review article pet-specific parameters and radiotracers in...

Review ArticlePET-Specific Parameters and Radiotracers inTheoretical Tumour Modelling

Matthew Jennings12 Loredana G Marcu13 and Eva Bezak12

1School of Chemistry amp Physics University of Adelaide Adelaide SA 5000 Australia2Department of Medical Physics Royal Adelaide Hospital Adelaide SA 5000 Australia3Faculty of Science University of Oradea 410087 Oradea Romania

Correspondence should be addressed to Matthew Jennings matthewjenningshealthsagovau

Received 17 July 2014 Accepted 15 September 2014

Academic Editor Iuliana Toma-Dasu

Copyright copy 2015 Matthew Jennings et al This is an open access article distributed under the Creative Commons AttributionLicense which permits unrestricted use distribution and reproduction in any medium provided the original work is properlycited

The innovation of computational techniques serves as an important step toward optimized patient-specific management of cancerIn particular in silico simulation of tumour growth and treatment response may eventually yield accurate information on diseaseprogression enhance the quality of cancer treatment and explain why certain therapies are effective where others are not Insilicomodelling is demonstrated to considerably benefit from information obtainable with PET and PETCT In particular modelshave successfully integrated tumour glucose metabolism cell proliferation and cell oxygenation from multiple tracers in order tosimulate tumour behaviour With the development of novel radiotracers to image additional tumour phenomena such as pH andgene expression the value of PET and PETCT data for use in tumourmodels will continue to grow In this work the use of PET andPETCT information in in silico tumour models is reviewed The various parameters that can be obtained using PET and PETCTare detailed as well as the radiotracers that may be used for this purpose their utility and limitations The biophysical measuresused to quantify PET and PETCT data are also described Finally a list of in silico models that incorporate PET andor PETCTdata is provided and reviewed

1 Introduction

Anatomic imaging modalities particularly X-ray computedtomography (CT) and magnetic resonance imaging (MRI)have long been the standard tools for the accurate localizationof organs and lesions in radiation oncology Today theyplay a routine role in three-dimensional treatment planningHowever the effectiveness of structural imaging techniquesin determining metabolic or functional tissue informationis limited Functional imaging has been demonstrated tobe invaluable for the initial diagnosis and staging of canceras well as the monitoring of therapy and the detection ofcancer recurrence [1] Metabolic changes in tissue com-monly precede the structural changes that are detected viaCT and MRI Thus imaging of metabolic changes mayenable the detection of malignant disease at earlier stages ofdevelopment [2] In addition the availability of functionalinformation is advantageous for cases in which there is poor

contrast between normal and malignant tissue when usingstructural imaging For example for the initial staging oflymphomas the metabolic information provided by positronemission tomography (PET) enables more accurate delin-eation of the extent of nodal disease as compared with CTand bone scans Similarly and perhaps most notably PETdemonstrates superior local staging capabilities for head andneck cancers over both CT and MRI FDG-PET alone has aplethora of indications for a wide variety of malignancies [3]Consequently functional imagingmodalities such as PET areplaying an increasingly important role in the management ofmalignant disease

PET scanning has progressed into widespread clini-cal practice since its first commercialization in the late1970s For oncological PET studies the most utilized andextensively researched radiotracer is 18F-fluorodeoxyglucose(FDG) FDG-PET has demonstrated superior accuracy overconventional imagingmodalities in multiple scenarios across

Hindawi Publishing CorporationComputational and Mathematical Methods in MedicineVolume 2015 Article ID 415923 11 pageshttpdxdoiorg1011552015415923

2 Computational and Mathematical Methods in Medicine

both the diagnosis and the staging of cancer [3 4] In a varietyof clinical settings FDG-PET exhibits improved values ofsensitivity specificity or both Examples of clinical scenariosin which FDG-PET has demonstrated efficacy include theevaluation ofmass lesions the staging and restaging of cancerthe planning of radiotherapy treatments the monitoring oftherapy and the detection of cancer recurrence [3] Howeverthe primary drawback of functional imaging is the lack ofanatomic information that it provides This necessitates itsaccompaniment with structural imaging for application inclinical oncology [1]

In order to take advantage of their inherent benefits thecombined use of both anatomic and physiologic imagingmodalities is optimal The accurate structure localizationcapabilities of CT complement the mapping of normal andabnormal tissue function performed by PET Because of thisPET images are routinely read alongside CT images in orderto both distinguish and localize metabolic irregularities Inthe first instance this has been achieved via the coregistrationof separately acquired PET and CT images using fusionsoftware a technique which has especially proved to beeffective for brain imaging However coregistration of imagesof other anatomical regions poses significant challenges [5]Difficulties in the registration process primarily arise fromvarying patient positioning between the two image setsWhilst such difficulties are minimal for brain scans theymay be significant in other anatomical regions wherein theremay be substantial organ movement or deformation In lightof this the development and subsequent commercializationof the PETCT scanner in 2001 has generally addressed theissues affecting the coregistration of separately acquired PETand CT images [6]

The use of integrated PETCT in oncologic imaging hassince become a widespread field of research particularlywith the utilization of 18F-FDG This combined modalityovercomes some of the drawbacks that are characteristic ofstandalone PET scanning namely the significant presence ofnoise in attenuation correction factors the lengthy durationof scans and the absence of anatomic markers [7] While thevast majority of published research concerns standalone PETPETCT has begun to show great promise across a multitudeof clinical settings [3] Studies have shown significantlyimproved accuracy in the staging of nonsmall cell lung cancerwith PETCT over the separate performance of PET and CT[8ndash10] Some centres perform a series of PETCT scans onnonoperative head and neck cancer patients following eitherradiation therapy or chemotherapy due to its indispensablecombined anatomical and functional information [3 7 11]For patients with recurrent or metastasized thyroid cancerthe localization capabilities of 18F-FDG-PETCT can lead toimproved diagnostic accuracy [12] Finally the incorporationof FDG-PETCT data into the radiation treatment planningprocess has repeatedly shown to improve target delineationand enhance the therapeutic ratio via its increased cancerstaging accuracy [13] The ongoing clinical evaluation andfurther innovation of PETCT namely via technologicalimprovements and the establishment of new tracers willcontinue to propel this technology into widespread use inoncology [3]

The significant interpatient variability in both tumourbehaviour and normal tissue response to cancer therapyhas prompted increasing demand for individualized treat-ment planning Simulations of radiobiological processes inmalignancies provide scope for the further optimization ofindividual treatment plans and to ultimately improve patientoutcomes The development of computer or in silico modelsof these radiobiological processes is particularly beneficial[14] It is here in which PET and PETCT as noninvasivefunctional imaging modalities can play a crucial role [15ndash17] The aim of this work is to review the application andeffectiveness of PET and PETCT for the in silico modellingof tumours In particular it examines the various parametersthat can be determined via PET and PETCT imaging alongwith the assortment of radiotracers available for these pur-poses The biophysical quantities used to quantify PET andPETCTdata including SUVand compartmentalmodels areassessed Finally an overview of the existing in silico modelsthat utilize PET or PETCT data is provided

2 Tumour Model Parameters Obtainable fromPET Imaging

With the objective of optimal individualized treatment plan-ningmany biological characteristics of a given tumour can beconsidered for computer simulation In the case of radiationtherapy the proliferative potential of a tumour and itsradiosensitivity are of particular interest Parameters whichdescribe tumour cell proliferation and repopulation charac-teristics are useful for in silico modelling of tumour growthand treatment response Similarly since it is well establishedthat tumourrsquos oxygenation level contributes significantly toits radioresistance modelling parameters that character-ize tumour hypoxia and angiogenesis are also commonIncreasingly in silico models have expanded to incorporateadditional contributing factors to tumour radiosensitivityincluding intracellular tumour pH gene expression andcell-cycle simulation The complexity of any given modeldictates the number of characteristics simulated as well as theaccuracy with which each tumour characteristic is simulatedSimpler models may use averaged macroscopic measures ofa given tumour characteristic while more complex modelsmay simulate a tumour and its behaviour at the microscopicmolecular and even atomic levels Indeed some of themost accurate contemporary models are ldquomultiscalerdquo theysimulate tumour behaviour across multiple biological scalesreconciling the macroscopic and microscopic levels [18]Details of the various parameters utilized for tumour mod-elling obtainable from PET and PETCT data are providedbelow

21 Cell Proliferation There are many approaches to mod-elling cell proliferation but it is useful to separately con-sider modelling at the macroscopic and microscopic scalesAt the macroscopic scale tumours are commonly simu-lated using continuum models Using this approach grosstumour morphology and behaviour are modelled undervarious environmental conditions and typically governed

Computational and Mathematical Methods in Medicine 3

by a set of differential equations whose initial condi-tions serve as input parameters [18] Input parametersinclude such quantities as cell density (both viable andnecrotic) cell volume fractions and gross tumour prolif-eration and metabolic rates [19] Such parameters may begathered from PETCT data where a corresponding radio-tracer can be utilized [20] Details of various radiotracersand the functional information they provide are given inSection 3

In contrast to continuum tumour models discretetumour models developed at the microscopic scale simulateindividual cell behaviour Their input parameters corre-spondingly describe a set of biophysical rules applied to themodelled cells and the diversity of these rules varies acrossdifferent models Each cell is assigned an initial state andits status is subsequently tracked throughout the simulationConsequently discrete models directly simulate cell prolifer-ation a process which is mediated by input parameters suchas proliferative potential cell-cycle positions and durationsfor different cell types probability of cell division cell lossrates and maximum tumour radius or cell number [19]Patient-specific parameters such as tumour radius and prolif-erating cells per voxel are directly obtainable from functionalimaging data [15]

22 Hypoxia The adverse effects of tumour hypoxia onpatient outcomes following radiotherapy have been wellestablished Indeed hypoxia is often attributed to poortumour control probabilities in locally advanced head andneck cancers for which low oxygenation levels are common[21] Accordingly the simulation of oxygenation in tumourshas been a prolific area of research in the tumour modellingcommunity for more than 60 years [22] The simulationof tumour hypoxia varies greatly across different modelsthe most complex models concurrently simulate oxygendiffusion oxygen consumption by tissue and the interdepen-dence of oxygen with tumour proliferation and vasculature[23]

Oxygen information is generally modelled using ananalytic approach That is sets of differential equations areused to describe oxygen diffusion andor consumption ratesboth during tumour growth and in response to treatmentEven stochastic (discrete) tumour cell proliferation modelsthat incorporate oxygen information typically employ a com-posite approach oxygen and other substrate concentrationsare generally governed by continuous fields Since the oxygendistribution within a tumour is directly related to the extentand nature of its vasculature many tumour hypoxia modelsincorporate blood vessel information into their simulationsThe most sophisticated models simulate angiogenesis (seeSection 23) and its effect on tumour hypoxia [16 23 24]

Simulated oxygen distributions are typically quantifiedusing partial oxygen tension or pO

2values Thus for the

successful simulation of tumour hypoxia realistic initial pO2

conditions must be set In order to categorize the oxygenstatus of tissue binary approaches to oxygen modellingestablish a defined threshold pO

2value below which the

region is considered to be hypoxicMore robustmodels estab-lish a more detailed relationship between tumour growth

or treatment response and oxygen information and mayincorporate both oxygen diffusion and oxygen consumptionby tissue [23 25] Additional parameters of interest may beincluded for oxygen effects such as reoxygenation proba-bility distributions and hypoxic thresholds for which cellquiescence is initiated Multiple studies have demonstratedthe utility of PET for obtaining pO

2distributions for use in

patient-specific tumour models [16 23 26]

23 Angiogenesis The extent and nature of tumour vascula-ture significantly influence tumour growth and oxygen statusConsequently the simulation of angiogenesis serves as auseful complement tomodels of tumour cell proliferation andhypoxia Likecell proliferation models of tumour-inducedangiogenesis may be either continuous or discrete in natureMore advancedmodels utilize both approaches for simulatingthe various mechanisms involved in angiogenesis such ascapillary sprout formation endothelial cell migration bloodflow and vessel adaptation [19]

Input parameters related to vessel branching generallyinclude probabilities of random endothelial cell migrationchemotaxis with tumour angiogenesis factor concentration(especially vascular endothelial growth factor or VEGF)and haptotaxis with fibronectin gradients [18] The utilityof PET for the imaging of specific angiogenic markerssuch as 120572v1205733 integrins and tumour expression of VEGFhas been demonstrated by multiple groups [28 29] Todate tumour models of angiogenesis have traditionally notincorporated angiogenesis-specific PET data However sincecellular oxygenation and tumour vasculature are intimatelyrelated phenomena hypoxia-specific PET data has beenutilised in models of tumour vasculature [16 23] Specificallythe value of PET information for the conversion of oxygenmaps into capillary density maps has been demonstratedThe potential of angiogenesis-specific PET-based imaging forinput in modelling the temporal development of vasculatureor angiogenesis is well understood and will develop withongoing studies [16]

24 pH Though the intracellular pH of solid tumours ismaintained in a range similar to that of normal cells theextracellular pH of solid tumours is commonly acidic Theincreased glucose metabolism of solid tumours assistedby characteristically poor perfusion is the most probablecause of their low extracellular pH This is because glucosecatabolism results in net acid production and insufficientvasculature cannot remove excess acid from the extracellularenvironment [30]

Perhaps the most compelling value of including pHin tumour growth models arises from the acid-mediatedtumour invasion hypothesis This suggests that tumour cellsdevelop phenotypic adaptations to the harmful effects ofacidosis during carcinogenesis traits that are not presentin normal cells Consequently tumour cells are renderedrelatively impervious to the decreased pH in the tumourmicroenvironment resulting from increased anaerobic gly-colysis which is otherwise toxic to normal tissue Effectivelythe tumour provides for itself a selective growth advantageand a useful mechanism for invasion [31] Investigation of


Table 1 PET Radiotracers whose data has beenthe potential to be incorporated into in silicomodels

PET radiotracer Functional characteristic Corresponding tumour model parameters Use in in silicomodels

FDG Glucose metabolism (i) Intracellular Volume Fraction (ICVF) Yes(ii) Acid production rateslowast

FLT DNA replication Tumour cell proliferative rates (vector- or voxel-based) Yes

FMISO Hypoxia (i) Partial oxygen tension (pO2) Yes(ii) Relative hypoxic fraction (RH)

Cu-ATSM Hypoxia Partial oxygen tension (pO2) YespHLIP Acidosis Extracellular pH (pHe) PotentialGalacto-RGD Angiogenesis (i) 120572V1205733 expression rate PotentialFDOPA Malignancy (ii) L-DOPA activity PotentialFES Malignancy (iii) Oestrogen overexpression PotentiallowastThe specificity of FDG to glucose metabolism provides in indirect measure of acid production rates in tumour cells since anaerobic glycolysis is net acidproducing

this phenomenon is particularly suited for in silico tumourgrowth modelling and the role played by acid gradients intriggering tumour invasion has been evaluated this way [31]Excess H+ ion concentration may be simulated in a con-tinuummodel utilizing reaction-diffusion partial differentialequations where input parameters such as acid productionrate reabsorption rate and H+ ion diffusion coefficients maybe obtained from measurement [32]

As with angiogenesis in silico models incorporatingtumour pH have not utilized noninvasive PET data for inputThough PET has been used for the measurement of pH sincethe 1970s and was the first noninvasive in vivo pH meterit has historically been both an inaccurate and imprecisemeasurement tool [30] However the development of novelradiotracers that selectively target acidic tumours will enablethe incorporation of pH related PET data into in silicomodelsof tumour growth [33]

3 Radiotracers Used for Tumour Modelling

PET and PETCT are able to image an increasing varietyof physiological phenomena This versatility arises fromthe ability to select a radiotracer that specifically targetsa particular mechanism Additionally the diversity of PETtracers continues to expand with ongoing innovations inradiopharmaceutical production Today radiotracers existfor the imaging of metabolism proliferation perfusiondrugreceptor interactions and gene expression Despite thisvariety the extent of PET data incorporated into in silicotumourmodels has so far been limited to radiotracers specificto glucose metabolism cell proliferation and hypoxia Thereis significant potential for the use of alternative radiotracers toobtain additional functional information for in silicomodelsusing PET A list of radiotracers whose information has beendirectly incorporated into in silicomodels as well as those thatshow significant promise for such applications is provided inTable 1

31 FDG 18F-2-Fluoro-2-deoxy-glucose or FDG is by farthe most commonly used and extensively researched PETradiotracer Today FDG-PET plays an important role in

oncology It has been recommended for use as an imagingtool additional to traditional radiological modalities in theappropriate clinical setting In particular it has demonstratedefficacy in the diagnosis staging unknown primary discov-ery and the detection of cancer recurrence [1]

Increased glucose consumption is a typical characteristicof most cancers In hypoxic regions the Pasteur effect resultsin the upregulation of anaerobic glycolysis and the GLUT 1glucose transporter in tumour cells However even if oxygenis plentiful cancers undergo accelerated glycolysis Thisobservation called theWarburg effect is widely attributed tomutations in oncogenes and tumour suppressor genes [34]Since FDG is a glucose analogue it is a particularly suitableradiotracer to measure the increased glucose utilization typ-ical of cancers Along with increased glucose consumptionthe upregulation of appropriate enzymatic activity furtheramplifies FDG uptake in tumour cells

The primary drawback of FDG-PET for oncologic imag-ing is that FDG uptake is not specific to cancerThat is FDG-PET exhibits a poor level of specificity for certain applica-tions FDG uptake may be intense in benign diseases as wellas in areas of infectious disease and inflammatory tissueThat is there are many potential causes of false-positive PETsignals in oncologic imaging [1 35 36] Conversely somemalignant diseases do not exhibit high glycolytic activityBronchioloalveolar carcinoma and carcinoid tumours areexamples of cancers for which false-negative signals mayoccur for standalone FDG-PET imaging [35] CombiningFDG-PET with other imaging modalities has served tomediate this drawback somewhat FDG-PETCT has demon-strated superior performance than standalone FDG-PETin common cancers [37] Additionally the emergence ofnovel radiotracers that target biochemical processes that aremore specific to cancer promises to overcome the relativenonspecificity of FDG-PET in oncologic imaging

Commensurate with the predominance of FDG as thePET radiotracer of choice metabolic information providedby FDG-PET is often utilized in in silico models of bothtumour growth and treatment response [16 20] The specificinformation employed from FDG-PET varies across suchmodels Images may be used solely to identify existing


cancerous tissue particularly in simulations for the predic-tion of tumour response to therapy [16] Alternatively glucosemetabolism data from FDG-PET can be quantitatively usedto simulate metabolic processes in predictive models oftumour growth [20]

32 FLT PET radiotracers specific to cell proliferationare an effective alternative to those specific to glucosemetabolism such as FDG Of these tracers 18F-3-fluoro-3-deoxy-thymidine (FLT) is perhaps the most researchedand the most utilized FLT-PET is typically a less sensitiveimaging modality than FDG-PET the difference in FLTuptake between normal and malignant tissues is usually lesspronounced than that for FDG [38] However FLT uptakecorrelates very well with Ki-67 an index of cell proliferationConsequently FLT-PET is useful for aiding in the grading oftumours The combined FLT-PETCT modality has demon-strated efficacy for the early prediction of treatment response[39ndash41] as well as the assessment of cancer aggressiveness[42] The uptake of FLT in infectious or inflammatory tissueis less than that of FDG and FLT has lower backgroundactivity in the brain and thorax [43 44] Consequently thespecificity of FLT-PETCT exceeds that of FDG-PETCT incertain imaging applications [44]

The magnitude of FLT that is trapped in a tumourcell is proportional to the amount of DNARNA synthesisundertaken by the cell Since the growth of malignanttissue is intricately related to DNA replication the degreeof FLT uptake is strongly correlated with proliferation rateAccordingly FLT-PET data is particularly useful for patient-specific tumourmodelling where proliferative rates are oftenof paramount interest A group led by Benjamin Titz at theUniversity of Wisconsin has correspondingly acquired cellproliferation information from FLT-PET data for in silicomodels of tumour growth and treatment response [15 16]

33 FMISO In an effort to develop accurate noninvasivemeasurement techniques of tumour hypoxia a number ofPET radiotracers have been produced which irreversiblybind to cells in poorly oxygenated conditions Of these 18F-fluoromisonidazole (FMISO) is the most extensively studiedand clinically validated FMISO uptake is inversely propor-tional toO

2level and perfusion does not restrict its delivery to

malignant tissue Several studies have demonstrated FMISO-PET to be a viable prognostic indicator of tumour responseto treatment [45]

FMISO-PET has not been universally adopted into rou-tine clinical application because of a number of limitationsinherent in the radiotracer Since it relies on passive transportmechanisms its uptake is relatively slow in hypoxic tumoursusually requiring 2ndash4 hours to be selectively retained in thetarget following injection [46] FMISO-PET imaging alsoexhibits relatively low tumour-to-background ratios sinceits binding to malignant hypoxic cells is highly nonspe-cific Finally considerable levels of unwanted radioactivemetabolite products result from the nonoxygen dependentmetabolism of FMISO [45] Improvements in the quantifi-cation of hypoxia by modelling FMISO-PET dynamics as

opposed to using the standardized uptake value (SUV) in abinary manner may aid in overcoming contrast limitationsSeveral studies have demonstrated reasonable success in thesimulation of tracer transport and its application to tumourmodels (see Section 5) [25ndash27]

34 Cu-ATSM Alternative hypoxia-specific PET radiotrac-ers have been developed in order to overcome the var-ious limitations of FMISO Several other nitroimidazolecompounds have been developed for this purpose [47] In1997 an alternative hypoxia PET tracer was proposed thatdoes not suffer from the undesirable radioactive residuesof nitroimidazoles This tracer is Cu(ll)-diacetyl-bis(N4-methylthiosemicarbazone) or Cu-ATSM [48] Cu-ATSM hasevolved to become one of the most promising PET agentsfor hypoxia imaging It has demonstrably high hypoxictissue selectivity [45] It is able to rapidly identify hypoxictissue with high tumour-to-background ratios due to acombination of small molecular weight high cell membranepermeability rapid blood clearance and prompt retention inhypoxic tissues [45]

The effectiveness of Cu-ATSM for providing clinicallyrelevant tumour oxygenation information has been con-firmed in multiple studies and its predictive value of tumourbehaviour and treatment response has been demonstrated[49ndash51] Perhaps unsurprisingly it is a preferred radiotracerfor the determination of oxygenation information using PETfor incorporation into in silico tumour models [15 16 23]For example Titz and Jeraj chose a sigmoidal relationshipbetween the SUV of Cu-ATSM and local tissue oxygenation[15] following the findings of Lewis et al [52] Using pretreat-ment Cu-ATSM-PET spatial maps of tumour oxygenationit was demonstrated that lower oxygen levels resulted inreduced treatment efficacy However the group did notethat further investigation into the quantitative relationshipbetween partial oxygen tension and Cu-ATSM uptake iswarranted

35 Other Radiotracers Although in silico models are yetto incorporate PET or PETCT information beyond glucosemetabolism cell proliferation and tumour oxygenationthere is scope for the use of tracers that image additionalprocesses Tumour acidosis arising from amplified glycolysisis a common feature of cancers and is a likely trigger ofinvasion into surrounding tissue [53] Consequently severalmathematical models inclusive of tumour acidity have beendeveloped to study the glycolytic phenotype and the tumour-host interface [31 54] There is suggestive potential of PETand particularly PETCT for directly obtaining parametersof interest for such models including glucose metabolicrate and acid production rates One promising novel PETtracer that specifically targets acidosis is pH low insertionpeptide (pHLIP) [33] pHLIP binds to acidic cell membranesand has demonstrated ability to target areas of hypoxiaand carbonic anhydrase IX (CAIX) overexpression an acid-extruding protein [55]

As discussed in Section 2 the simulation of angiogenesisis of significant interest for in silico tumour modelling


Though parameters related to angiogenesis can be indirectlyobtained using hypoxia-specific PET-imaging alternativemarkers of angiogenesis can instead be targeted For examplevascular integrins are targeted by PET radiotracers contain-ing the tripeptide sequence arginine-glycine-aspartic acid(RGD) [28] In particular the 120572V1205733 integrin is a receptorrelated to cell adhesion and involved in tumour-inducedangiogenesis that can be imaged using radiotracers such as18F-Galacto-RGD [29]

In models that simulate specific tumour types PET infor-mation from alternative tracers might be useful For exampleneuroendocrine tumours are typically characterised by anincreased L-DOPA decarboxylase activity [56] The imagingof advanced neuroendocrine tumours has been validatedwith PET using 18F-dihydroxyphenylalanine (FDOPA) [57]and may be of value in the modelling of these tumoursSimilar arguments may be made for the simulation of breastcancers and the imaging of oestrogen receptor expressionusing 18F labelled oestrogens such as 18F-fluoroestradiol(FES) [58]

4 Biophysical Parameters Used inPET and PETCT

The integration of reliable imaging-based information intoin silico models of tumour growth and treatment responsegreatly relies on the accurate and precise quantification ofimaging dataThis is especially relevant for PET and PETCTforwhich radiotracer uptake is dependent on a host of factorsSUV is the most extensively used parameter clinically for theanalysis of PET tracers but its high degree of sensitivity tomultiple variables can render the comparison of SUVs takenat different times or between different centres to be extremelydifficult [59]

Details of the biophysical parameters used to quantifyPET and PETCT data are provided belowTheir applicationsand limitations are discussed and compared as well as thevarious methods that have been developed to overcome thepotential pitfalls of a given measure The present use of PETand PETCT quantification measures in in silico models oftumour growth and treatment response is also discussed

41 SUV The standardized uptake value is the quintessentialparameter employed to analyse and quantify PET radiotracerdata It is defined as follows

SUV = radiotracer concentration in ROItotal injected activity119873

gmL (1)

where the concentration is as measured with PET in kBqmLROI is the region or volume elements of interest and 119873 is afactor normalizing for bodyweight body surface area or leanbody mass The overall denominator has units kBqg Theradiotracer is commonly computed by scanning the patientfor a 5ndash15-minute interval after a predetermined period (eg1 hour) after radiotracer injection

In general SUV depends on the time between injectionand scanning as well as multiple image acquisition settings

such as the reconstruction algorithm and scatter and atten-uation corrections [59] Its comparative value is hamperedby methodology differences such as choice of normalizationfactor and choice ofmaxmean peak or total SUV [60] SUVmay also be confounded by biological mechanisms such asvariations in plasma clearance before and after treatment andplasma glucose concentration (in the case of FDG) [59]

Despite its limitations SUV poses advantages over alter-native quantification techniques such as compartmentalmethods (see Section 42) It is the only method of PETquantitative analysis that can be realistically employed forroutine clinical use due to its sheer simplicity and theefficiency of the associated scan protocols In addition theeffectiveness of SUV for the assessment of cancer therapyresponse by comparing values for scans taken before and aftertreatment has been extensively validated This is particularlytrue for FDG-PET whose efficacy has been confirmed formultiple cancer types [59] Accordingly in spite of its largevariation in some situations the use of SUV is commonfor the acquisition of metabolic information proliferationrates and pO

2values for use in in silico tumour models

[16 20 23]

42 Compartmental Models Compartmental or kineticmodelling (CM) is the ldquogold standardrdquo of quantificationmethods for PET data [59] In CM the exchange of thePET radiotracer between a number of physiological entities(called compartments) is simulated These compartmentsare homogeneous in nature and the tracer transport andbinding rates between them is modelled by a set of first-orderdifferential equations [61] The set of equations are solvednumerically to obtain the rate constants kinetic parametersanalogous to those outlines in Section 2 such as glucosemetabolic rates or blood flow [59]

Despite its accuracy and relative independence of con-founding effects as compared to SUV CM has the disad-vantage of requiring a complex time-intensive acquisitionprotocol Techniques with which the requisite scan protocolcomplexity of CM can be overcome CM are an ongoing fieldof research [62 63] In the case of radiotracers such as FDGfor which the use of SUV has been strongly validated (viacomparison with CM in some cases) the added benefit ofemploying CM techniques for PET analysis is likely to beinsubstantial [59]

However in the case of FMISO-PET imaging of hypoxiathe use of CM is warranted Whilst FMISO can be used toidentify and image tumour hypoxia it typically exhibits poortumour-to-background ratios using the standard SUV mea-sure generating highly variable results [25 27] Compart-mental modelling of hypoxia imaging for dynamic FMISO-PET data has shown great success in ameliorating thisproblem with particularly promising contributions fromThorwarth et al [25] and Wang et al [27]

43 Hounsfield Units The process of positron emissiontomography is based on the coincident detection of colinear511 keV photons originating from an annihilation event Thismeasurement may be affected by an interaction between oneor both of the photons and the attenuator prior to reaching


the detectors A colinear coincident event is consequently notdetected and may instead register as scatter coincidence orno coincidence To account for this signal loss attenuationcorrections are performed during PET image reconstructionin an effort to salvage the true radiotracer distribution Untilthe development of PETCT attenuation correction in PETwas performed using a transmission scan taken immediatelyprior to the imaging scan effectively doubling the total scantime In modern PETCT scanners CT-based attenuationcorrection of PET images can be performed using the imme-diately available CT images 511 keV linear attenuation valuesare obtained from the Hounsfield unit (HU) data providedby the CT using an appropriate transformation scheme [64]usually a bilinear relationship

CT scanners convert attenuation coefficient distributions120583(119909 119910 119911) into HU for display Since attenuation is directlyproportional to attenuator density the HU of a particularvoxel may be interpreted as the density of the object withinthat voxel relative to that of water Typical scans consist ofimage noise within the range of 10ndash50HU corresponding toa relative error of 1ndash5 [65] Consequently CT is a powerfulimager of tumour density For oncologic scenarios in whichlesions and background tissues are characterized by similarHU values assessment with CT is facilitated by the use ofpositive and negative contrast agents Contrast enhancementin PETCT has been reported to improve lesion detectioncharacterization and localization in some clinical settings[66ndash68]

Positive contrast agents within PETCT may cause over-estimation of PET attenuation with contrast-enhanced CTbased attenuation corrections This can lead to artifactsof apparently increased tracer uptake in regions of highcontrast concentration within the PET image However suchartifacts can often be attributed to an underlying vessel andhence do not cause problems with image interpretation [65]Furthermore several research studies have confirmed theclinical insignificance of this effect since the typical SUVmeasure is negligibly affected by contrast [69 70]

5 Review of Computational andMathematical Models of Tumour Growthand Prediction to Treatment ResponseBased on PET Imaging Data

With the advances in technology the current imagingmodal-ities offer a great variety of biological biophysical and clinicalparameters to be further studied and implemented into com-plex tumour models Computational modelling is an ever-increasing area of research in tumour biology and therapyDepending on their design (ie continuum or discrete)models offer various levels of understanding of biologicalbiochemical and biophysical processes occurring in tumoursbefore and during treatment Models are versatile in terms ofinput parameters equations used phenomena simulated andend points While never perfectly illustrating the biologicalreality models are valuable complements to kinetic analysisof tumour growth and development treatment outcome

prediction patient selection and important decision-makingtowards personalized medicine

There are a large number of computational and mathe-matical tumour models that incorporate functional imagingdata in the scientific literature This is particularly true forPET and PETCT A detailed list is provided in Table 2 whichincludes models of tumour growth tumour characteristicsand response to treatment The aims of the models thecorresponding imaging techniques used and physiologicalparameters imaged and the relevant grouprsquos findings aregiven

Tumour growth models are an important initial stepwhen modelling treatment response Using dual-phase CTand FDG-PET imaging modalities Liu et al have developeda tumour growth model for pancreatic cancers [20 71] Theyhave introduced the intracellular volume fraction (ICVF) asbiomarker for the estimation and evaluation of the modelrsquosparameters based on longitudinal dual-phase CT imagesmeasured on pre- and postcontrast images SUVwas used as asemiquantitative measure of tumour metabolism (metabolicrate) which was further related to tumour proliferation rateThe model was validated by comparing the virtual tumourwith a real pancreatic tumour in terms of average ICVF dif-ference of tumour surface relative tumour volume differenceand average surface distance between the predicted tumoursurface and the CT-segmented (reference) tumour surface[20]

Perhaps the vast majority of the models address the chal-lenge of tumour hypoxia and neovascularization [15 16 2327] The approach used in the models varies among researchgroups Given that compartmental models are great tools inkinetic modelling of perfusion diffusion and pharmacoki-netics of various tracers they were chosen by some groups toquantitatively estimate the levels of hypoxia in head and necktumours and also to assess the hypoxic distributionwithin thetumour [25 27] Considering the controversies around SUVand its correlation with the partial oxygen tension (pO

2)

Thorwarth et al came to a practical conclusion wherebycompartmental kinetic models are more reliable for hypoxiaassessment than early static SUV measurements due to thelow uptake of FMISO by severely hypoxic cells

Experimental probability density functions were em-ployed by other groups to simulate the direction and spatialarrangement of microvascular tumour density using patient-specific PET imaging information [23] The discovered cor-relation between microvessel density and tumour oxygena-tion levels (ie pO

2) suggests that patient-based simulation

can contribute towards individualized patient planning andtreatment

Hybrid models (or multiscale models) are often usedfor complex assessment of tumour growth and behaviourunder therapy due to their versatility and ability to integratemathematicalcomputational modelling with experimentaldata on different physical scalesThe hybridmodel developedby Titz and Jeraj is an example of this [15] Depending onthe input parameters chosen in terms of relevance and reli-ability such hybrid models can predict with high accuracytumour response to various treatments As illustrated inSection 3 functional imaging and particularly PET imaging


Table 2 Models of tumour growth and prediction to treatment response based on PET imaging data

Aim of the model[references]

Imaging techniqueused Model parameters Resultsobservations

Models of tumour growth

Spatial-temporalcharacterization ofpancreatic tumourgrowth and progression[20]

Dual-phase CT andFDG-PET

Intracellular Volume Fraction (ICVF)which reflects tumour cell invasion andSUV used for determination of cellmetabolic rate growth rate cell motiondiffusion and advection (for mass effect)

The model was successfully validatedagainst a real tumour using average ICVFdifference of tumour surface relativetumour volume difference amp averagesurface distance between predicted andsegmented tumour surface

Models of tumour characteristics

Evaluation of tumourhypoxia in head andneck tumours [25]

Dynamic FMISO-PET

Tracer transport and diffusion modelvoxel-based data analysis used todecompose time-activity curves intocomponents for perfusion diffusion andhypoxia-induced retention

Quantification of hypoxia hypoxicregions are spatially separated from bloodvessels tracer uptake occurs in viablehypoxic cells-onlyThe kinetic model is more accurate thanstatic SUV values

Simulation of tumouroxygenation [26] Dynamic FMISO-PET

Model input parameters for steady-stateO2 distribution 2D vascular map oxygentension and rate of oxygen consumptionBinding rates of FMISO estimated andspatial-temporal O2 distribution foundProbability density function was used tomodel tumour vasculature to identifyhypoxic sub-regions

Hypoxic sub-region distribution andshape resulting from the simulation agreewith real imaging data It was shown thatthe extent of vasculature is of greaterimportance than the level of tissueoxygen supply The model allows forquantitative analysis of tumourparameters when physiological changesoccur in tumour microenvironment

Estimation of tumourhypoxia in head andneck tumours [27]

Dynamic FMISO-PET

Region of interest and arterial blood areidentified via PET Values of kineticparameters (for oxic hypoxic andnecrotic areas) are taken fromPET-scanned patient data

Voxel-based compartmental analysis isfeasible to quantify tumour hypoxia andmore reliable than static PET-SUVmeasurements

Simulation of tumourvasculature [23]

Cu-ATSM PET andcontrast CT

Capillaries were simulated usingprobability density functions(micro-vessel density) and patientimaging data Capillary diameter wasmodelled in conjunction with voxel sizea relationship between vessel density andpO2 was employed

Simulation of homogenous andheterogeneous oxygen and vasculardistribution The model was tested onmouse tumour the simulated vasculatureand the Cu-ATSM PET hypoxia maprepresent the image-based hypoxiadistribution The model can be used foranti-angiogenic treatment simulation

Models of treatment response

Tumour growth andresponse model withhypoxia effects [15]

18F-FLT (forproliferation) ampCu-ATSM PET(for hypoxia) and CT

CT used for tumour anatomy Behaviourof tumour voxels modelled upon PETdata FLT uptake was used as proliferationindex A sigmoid relationship wasconsidered between Cu-ATSM SUV andpO2 The Linear Quadratic model wasused for cell survival

The model accurately reproduced tumourbehaviour for different oxygendistribution patterns Treatmentsimulations resulted in poor control forhypoxic tumours heterogeneous oxygendistribution resulted in heterogeneoustumour response (ie higher survivalamong hypoxic cells)y

Evaluation of tumourresponse toanti-angiogenic therapy[16]

18F-FDG (for metabolicactivity)18F-FLT (forproliferation) ampCu-ATSM PET(for hypoxia) and CT

Model based on previous work [15] withan added vascular componentMicrovessel density was used as modelparameter in direct relationship with thevascular growth fraction Probabilitydensity functions were used to samplecapillary properties and geometry

The maximum vascular growth fractionwas found to be the most sensitive modelparameter The dosage of theanti-angiogenic agent bevacizumab canbe adjusted to improve oxygenation Themodel was validated on imaging data of aphase I trial with bevacizumab on headand neck cancer patients


employing tumour-specific radiotracers play an importantrole in fulfilling this task Therefore information regardingtumour kinetics and proliferation can be obtained fromproliferation-specific agents (such as FLT) while oxygendistribution data is gained from hypoxia-specific radio-tracers (such as FMISO or Cu-ATSM) Additionally withongoing radiotracer development and evaluation thereis scope for obtaining additional tumour characteristicssuch as acidosis using pHLIP and gene expression usingprotein-specific agents such as Galacto-RGD FDOPA andFES

To further prove the usefulness of complex multiscalemodels the same group has simulated the effect of beva-cizumab an anti-VEGF agent which is administered fortargeting endothelial cell population in tumours [16] Tumourhypoxia and proliferation data were gathered from PETimages taken before and after the antiangiogenic treatmentSimulated hypoxia levels were compared with mean SUVvalues and changes in mean SUV after the administrationof bevacizumab for various levels of hypoxia proliferationand VEGF expression were analysed The findings wereimplemented on imaging data of a phase I clinical trial thatinvolved eight head and neck cancer patients showing thepotential of such models to optimise treatment outcome

6 Conclusion

The incorporation of patient-specific data into multiscalemodels is necessary for individualized predictive simula-tion This is an essential component of predictive oncologyImage-based information can be transformed into inputparameters and incorporated into either probabilistic ordeterministic equations governing their relationships andinterdependences Using these tools countless hypothesescan then be generated and scenarios of ldquowhat if rdquo can besimulated and solved Models usually have the benefit ofindependence from the manner in which input parametersare obtained This allows for the constant refinement ofparameters with future innovations in measurement tech-niques particularly in PET and PETCT Additionally mostmodels can be readily adapted to include new parametersin order to better resemble the real tumour environmentThe widespread continuing research into in silico modeldevelopment and refinement permits the simulation of can-cer with ever-increasing accuracy with the goal of optimallyindividualizing cancer management and improving overallpatient outcome

Conflict of Interests

The authors declare that there is no conflict of interestsregarding the publication of this paper

Acknowledgment

L G Marcu would like to acknowledge the support offeredby a Grant of the Ministry of National Education CNCS-UEFISCDI Project no PN-II-ID-PCE-2012-4-0067

References

[1] J W Fletcher B Djulbegovic H P Soares et al ldquoRecommen-dations on the use of 18F-FDG PET in oncologyrdquoThe Journal ofNuclear Medicine vol 49 no 3 pp 480ndash508 2008

[2] O Israel and A Kuten ldquoEarly detection of cancer recurrence18F-FDG PETCT can make a difference in diagnosis andpatient carerdquoThe Journal of Nuclear Medicine vol 48 no 1 pp28Sndash35S 2007

[3] D Papathanassiou C Bruna-Muraille J-C Liehn T DNguyen and H Cure ldquoPositron emission tomography inoncology present and future of PET and PETCTrdquo CriticalReviews in OncologyHematology vol 72 no 3 pp 239ndash2542009

[4] S S Gambhir J Czernin J Schwimmer D H S Silverman RE Coleman and M E Phelps ldquoA tabulated summary of theFDG PET literaturerdquo The Journal of Nuclear Medicine vol 42no 5 pp 1Sndash93S 2001

[5] T M Blodgett C C Meltzer and D W Townsend ldquoPETCTform and functionrdquoRadiology vol 242 no 2 pp 360ndash385 2007

[6] T Beyer D W Townsend T Brun et al ldquoA combined PETCTscanner for clinical oncologyrdquoThe Journal of Nuclear Medicinevol 41 no 8 pp 1369ndash1379 2000

[7] T Ishikita N Oriuchi T Higuchi et al ldquoAdditional value ofintegrated PETCT over PET alone in the initial staging andfollow up of head and neck malignancyrdquo Annals of NuclearMedicine vol 24 no 2 pp 77ndash82 2010

[8] R J Cerfolio B Ojha A S Bryant V Raghuveer J M Mountzand A A Bartolucci ldquoThe accuracy of integrated PET-CTcompared with dedicated PET alone for the staging of patientswith nonsmall cell lung cancerrdquo Annals of Thoracic Surgery vol78 no 3 pp 1017ndash1023 2004

[9] G Antoch J Stattaus A T Nemat et al ldquoNon-small celllung cancer dual-modality PETCT in preoperative stagingrdquoRadiology vol 229 no 2 pp 526ndash533 2003

[10] W D Wever S Ceyssens L Mortelmans et al ldquoAdditionalvalue of PET-CT in the staging of lung cancer ComparisonwithCT alone PET alone and visual correlation of PET and CTrdquoEuropean Radiology vol 17 no 1 pp 23ndash32 2007

[11] G W Goerres G K von Schulthess and H C Steinert ldquoWhymost PET of lung and head-and-neck cancer will be PETCTrdquoThe Journal of NuclearMedicine vol 45 no 1 pp 66Sndash71S 2004

[12] A Shammas B Degirmenci J M Mountz et al ldquo 18F-FDGPETCT in patients with suspected recurrent ormetastatic well-differentiated thyroid cancerrdquo The Journal of Nuclear Medicinevol 48 no 2 pp 221ndash226 2007

[13] M MacManus U Nestle K E Rosenzweig et al ldquoUse of PETand PETCT for Radiation Therapy Planning IAEA expertreport 2006ndash2007rdquoRadiotherapy andOncology vol 91 no 1 pp85ndash94 2009

[14] P Tracqui ldquoBiophysical models of tumour growthrdquo Reports onProgress in Physics vol 72 no 5 Article ID 056701 2009

[15] B Titz and R Jeraj ldquoAn imaging-based tumour growth andtreatment response model investigating the effect of tumouroxygenation on radiation therapy responserdquo Physics inMedicineand Biology vol 53 no 17 pp 4471ndash4488 2008

[16] B Titz K R Kozak and R Jeraj ldquoComputational modelling ofanti-angiogenic therapies based on multiparametric molecularimaging datardquo Physics in Medicine and Biology vol 57 no 19pp 6079ndash6101 2012

[17] S Sanga H B Frieboes X Zheng R Gatenby E L Bearer andV Cristini ldquoPredictive oncology a review of multidisciplinary


multiscale in silico modeling linking phenotype morphologyand growthrdquo NeuroImage vol 37 pp S120ndashS134 2007

[18] T S Deisboeck Z Wang P MacKlin and V Cristini ldquoMulti-scale cancer modelingrdquo Annual Review of Biomedical Engineer-ing vol 13 pp 127ndash155 2011

[19] J S Lowengrub H B Frieboes F Jin et al ldquoNonlinear mod-elling of cancer bridging the gap between cells and tumoursrdquoNonlinearity vol 23 no 1 pp R1ndashR9 2010

[20] Y Liu S M Sadowski A B Weisbrod E Kebebew R MSummers and J Yao ldquoPatient specific tumor growth predictionusing multimodal imagesrdquo Medical Image Analysis vol 18 no3 pp 555ndash566 2014

[21] D M Brizel G S Sibley L R Prosnitz R L Scher and MW Dewhirst ldquoTumor hypoxia adversely affects the prognosisof carcinoma of the head and neckrdquo International Journal ofRadiation Oncology Biology Physics vol 38 no 2 pp 285ndash2891997

[22] A L Harris ldquoHypoxiamdasha key regulatory factor in tumourgrowthrdquo Nature Reviews Cancer vol 2 no 1 pp 38ndash47 2002

[23] V Adhikarla and R Jeraj ldquoAn imaging-based stochastic modelfor simulation of tumour vasculaturerdquo Physics in Medicine andBiology vol 57 no 19 pp 6103ndash6124 2012

[24] W Tuckwell E Bezak E Yeoh and L Marcu ldquoEfficient MonteCarlo modelling of individual tumour cell propagation forhypoxic head and neck cancerrdquo Physics inMedicine and Biologyvol 53 no 17 pp 4489ndash4507 2008

[25] D Thorwarth S M Eschmann F Paulsen and M AlberldquoA kinetic model for dynamic 18F-Fmiso PET data to analysetumour hypoxiardquo Physics in Medicine and Biology vol 50 no10 pp 2209ndash2224 2005

[26] C J Kelly and M Brady ldquoA model to simulate tumouroxygenation and dynamic [18F]-Fmiso PET datardquo Physics inMedicine and Biology vol 51 pp 5859ndash5873 2006

[27] W Wang J-C Georgi S A Nehmeh et al ldquoEvaluation of acompartmentalmodel for estimating tumor hypoxia via FMISOdynamic PET imagingrdquo Physics inMedicine and Biology vol 54no 10 pp 3083ndash3099 2009

[28] R Haubner H J Wester W A Weber et al ldquoNoninvasiveimaging of 120572

1199071205733integrin expression using 18F-labeled RGD-

containing glycopeptide and positron emission tomographyrdquoCancer Research vol 61 no 5 pp 1781ndash1785 2001

[29] A J Beer R Haubner M Sarbia et al ldquoPositron emissiontomography using [18F]Galacto-RGD identifies the level ofintegrin 120572

1199071205733expression in manrdquo Clinical Cancer Research vol

12 no 13 pp 3942ndash3949 2006[30] X Zhang Y Lin and R J Gillies ldquoTumor pH and its measure-

mentrdquo Journal of Nuclear Medicine vol 51 no 8 pp 1167ndash11702010

[31] R A Gatenby E T Gawlinski A F Gmitro B Kaylor and RJ Gillies ldquoAcid-mediated tumor invasion a multidisciplinarystudyrdquo Cancer Research vol 66 no 10 pp 5216ndash5223 2006

[32] R A Gatenby and E T Gawlinski ldquoA reaction-diffusion modelof cancer invasionrdquo Cancer Research vol 56 no 24 pp 5745ndash5753 1996

[33] A L Vavere G B Biddlecombe W M Spees et al ldquoAnovel technology for the imaging of acidic prostate tumors bypositron emission tomographyrdquoCancer Research vol 69 no 10pp 4510ndash4516 2009

[34] D Grander ldquoHow domutated oncogenes and tumor suppressorgenes cause cancerrdquoMedical Oncology vol 15 no 1 pp 20ndash261998

[35] J M Chang H J Lee J M Goo et al ldquoFalse positive and falsenegative FDG-PET scans in various thoracic diseasesrdquo KoreanJournal of Radiology vol 7 no 1 pp 57ndash69 2006

[36] A D Culverwell A F Scarsbrook and F U ChowdhuryldquoFalse-positive uptake on 2-[18F]-fluoro-2-deoxy-D-glucose(FDG) positron-emission tomographycomputed tomography(PETCT) in oncological imagingrdquo Clinical Radiology vol 66no 4 pp 366ndash382 2011

[37] D Delbeke H Schoder W H Martin and R L Wahl ldquoHybridimaging (SPECTCT and PETCT) improving therapeuticdecisionsrdquo Seminars in Nuclear Medicine vol 39 no 5 pp 308ndash340 2009

[38] A K Buck G Halter H Schirrmeister et al ldquoImaging prolifer-ation in lung tumors with PET 18F-FLT versus 18F-FDGrdquo TheJournal of Nuclear Medicine vol 44 no 9 pp 1426ndash1431 2003

[39] W Chen S Delaloye D H S Silverman et al ldquoPredictingtreatment response of malignant gliomas to bevacizumab andirinotecan by imaging proliferation with [18F] fluorothymidinepositron emission tomography a pilot studyrdquo Journal of ClinicalOncology vol 25 no 30 pp 4714ndash4721 2007

[40] B S Pio C K Park R Pietras et al ldquoUsefulness of 31015840-[F-18]fluoro-31015840-deoxythymidine with positron emission tomogra-phy in predicting breast cancer response to therapyrdquoMolecularImaging and Biology vol 8 no 1 pp 36ndash42 2006

[41] H Barthel M C Cleij D R Collingridge et al ldquo31015840-Deoxy-31015840-[18F]fluorothymidine as a new marker for monitoring tumorresponse to antiproliferative therapy in vivo with positronemission tomographyrdquoCancer Research vol 63 no 13 pp 3791ndash3798 2003

[42] J S Rasey J R Grierson L W Wiens P D Kolb and J LSchwartz ldquoValidation of FLT uptake as a measure of thymidinekinase-1 activity in A549 carcinoma cellsrdquo The Journal ofNuclear Medicine vol 43 no 9 pp 1210ndash1217 2002

[43] W Chen T Cloughesy N Kamdar et al ldquoImaging proliferationin brain tumors with 18F-FLT PET comparison with 18F-FDGrdquoJournal of Nuclear Medicine vol 46 no 6 pp 945ndash952 2005

[44] A F Shields ldquoPET imaging with 18F-FLT and thymidineanalogs promise and pitfallsrdquoThe Journal of Nuclear Medicinevol 44 no 9 pp 1432ndash1434 2003

[45] G Mees R Dierckx C Vangestel and C van de WieleldquoMolecular imaging of hypoxia with radiolabelled agentsrdquoEuropean Journal of Nuclear Medicine and Molecular Imagingvol 36 no 10 pp 1674ndash1686 2009

[46] W J Koh J S Rasey M L Evans et al ldquoImaging of hypoxiain human tumors with [F-18]fluoromisonidazolerdquo InternationalJournal of Radiation Oncology Biology Physics vol 22 no 1 pp199ndash212 1992

[47] S T Lee and AM Scott ldquoHypoxia positron emission tomogra-phy imaging with 18F-fluoromisonidazolerdquo Seminars in NuclearMedicine vol 37 no 6 pp 451ndash461 2007

[48] Y Fujibayashi H Taniuchi Y Yonekura H Ohtani J Konishiand A Yokoyama ldquoCopper-62-ATSM a new hypoxia imagingagent with high membrane permeability and low redox poten-tialrdquo Journal of Nuclear Medicine vol 38 no 7 pp 1155ndash11601997

[49] F Dehdashti P W Grigsby M A Mintun J S Lewis B ASiegel and M J Welch ldquoAssessing tumor hypoxia in cervicalcancer by positron emission tomography with 60Cu-ATSMrelationship to therapeutic responsemdasha preliminary reportrdquoInternational Journal of Radiation Oncology Biology Physics vol55 no 5 pp 1233ndash1238 2003


[50] Y Minagawa K Shizukuishi I Koike et al ldquoAssessmentof tumor hypoxia by 62Cu-ATSM PETCT as a predictor ofresponse in head and neck cancer a pilot studyrdquo Annals ofNuclear Medicine vol 25 no 5 pp 339ndash345 2011

[51] J P Holland J S Lewis and F Dehdashti ldquoAssessing tumorhypoxia by positron emission tomography with Cu-ATSMrdquoTheQuarterly Journal of Nuclear Medicine and Molecular Imagingvol 53 no 2 pp 193ndash200 2009

[52] J S Lewis D W McCarthy T J McCarthy Y Fujibayashi andM J Welch ldquoEvaluation of 64Cu-ATSM in vitro and in vivo ina hypoxic tumor modelrdquo Journal of Nuclear Medicine vol 40no 1 pp 177ndash183 1999

[53] K Smallbone D J Gavaghan R Gatenby and P K MainildquoThe role of acidity in solid tumour growth and invasionrdquoJournal ofTheoretical Biology vol 235 no 4 pp 476ndash484 2005

[54] K Smallbone R A Gatenby and P K Maini ldquoMathematicalmodelling of tumour acidityrdquo Journal ofTheoretical Biology vol255 no 1 pp 106ndash112 2008

[55] N Viola-Villegas V Divilov O Andreev Y Reshetnyak andJ Lewis ldquoTowards the improvement of an acidosis-targetingpeptide PET tracerrdquo The Journal of Nuclear Medicine vol 53supplement 1 abstract no 1673 2012

[56] C Nanni S Fanti and D Rubello ldquo18F-DOPA PET andPETCTrdquo Journal of Nuclear Medicine vol 48 no 10 pp 1577ndash1579 2007

[57] A BechererM Szabo GKaranikas et al ldquoImaging of advancedneuroendocrine tumors with 18F-FDOPA PETrdquo The Journal ofNuclear Medicine vol 45 no 7 pp 1161ndash1167 2004

[58] L M Peterson D A Mankoff T Lawton et al ldquoQuantitativeimaging of estrogen receptor expression in breast cancer withPET and 18F-fluoroestradiolrdquo The Journal of Nuclear Medicinevol 49 no 3 pp 367ndash374 2008

[59] G Tomasi F Turkheimer and E Aboagye ldquoImportance ofquantification for the analysis of PET data in oncology reviewof current methods and trends for the futurerdquo MolecularImaging and Biology vol 14 no 2 pp 131ndash136 2012

[60] M Vanderhoek S B Perlman and R Jeraj ldquoImpact of differentstandardized uptake value measures on PET-based quantifica-tion of treatment responserdquo The Journal of Nuclear Medicinevol 54 no 8 pp 1188ndash1194 2013

[61] H Watabe Y Ikoma Y Kimura M Naganawa and M Shida-hara ldquoPET kinetic analysismdashcompartmental modelrdquo Annals ofNuclear Medicine vol 20 no 9 pp 583ndash588 2006

[62] L G Strauss A Dimitrakopoulou-Strauss and U HaberkornldquoShortened PET data acquisition protocol for the quantificationof 18F-FDG kineticsrdquo The Journal of Nuclear Medicine vol 44no 12 pp 1933ndash1939 2003

[63] L G Strauss L Pan C Cheng U Haberkorn and ADimitrakopoulou-Strauss ldquoShortened acquisition protocols forthe quantitative assessment of the 2-tissue-compartment modelusing dynamic PETCT18F-FDG studiesrdquo Journal of NuclearMedicine vol 52 no 3 pp 379ndash385 2011

[64] C Burger G Goerres S Schoenes A Buck A Lonn and Gvon Schulthess ldquoPET attenuation coefficients from CT imagesexperimental evaluation of the transformation of CT into PET511-keV attenuation coefficientsrdquo European Journal of NuclearMedicine and Molecular Imaging vol 29 no 7 pp 922ndash9272002

[65] Clinical PET-CT in Radiology Integrated Imaging in OncologySpringer Science+Business Media New York NY USA 2011

[66] A C Pfannenberg P Aschoff K Brechtel et al ldquoLow dose non-enhanced CT versus standard dose contrast-enhanced CT incombined PETCT protocols for staging and therapy planningin non-small cell lung cancerrdquo European Journal of NuclearMedicine and Molecular Imaging vol 34 no 1 pp 36ndash44 2007

[67] C G Cronin P Prakash andMA Blake ldquoOral and IV contrastagents for the CT portion of PETCTrdquoThe American Journal ofRoentgenology vol 195 no 1 pp W5ndashW13 2010

[68] S K Haerle K Strobel N Ahmad A Soltermann D TSchmid and S J Stoeckli ldquoContrast-enhanced 18F-FDG-PETCT for the assessment of necrotic lymph nodemetastasesrdquoHead and Neck vol 33 no 3 pp 324ndash329 2011

[69] E Dizendorf T F Hany A Buck G K Von Schulthess andC Burger ldquoCause and magnitude of the error induced by oralCT contrast agent in CT-based attenuation correction of PETemission studiesrdquo The Journal of Nuclear Medicine vol 44 no5 pp 732ndash738 2003

[70] O Mawlawi J J Erasmus R F Munden et al ldquoQuantifyingthe effect of IV contrast media on integrated PETCT clinicalevaluationrdquoTheAmerican Journal of Roentgenology vol 186 no2 pp 308ndash319 2006

[71] Y Liu S M Sadowski A B Weisbrod E Kebebew RM Summers and J Yao ldquoMultimodal image driven patientspecific tumor growth modelingrdquo Medical Image Computingand Computer-Assisted Intervention vol 16 no 3 pp 283ndash2902013

Submit your manuscripts athttpwwwhindawicom

Stem CellsInternational

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014


MEDIATORSINFLAMMATION

of


Behavioural Neurology

EndocrinologyInternational Journal of



Disease Markers


BioMed Research International

OncologyJournal of



Oxidative Medicine and Cellular Longevity


PPAR Research

The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014

Immunology ResearchHindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Journal of

ObesityJournal of



Computational and Mathematical Methods in Medicine

OphthalmologyJournal of


Diabetes ResearchJournal of



Research and TreatmentAIDS


Gastroenterology Research and Practice


Parkinsonrsquos Disease

Evidence-Based Complementary and Alternative Medicine

Volume 2014Hindawi Publishing Corporationhttpwwwhindawicom


both the diagnosis and the staging of cancer [3 4] In a varietyof clinical settings FDG-PET exhibits improved values ofsensitivity specificity or both Examples of clinical scenariosin which FDG-PET has demonstrated efficacy include theevaluation ofmass lesions the staging and restaging of cancerthe planning of radiotherapy treatments the monitoring oftherapy and the detection of cancer recurrence [3] Howeverthe primary drawback of functional imaging is the lack ofanatomic information that it provides This necessitates itsaccompaniment with structural imaging for application inclinical oncology [1]

In order to take advantage of their inherent benefits thecombined use of both anatomic and physiologic imagingmodalities is optimal The accurate structure localizationcapabilities of CT complement the mapping of normal andabnormal tissue function performed by PET Because of thisPET images are routinely read alongside CT images in orderto both distinguish and localize metabolic irregularities Inthe first instance this has been achieved via the coregistrationof separately acquired PET and CT images using fusionsoftware a technique which has especially proved to beeffective for brain imaging However coregistration of imagesof other anatomical regions poses significant challenges [5]Difficulties in the registration process primarily arise fromvarying patient positioning between the two image setsWhilst such difficulties are minimal for brain scans theymay be significant in other anatomical regions wherein theremay be substantial organ movement or deformation In lightof this the development and subsequent commercializationof the PETCT scanner in 2001 has generally addressed theissues affecting the coregistration of separately acquired PETand CT images [6]

The use of integrated PETCT in oncologic imaging hassince become a widespread field of research particularlywith the utilization of 18F-FDG This combined modalityovercomes some of the drawbacks that are characteristic ofstandalone PET scanning namely the significant presence ofnoise in attenuation correction factors the lengthy durationof scans and the absence of anatomic markers [7] While thevast majority of published research concerns standalone PETPETCT has begun to show great promise across a multitudeof clinical settings [3] Studies have shown significantlyimproved accuracy in the staging of nonsmall cell lung cancerwith PETCT over the separate performance of PET and CT[8ndash10] Some centres perform a series of PETCT scans onnonoperative head and neck cancer patients following eitherradiation therapy or chemotherapy due to its indispensablecombined anatomical and functional information [3 7 11]For patients with recurrent or metastasized thyroid cancerthe localization capabilities of 18F-FDG-PETCT can lead toimproved diagnostic accuracy [12] Finally the incorporationof FDG-PETCT data into the radiation treatment planningprocess has repeatedly shown to improve target delineationand enhance the therapeutic ratio via its increased cancerstaging accuracy [13] The ongoing clinical evaluation andfurther innovation of PETCT namely via technologicalimprovements and the establishment of new tracers willcontinue to propel this technology into widespread use inoncology [3]

The significant interpatient variability in both tumourbehaviour and normal tissue response to cancer therapyhas prompted increasing demand for individualized treat-ment planning Simulations of radiobiological processes inmalignancies provide scope for the further optimization ofindividual treatment plans and to ultimately improve patientoutcomes The development of computer or in silico modelsof these radiobiological processes is particularly beneficial[14] It is here in which PET and PETCT as noninvasivefunctional imaging modalities can play a crucial role [15ndash17] The aim of this work is to review the application andeffectiveness of PET and PETCT for the in silico modellingof tumours In particular it examines the various parametersthat can be determined via PET and PETCT imaging alongwith the assortment of radiotracers available for these pur-poses The biophysical quantities used to quantify PET andPETCTdata including SUVand compartmentalmodels areassessed Finally an overview of the existing in silico modelsthat utilize PET or PETCT data is provided

2 Tumour Model Parameters Obtainable fromPET Imaging

With the objective of optimal individualized treatment plan-ningmany biological characteristics of a given tumour can beconsidered for computer simulation In the case of radiationtherapy the proliferative potential of a tumour and itsradiosensitivity are of particular interest Parameters whichdescribe tumour cell proliferation and repopulation charac-teristics are useful for in silico modelling of tumour growthand treatment response Similarly since it is well establishedthat tumourrsquos oxygenation level contributes significantly toits radioresistance modelling parameters that character-ize tumour hypoxia and angiogenesis are also commonIncreasingly in silico models have expanded to incorporateadditional contributing factors to tumour radiosensitivityincluding intracellular tumour pH gene expression andcell-cycle simulation The complexity of any given modeldictates the number of characteristics simulated as well as theaccuracy with which each tumour characteristic is simulatedSimpler models may use averaged macroscopic measures ofa given tumour characteristic while more complex modelsmay simulate a tumour and its behaviour at the microscopicmolecular and even atomic levels Indeed some of themost accurate contemporary models are ldquomultiscalerdquo theysimulate tumour behaviour across multiple biological scalesreconciling the macroscopic and microscopic levels [18]Details of the various parameters utilized for tumour mod-elling obtainable from PET and PETCT data are providedbelow

21 Cell Proliferation There are many approaches to mod-elling cell proliferation but it is useful to separately con-sider modelling at the macroscopic and microscopic scalesAt the macroscopic scale tumours are commonly simu-lated using continuum models Using this approach grosstumour morphology and behaviour are modelled undervarious environmental conditions and typically governed
























































gmL (1)






[16 20 23]















2)

















Dynamic FMISO-PET







Dynamic FMISO-PET



















6 Conclusion




Acknowledgment


References




















































































of






Disease Markers



OncologyJournal of





PPAR Research



Journal of

ObesityJournal of







































































gmL (1)






[16 20 23]















2)

















Dynamic FMISO-PET







Dynamic FMISO-PET



















6 Conclusion




Acknowledgment


References




















































































of






Disease Markers



OncologyJournal of





PPAR Research



Journal of

ObesityJournal of


























2)

















Dynamic FMISO-PET







Dynamic FMISO-PET



















6 Conclusion




Acknowledgment


References




















































































of






Disease Markers



OncologyJournal of





PPAR Research



Journal of

ObesityJournal of



























Dynamic FMISO-PET







Dynamic FMISO-PET



















6 Conclusion




Acknowledgment


References




















































































of






Disease Markers



OncologyJournal of





PPAR Research



Journal of

ObesityJournal of



















6 Conclusion




Acknowledgment


References




















































































of






Disease Markers



OncologyJournal of





PPAR Research



Journal of

ObesityJournal of


















































































of






Disease Markers



OncologyJournal of





PPAR Research



Journal of

ObesityJournal of
















review article pet-specific parameters and radiotracers in...

Documents