niche in the medical imaging armamentarium.The same is likely true of PET/CT and PETIMRI-thetechnical challenges will be solved and simultaneos acqui-sition of MRI and PET will undoubtedly open new doors inelinical research and eventually alsa in the clinic.

,1J

26

a ..

1.

1. Patient prepartionJpositioning

2. Topogram

3. Low-dose CT

~ cr- ~OO attenuation ccrrection4. Multi-step PET

5. Dedicated CT (contrast, 9ating, breath- hold)

6. Escorting the patient out

b

1. Patient preparationlpositioning

2. Scout (MRI)

3. + 4. Multi-step PET/MR include. MR-AC (3)

5. PET and MR image reconstruction,Escorting the patient out

Fig.2.15 Relative conrnbutions to patient and staff exposure during nwhole-body oneology examination in PET/cr Ca) and PET&.>IR (b).The amount of radioactivity injected into the patient for a PET/cr (a.step i) and PETI?v1R (n. step i) is assuned to be identica!. Note. patient

Acknowlcdgcmcnt Wc are indebted to Caspar Delso (GEHe), HansHerzog (Research Centre Jlich). Jens-Christoph Georgi (SiemensHealthcare). Antenis Kalemis (Philips Healthcare). Bernd Pichler(University of Tbingen). Nina Schwenzer (University of Tbingen).

T. Beyer et a'

-ii

i- irel.exposure patient Staff

--i

rel. exposure Patient Staff

set-up in PETlMR is more elaborate and. therefore, relative andpotentially total sraff cxposures are expected to be higher than those inPET/CT

Jrgen Scbeins (Research Centre Jlich). Hclger Schnidt (Universityof Tbingen). David W Townsend (Singapore). Rtiner Veigel (PhilipsHealthcare). Patrick Veit-Haibach (Zurich) for helpful discussions andthe provision of support materials.

2 PET/MR Instrumentation

References

27

i. Gambhir SS et al (2001) A tabulated summary of the FDG PETlitcrature. J Nucl Med 42(5 supp!): iS-935

2. Herholz K. Heiss W-D (2004) Positron emission tomography inelinical neurology. Mol Imaging BioI6(4):239-269

3. Knuuti J (2004) Clinical cardiac PET in the future. Eur J Nucl Med31(4):467-468

4. Wahl RL et al (1993) "Anatcmetabolic" turnor imaging: fusion ofFDG PET with CT or MRI to localize foci of inerensed activity.J Nucl Med 34:1190-1197

5. Wagner HN (1998) A brief history of positrcn emission tomogra-phy. Sernin Nucl Med XXVLLL(3):213-220

6. Townsend D (20081 Multimodality imaging of structure and func-tion. Phys Med BioI53(4):RI-R39

7. von Sehulthess GK (2O

28T. Beyer et al.

52. Tnhn D. Kiessling F, Schulz V (2011) Optinized RF sbieldingtechniques for simulaneous PET/MR. Med Phys 38(7):3995-4000

53. Hammer B (1990) NMR-PET scanner apparatus. U.S. Patent, USA.3July 1990

54. Lecomte R (2009) Novel detectcr technology for elinical PET. EurJ Nucl Med Mollmaging 36(suppl 1):S69-585

55. Degenhardt C et al (2009) The digital silicon photomultiplier - anovel sensor for the deteetion of scirulllaion light. In: IEEE nuelearscicnce symposium corference (NSSIMIC). Orlando. 24 Oct- Nov 2009

56. Keereman V et al (2013) Chal1enges and current methods for atten-uaticn correction in PET/MR. MAGMA 26(1):81-98

57. Zaidi H. Hasegawa B (2003) Deterrnination of the attenuation mapin emission tomography. J Nucl Med 44(2):291-315

58. Beyer T et al (2008) MR-bascd attenuation correetion for torso-PETlrvlR imaging: pitfalls in mapping MR to cr data. Eur J NuclMed Mollmaging 35(6):1142-1146

59. Hofmanr M et al (2009) Towards quantitative PETIMRI: a reviewof MR-based atteruaticn correction techniques. Eur J Nucl MedMol lmaging 36(suppl ~):593-5103

60. Hofmann M et al (2011) MRI-based attenuation correction forwhole-body PET/MRI: quantitative evaluation of segmentation-and atlas-besed methods. J Nucl Med 52(9):1392-1399

61. Martinez-Mller A et al (2009) Tissue classification as a potentialapproach for auenuation correction in whole-body PET/MRI: eval-uation with PET/CT data. J Nucl Med 50(4):520-526

62. Delso G et al (2009) Study of MR head and neck coil for its use inan irtcgrated MR/PET scanner. J Nucl Med 50(suppl 2):294-295P

63. Delso G et al (2010) Evaluntionof the attenuation properties of MRequipment for its use in a whcle-body PETlMR scanner. Phys MedBioI55(15):4361-4374

64. Tellmann L et al (20 ll) The effect of MR coils on PET quantificationin whole-body PET/MR: results from u pseudo-PET/MR phartcmstudy. Med Phys 38(5):2795-2805

65. MacDonald L ct al (2011) Effetcs of MR surface coils on PETquantifcation. Med Phys 38(6):2948-2953

66. Lois C et al (2011) Effect of MR centrast agerus on quantitativeaccuracy of PET in eombined whole-body PET/MR imaging. Eur JNucl Med Mol Imaging 38(suppl 2):S 156

67. Lee W et al (2011) Effects of MR centrast agerts on PET quantita-tion in PET/MR! study. J Nucl Med 52(suppl 1):53

68. Beyer T e al (2006) Whole-body 18F-FDG PET/CT in he pres-ence of truncation anifacts. J Nucl Med 47(1):91-99

69. Delso G et al (2010) The effect of limited MR field of view in MRJPET attenation correction. Med Phys 37(6):2804-2812

70. Nuyts J et al (1999) Sirnultareous maximum a posterieri recon-struction of atteruation and activity distributions from emissionsinograms. IEEE Trans Med Imaging 18(5):393-403

7 i. Salomon A et al (2009) Iterative generation of attenuation maps in TOF-PETlMR using consistency ccrditionsJ Nucl Med 50(suppl 2):425P

72. Leahy R. Yan X (1991) Incorporation of anatemical MR data forimproved furctional imaging with PET. In: Xllth IPMI irtema-ticnal conference. Wye. 1991

73. Lipinski B et al (1997) Expectaticn maxirnization reconsrructior ofpositron emission tomography irnages using anatomical magneticresonanee information. IEEE Trans Med Irraging 16(2): 129-136

74. Meltzer C et al (1990) Correcion of PET data for partial volumeeffeets n human cerebral cortex by MR imaging. J Comput AssistTomogr 14(4):561

75. Meltzer CC et al (1999) Comparative evaluation of MR-basedpartial-volume correction sehemes for PET. J Nucl Med40( 12):2053-2065

76. Beyer T et al (2003) A limited breath-hcld technique for imprcvedimage quality in multi-siice PET/CT exarrs. 1 Nucl Med44(5):274P-275P

77. Beyer T et ai (2003) Dual-rnodality PET/cr imaging: the effec ofrespiratory molion on combined image quaiity in clinicaloncology.Eur J Nucl Med 30(4):588-596

78. Nehmeh 5 e al (2007) Deep-inspiration breath-hold PET/CT of thethcrax. J Nucl Med 48(1):22-26

79. Mori S et al (2009) Effective doses in four-dimensional conputedtomography for lung radiotherapy planning. Med Dosim34(1):87-90

80. Chang G et al (20 o) Implementaricn of an automated respiratoryamplitude gating technique for PET/CT: elinical evaluation. J Nuc1Med 51(1):16-24

81. van der Kouwe A. Benner T. Dale A (2006) Real-time rigid bodyrnorion correetion and shimming using cloverleaf navigatcrs. MagnReson Med 56(5):1019-1032

82. Kcllman P et al (2008) Fully automatic. retrospective enhancementof real-tirne acquired cardiac cine MR irnages using image-besednavigators and respiratory rnotior-corrected averaging. MagnReson Med 59(4):771-778

83. Tsoumpas C et al (2010) Simultaneous PET/MR acquisition andMR-derivcd notion felds for eorreetion of non-rigid motion inPET. Ann Nucl Med 24(10):745-750

84. King A et al (2012) Thoracic respiratory motion estimation fromMRl using a st~tistieal model and a 2-0 image navigator. MedImaging Anal 16(1):252-264

S5. Buerger C et al (2012) Nonrigid rrotion modeling of the liver from3-D undersampled self-gated golden-radial phase encoded MRI.IEEE Trans Med lmaging 31(3):805-815

S6. Brix G et al (2005) Radiauon exposure of patients undcrgoingwhole-body dual-modality FDG-PET/CT examinations. J NuclMed 46(4):608-613

87. Brix G ct al (2009) Risks and safety aspects related to PET&-'lRexarninauons. Eur J Nucl Med Mol lmaging 36(suppl i):5131-S138

S8. Goyen M. Debatin J (2009) Healthcare costs for new technologies.Em J Nucl Med Mol Imaging 36(suppl 1):5139-5143

Technical Principles and Protocolsof PET/MR Imaging 3A. Kalemis ~

Contents IntroductionIntroduction .. 29

30

30

31

31

PET and MRl are two well-established medical imagingmodalities that are used frequently os diagnostic tools in awide range of elinical irdications providing complementaryinfannation [Il. MRI can provide anatomical informationwith very high spatial resolution. and functional measurernentsat organ and tissue level with high diagnosric sensitivity. Onthe other hand, PET images functional processes at cellularand sub-cellular level with very high diagnostic specificityand high tracer detection sensitivity (I 0-11-~ 0-12 mol/l) [2] butwith a sparial resolution inferior to MR!. This complementarymarching of capabilities renders both PET(lCT) and MR] nec-essary in sevemi disease pathways, particularly in oneologyand neurology. A significant workflow lirnitation, when bothmodalities are needed, is the physical and organisationa! sepa-raion of the two systems [3]. Patients who need both PET andMR imaging are of ten referred for such seans independentlyand. often, theyare scanned with a significart time difference.This renders the fusion of information from both exarninationsdifficult or even impossible due to disease status changes orseveralother technical and organisational factors 14]. Hence,when MRl is the preferred imaging rnodality versus CT, PET/MR should be more clinically useful than PET/CT and a com-bined single examination could provide significant benefits toboth the patient and the hospital.

Today, there are two different designs for combined PET/MR sysems: positioning PET inside the MR! magne or intandem, similar to PET/CT [5J. Both philosophies attemptto balance elinical utility, user flexibility in developingclinically-relevant PETlMR-dedicated imaging protocols,potertial sacrifkes in relation to stand-alone state-of-the-artPET and MRI imaging capabilities and co st. lrrespectively,of the design and techrical differences with the stand-al onesystems, PETlWIR, as a novel imaging option, requiressignificant innovation at various levels n order to surpasscurrent concems and scepticism. The latter are of elini-caL. organisatioral and techrological nature. Questiorsabout how it compares with PET/CT, under which elinical

Opera tion al Requirements ..

PET/MR Applicatiors ...

Clinical ProtoeoIs and Workflow Considerations ..

Single-Organ Imaging

Whole-Body lmaging .. 32

Tcchnical Requirements

Sean

34

34

35Image Quality and Quantifieation ..

37

A. KalemisPETIMR,.Philips Healthcare.Guildford Business Park. Guildford, Surrey GU2 8XH. UKe-mail: anonis.kalemis@philips.com

._~__ O. Ranb et aL. (cds.). Ar/as o/PET/MR Imagig in Oneology.....:::: 001 10.1007/978-3-642-31292-2_3, Springer- Verlag Berlin Heidelberg 2013

29

30A. Kalemis

seenarios either of them should be used, patient throughput.ease of worktlow, building and running costs, ownership ofdevice and staff (betweer different departrnents). reliabilityof new the technologyand image quality in comparison tostand-alane systems are com mo n discussion points.

Operational Requirements

l is common in healthcare for radcal new technologies torequire a second innovation wave, in the area of its structureand organisation. in order to optirnise their contribution indisease managernent and patient pathway [6J. An organisa-tion wishing to adopt PETlMR, in particular, has to over-corne its complicated logistics, of ten the single-modalitytrained technical personnel, the saffing requirement fromtwo different departments and the excessive scanning timerequired to acquire both PET and MR seans, which raisessignificantly the cost of each sean. At this level, institutionsare requested to innovate in order to successfully adopt thisnovel technology.

At infrastructure level, so me degree of cross-departmentalcollaboration and in sorne cases even restructuring is neces-sary to bring the Nuclear Medicine (NM) and Radiologystaff much c1oser. Dual-raining of the technieal personnel isessential in order to operate the scanner while arother long-terrn consideration is the need to eress-trnin Radiologists andNuclear Medicine physicians from both modalities [7J.However, such needs are not sttuightforward due several fac-ters. amongst ohers the exising territorial and protectivepractices in many healthcare faciluies and the variatiors inthe legal frameworks for imaging technologists [8L In manyhospitals Radiology and NM departments are far from eachother creating further complications in staff allocation and/orlogisucs of the new scanners, Therefore, a suecessful modelneeds to be devised for the placement of the scanner corsid-er