In nuclear medicine, the patternsseenin imagesrepresentspatial and temporal arrangementsandrearrangementsof the physiological or biochemi-cal processesunder investigation.How are thesepatternsbestdetectedandcompared?

J.M. Links,1994

Chapter 7

Inter pretation of 3D FusionofClinical SPECTand MR BrainImages:Preliminary Results

AbstractA preliminaryqualitative interpretationof 3D fusion for clinical casesis conducted.TheSPECTand MRI dataof two patientgroups, i.e., casesdiagnosedwith the Gilles de laTouretteSyndromeand Attention-Deficit Hyperactivity Disorder, are presentedusing theNormalFusiontechnique.Thesefusionimagesarefirstcomparedwith the2DSPECTimagesto assesswhetherNormalFusionaccuratelycharacterizesthefunctionaldataof thecorticalsurfacelayer. Thenthefusionimagesof theindividual casesarequalitatively evaluatedandthetwo patientgroupsarecompared.Theresultsindicatethatthesignalingof corticalactiv-ity by theNormalFusionimagesis in goodagreementwith 2D SPECTreconstructionsandthat specificpatternsof hypo or hyperactivity aremoreeasilyrecognizedwith the NormalFusiontechnique.

86 Interpretationof 3D Fusionof Clinical SPECTandMR Brain Images:PreliminaryResults

7.1 Intr oduction

Theroutineavailability of multiple imagingmodalitiesfor patientinvestigationpro-videsnew insightsto the clinician, but alsoposeschallengingdemandson the ex-tractionandpresentationof themultivariatedata.Thementalintegrationof multipletomographicdatasetsinto oneconsistent3D representationof the patientis an ex-tremelydifficult taskwhich calls for computerassistance.Integrationor fusion ofimagedatarequiresregistrationof thedifferentsetsinto thesamecoordinateframeandpresentationof theinformationin anintegratedfashion.

This thesisfocusseson techniquesfor the integratedvisualizationof functionalandanatomicalbraindata.Evaluationof severalof thesetechniquesis presentedinChapter6 for thelocalizationof abnormalities.Hereweperformapreliminaryinter-pretationof integratedvisualizationusingtheNormalFusionimagesfrom Chapter6.

In Figure4.2 NormalFusionimagesof two children,diagnosed,resp. with theGilles de la TouretteSyndrome(TS) andautisticbehavior, arepresented.The pat-ternsof cerebralblood flow observed in theseimagespromptedthis study. In thecurrentchaptertwo patientgroups, i.e., children diagnosedwith either TS (n=6)or Attention-DeficitHyperactivity Disorder(ADHD) (n=6), are used. Only thesetwo groupswereavailablewith a combinationof a sufficiently high resolutionMRIdatasetandHMPAO–SPECTimages.

Theaimof thischapteris twofold: i) To comparetheNormalFusionimageswiththe original 2D SPECTdatato determinewhetherthe integrated3D presentationsaccuratelycharacterizetheSPECTinformationof thesurfacelayerof thebrain. ii) Toqualitatively investigatepatternsof cerebralbloodflow in theintegratedimages.

We will first provide somebackgroundon bothTS andADHD, andthengive adescriptionof thepatientgroupsandthesetupof thestudy.

7.1.1 The Gilles de la TouretteSyndrome

TSis achronicneuropsychiatricdisorderthathasits onsetin childhoodandis primar-ily characterizedby multiplemotorandvocaltics. Virtually any musclegroupcanbeaffectedby tics,but theonsetis usuallyin theorofacialregion(Petersonetal. 1993).A geneticetiologyis supportedby family-geneticandtwin studies,with mostlikelya sex-influencedautosomaldominantmodeof inheritanceand variableexpressiv-ity asTS, chronicmotor tic disorderor obsessive-compulsive disorder(CohenandLeckman1994,Robertson1994).Subjectswith TSoftenexhibit comorbiddisordersin additionto thewide rangeof tics,suchasADHD in childrenandadolescentsandobsessive-compulsive disorderat adultage. About 40%of thechildrenandadoles-centswith TShaveacomorbidADHD.

A substantialbody of dataimplicatesthe basalgangliaandrelatedcorticalandthalamicstructuresin thepathophysiologyof TS(Leckmanetal.1992).Post-mortemstudieshave so far not establishedspecificabnormalities,exceptfor onecasewith

7.1Introduction 87

smallerneuronsin theputamenandcaudatenuclei(Richardson1982).Thissuggestshypoplasiaof partsof the basalgangliawhich may be dueto somedevelopmentalarrest. In vivo neuroradiologicstudiesprovide supportfor the involvementof thebasalganglia in the pathophysiology of TS. Structuralneuro-imagingby meansof3D reconstructedMRI hasdemonstratedsmallervolumesof thecaudate,lenticularand globus pallidus nuclei in TS subjectscomparedto controls. Furthermore,anabsenceof thenormalvolumetricasymmetry(left greaterthanright) wasshown inTSsubjects(Petersonetal. 1993,Singeretal. 1993).

A numberof SPECTstudieshaveexaminedthecerebralbloodflow in TS(mainlyadults)usingHMPAO. Thefirst study(Riddleetal. 1992)comparedninecontrolstonine right-handedadultswith TS. A significantlyreducedcerebralperfusionin theleft putamenand left globus pallidus of the TS subjectswas reported. A secondstudy (George et al. 1992) included20 adolescentsand adultswith TS and eightnormaladults,andfoundanincreasedright frontalactivity in TS.Thisfinding,how-ever, demandscautionbecausetheanalysisdid not includea correctionfor multiplecomparisonsand the frontal perfusionin the controlswasunusuallylow. A thirdstudy(Moriarty etal. 1995)including50subjects(agerange7 to 65)and20controlsfound significantlylower HMPAO uptake in the left caudateandanteriorcingulateareain TS.

In summary, the functionalandanatomicaldatasupportthe involvementof thecortico-striato-thalamo-corticalcircuitsin TS(AlexanderandCrutcher1990).

7.1.2 Attention-Deficit Hyperactivity Disorder

ADHD is a childhood-onsetdisordercharacterizedin DSM-IV by hyperactivity, im-pulsiveness,andpoorsustainedattention.Threesubtypeshave beendescribed:pre-dominantlyhyperactive/impulsive, predominantlyinattentive, andcombinedtypes.ADHD is themostprevalentpsychiatricdisorderin childhoodandoccursin 3-5%ofall childrenbetween7 and12 years.Boys areaboutfour timesmoreoftenaffectedthangirls. Thedisordershowsastrongpersistenceoverdevelopment;theprevalenceof ADHD amongadolescentsis 1.5-2%andamongadults0.5-1%.ThediagnosisofADHD is madeonthebasisof aclinical picture,currentlyno laboratorytestor setoftestscanbeusedto form adefinitediagnosis.(Cantwell1996)

Psychosocialfactorsarenot thoughtto playaprimaryetiologicalrole in ADHD,but family geneticfactorshavebeenimplicated.Furthermore,some”environmental”etiologicalfactors(includingpreandperinatalabnormalities,toxins,sugar intoxica-tion, andorthomoleculartheoriesof greatneedfor vitaminsandnutrientsin childrenwith ADHD) havebeenproposed.(Cantwell1996)

ADHD hasanenormousimpacton theutilizationof medicalandhealthcareser-vices. Both in clinical andepidemiologicsamplesADHD in childrenis associatedwith ahighrateof comorbiddisorders,suchasoppositionalandconduct(aggressive)disorders(50%),affective disorders(20%) andlearningdisorders(30%) (Cantwell

88 Interpretationof 3D Fusionof Clinical SPECTandMR Brain Images:PreliminaryResults

1996). The presenceof comorbidaggressionhasemergedasthe mostcritical fac-tor in exacerbatingthesedevelopmentalrisks. Childrenwith ADHD andcomorbidaggressionarealsofairly resistentto successfulbehavioral andpharmacologicinter-ventions.

Pathophysiology of ADHD hasbeeninvestigatedwith MRI (Castellanoset al.1994,Gieddet al. 1994),SPECTandPET(Lou et al. 1984,Zametkinet al. 1990),implicatingthe fronto-striatalareasasbeingabnormal.TheMRI studyhasdemon-stratedareducedvolumein therostrumandrostralbodyof thecorpuscallosum,anddecreasedvolumesof thecaudatenuclei.SteereandArnstein(1995)interpretedthisasanalterationof functioningof theprefrontalandanteriorcingulatecorticesof thebrainin additionto alteredpremotorfunction.

SPECTstudiesrevealedfocal cerebralhypoperfusionof striatumandhyperper-fusion in sensoryandsensorymotorareas.ThePETstudyin adultADHD revealedlowercerebralglucosemetabolismin thepremotorcortex andin thesuperiorprefron-tral cortex (Zametkinet al. 1990). Theseareasareinvolved in thecontrolof motoractivity andattention.

7.2 Materials and Methods

7.2.1 The patient data

At theUniversityHospitalUtrechtatotalof 41childrenandadolescentsarepartof anongoingstudyinvolvingMRI andSPECTacquisitions(for acquisitioncharacteristicsseeSection6.2.1).Fromthesecasesweselected16patientsonthebasisof absenceofcomorbiddisordersaccordingto DSM-IV criteria.Anotherfour caseswereremovedfrom theselectionowing to thelow qualityof thecorticalSPECTdatayieldingatotalof six TSandsix ADHD cases.Theageof theTSpatientsrangedfrom 8 to 13years,andthe ADHD rangewas8 to 11 years. All subjectswereon medication,but notduring theacquisitionof thedata. Subjectswith TS usedeitherclonidine(N=3) orlow dosesof neuroleptics(N=3). ADHD subjectsweretreatedwith methylphenidate(N=5) or clonidine(N=1).

Ethicalconcernsaboutradiationexposureprecludedtheuseof a normalcontrolgroup. All childrenandtheir parentsgave their informedconsentto participateinthis study, which wasapprovedby theethicalcommitteeof theUniversityHospitalUtrecht.Participantswerevolunteeredfrom childrenreferredto theOutpatientChildPsychiatricUnit, wherethey underwentextensive diagnosticprocedures,includingstandardizedpsychiatricexaminations,developmentalhistory, andneuropsychologi-cal testing.

7.3Results 89

7.2.2 Setup

For eachpatientNormalFusionvisualizationsweremadefor thesix principaldirec-tionssupplyingaroundaboutview of thebrain.Theimagesareviewedonscreenwithanoptionto manipulatethecolorencodingof thefunctionalinformationaccordingtotheapproachdiscussedin Chapter5. For quick comparisonof thevisualizationsfordifferentpatientswe alsousedcolor printsof theNormalFusionimages.For theseprints,colorencodingwasqualitatively setby theprincipalauthor(RS,seepagevii)by increasingtheamountof red to anarbitrarylevel andthenperformingthesamefor thecolor blue. Theauthorwasnot oneof theratersandmadeno useof clinicalinformationduringtheseoperations.

The quality of the 2D SPECTdatawas first investigatedin the usualway bya nuclearmedicinephysician(JvI). The Normal Fusionimagesweresubsequentlyinvestigatedandcomparedwith the 2D SPECTdatato evaluatetheir accuracy forsignalingof hypoandhyperperfusedareasin thecerebralcortex.

Finally, theNormalFusionvisualizationwerequalitatively evaluatedby thenu-clearmedicinephysician,aneuroradiologist(LM) andachild psychiatrist(JB)soasto interprettheindividualcasesandto comparethepatientgroupsTSandADHD.

7.3 Results

As a first steptheareasof increasedanddecreasedSPECTactivity indicatedby theNormalFusionvisualizationswerecomparedwith theinformationfrom theoriginal2D SPECTimagesandvice versa. An averagenumberof five cold-spotsor areasand seven hot-spotsor areaswere verified. Upon careful comparisonthe nuclearmedicinephysicianconcludedthatthe3D NormalFusionimagesaccuratelycharac-terizethecorrespondinginformationfrom the2D SPECTimages.For someof thesecomparisonscolormanipulationof theNormalFusionimagesprovedof importance.

Thesecondaim, i.e., thestudyof flow patterns,wasevaluatedby all observers.Table7.1presentsanoverview of themostrelevantobservationsandFigure7.1de-picts the Normal Fusionimagesof four of the cases,two for TS (TS1 and TS3)andtwo for ADHD (ADHD2 andADHD3). Theopinionof theobserverswasthattheNormalFusionimagespresenteda valuableandfastoverview of a largepartofthe functional informationof the grey matter. In combinationwith the individualanatomy, NormalFusionimageswereconsideredhelpful in the recognitionof per-fusioncharacteristicsof grey matterandin thegenerationof hypothesesfor corticalinvolvementin thesegroups.

90 Interpretationof 3D Fusionof Clinical SPECTandMR Brain Images:PreliminaryResults

A B

C D

Figure 7.1 Examplesof theNormalFusionvisualizationsfor theTS(toprow)andADHD (bottomrow) groups. Frame(A) is caseTS1,(B): TS3,(C): ADHD2, and(D): ADHD3(seeTable7.1).

7.4Discussion 91

Item Feature TS ADHD1 2 3 4 5 6 1 2 3 4 5 6

I R hyperoverL + + + + + + + +

II FTPconflR + + + + + + +

III FShypo L +p + +p +p +p +p +p +p +p +R +p + +p +p +p +p +p +p +p

IV Occ/Cerebellum L � � � � � � � � � � � �R � � � � � � � � � � � �

Table 7.1 The overall resultsof the qualitativeanalysisof the Normal Fusionimages.R = right, L = left, hyper= hyperperfusion,hypo= hypoperfusion,FTP =fronto-temporo-parietalregion,confl= confluent,FS= frontal superiorgyrus,Occ= occipital,’ � ’ = present,’ � ’ = lowerthan1, ’ � ’ = approximately1, ’ � ’ = higherthan1, The’p’ scoresin itemIII indicatethat thehypoperfusedareaextendedovertheparietal lobe. For anexplanationof thefeatureswereferto thetext.

Theevaluationof thecorticalactivity of theindividualcasesandsubsequentcom-parisonbetweentheTSandADHD patientgroups(seeFigure7.1)yieldedinterestingresults(seeTable7.1),from which thefollowing itemswerededuced:

I. Hyperperfusionof right cerebralareasin comparisonwith the correspondingleft cerebralareasfor theADHD group(n=5/6).

II. A patternof confluentregionsof hyperperfusionin the right fronto-temporo-parietalareafor theADHD patients(n=5/6).

III. Hypoperfusionin the left and/orright frontal superiorgyri in a majority ofcases(n=10/12). In mostof thesecases(n=8/10)the hypoperfusedareawasnot limited to the FS, but extendedover the parietallobe. This arealargelycorrespondswith thedistributionareaof thecerebralanteriorartery.

IV. Theoccipital lobeandthecerebellumarefrequentlyusedasa referencearea.We found that the ratio of theoccipital lobeover thecerebellum(Occ/cereb-ellum) is notconsistentfor theTSandADHD patients.In mostcases(n=8/12)theoccipitallobehasa lowerperfusionthanthecerebellum,which is in accor-dancewith the generalbelief that the cerebellumsignalsthe highestactivityfor the brain. However, the othercases(n=4/12)indicatethat at leastoneoftheseregionsis notsuitedasageneralreferencearea.

7.4 Discussion

The comparisonof the informationasprovidedby the NormalFusionimageswiththatof theoriginal2DSPECTdatacallsfor severalcomments.Analysisof functionaldatasuchasSPECTis typically enhancedby colormanipulationandthisalsoappliesto Normal Fusionpresentations.The useof the color manipulationcapabilitiesonscreenis favoredover theprintedimagesfor investigationof thedata.However, theprintedimageswereof considerablevaluefor quick comparisonbetweenmultiple

92 Interpretationof 3D Fusionof Clinical SPECTandMR Brain Images:PreliminaryResults

cases.Furthermore,we adviseagainst the stand-alonepresentationof the NormalFusionimagesfor viewing of the data. In our opinion the correlationwith the 2DSPECTinformationis vital for verificationof the observationsperformedwith thevolumetricdisplay(seealso(Webbetal. 1987,Wallis 1992)).

The resultsindicatethat thesignalingof corticalactivity by theNormalFusionimagesis in goodagreementwith 2D SPECTreconstructions.Thevisualcomparisonof theNormalFusionimageswith the2D SPECTdatarevealedthatunderstandingofthelatterwasimprovedby theintegratedvolumetricdisplay. Detectionof 3Dpatternsis problematicin a 2D SPECTdisplay, especiallywhenthe patternsarenot in oneof thethreeorthogonaldirections.In this respect,the’signaling’functionof NormalFusionis greatlyappreciated,becausespecificpatternsof hypo or hyperactivity aremoreeasilyrecognized.

In thepresentedpatientdatasets,NormalFusionvisualizationfacilitatedtherecog-nitionof theabovedescribed(Section7.3)abnormalflow patterns.Assessmentof theindividual casesandcomparisonbetweenthetwo patientgroupsyieldeda few note-worthy features.Forexample,ourobservationsindicatethattheuseof thecerebellumor thevisualcortex asa referenceareaashasbeenimplementedin severalpublica-tions(Georgeetal.1992,Moriartyetal.1995,Hashikawaetal.1995)appearsaques-tionablechoicefor thesepatientgroups(seealso(Lamoureuxet al. 1990,Crossonetal. 1994)ontheuseof thecerebellumasreferencearea).Until quantitativeSPECTwill becomeavailable(Beekmanet al. 1996),theuseof overall SPECTactivity asareferenceseemsmoreappropriate(Crossonetal. 1994).

In comparingthetwo patientgroups,theADHD casespresenteda relatively ho-mogeneouspatternof surfaceactivity. TheTSgroupwasquitediverse.

In thispreliminary, qualitatively evaluationtheNormalFusiontechniquepresentsa quick andaccuratecharacterizationof the cortical cerebralblood flow in clinicalcases.Recognitionof (possiblydisease-related)abnormalflow patternsis facilitated.However, theabsenceof anormalgroupandthelow numberof casespreventdefiniteconclusionsto bedrawn from thisstudy.