AnRDoC frameworkintegratingstructuralMRIwithcognitiveControlandWorkingMemory 1,3 2 2 2 3 ArianaE.Anderson ,MaxwellMansolf ,AnthonyRodriguez ,StevenP.Reise ,SpencerFrei ,Levon Demirdjian 3,RobertM.Bilder 1 (1)DepartmentofPsychiatryandBiobehavioral Sciences,(2)DepartmentofPsychology,(3)DepartmentofStatistics UniversityofCalifornia,LosAngeles INTRODUCTION Thegrowthofbiologicalpsychiatryoverthelast50yearshas increasedaccessibilityofmeasuresatmultiplelevelsfromthe genomicthroughneuroimagingofbraincircuitstructureandfunction tosophisticatedassessmentofcognitiveandbehavioralprocesses. TheRDoC initiativeaimstoleverageourcapacitytoassessthese underlyingdimensionswiththeaimtoprovidestrongerbasesfor biologicalpsychiatry.Here,weidentifystructuralMRIcorrelatesof workingmemory(visualandspatial)andcognitivecontrol(response inhibition)usingBayesiannetworksin278subjectswithandwithout aDSMdiagnoses. Units of Analyses DISC1 ... DRD1 non.WM.hypointensities HOPKINS_INTSENSITIVITY SWM_G ADHD BP Control SZ SubCortGrayVol SCHOOL_YRS RI_G Right.Thalamus.Proper VWM_G HOPKINS_OBSCOMP n Age Gender(%Male) Education(years) SomatizationScore(Hopkins ObsessiveCompulsive(Hopkins) 42.00 32.53 0.45 14.63 0.35 1.18 48.00 35.49 0.42 14.61 0.66 1.18 133.00 31.34 0.48 15.03 0.22 0.50 55.00 36.03 0.24 12.69 0.63 1.05 InterpersonalSensitivity(Hopkins) Depression(Hopkins) Anxiety(Hopkins) ResponseInhibition(g) VisualWorkingMemory(g) SpatialWorkingMemory(g) ChapmanPerceptionAberration ChapmanSocialAnhedonia ChapmanPhysicalAnhedonia 0.75 0.66 0.44 0.01 0.18 0.18 4.71 13.40 13.59 1.11 0.97 0.75 0.03 0.11 0.12 6.04 16.46 15.83 0.40 0.37 0.21 -0.28 0.28 0.37 3.14 9.62 11.47 0.94 0.84 0.84 0.37 -0.27 -0.20 8.93 15.14 17.41 Left.Amygdala CSF Left.Hippocampus HOPKINS_ANXIETY CHAPPER_TOTAL Left.Caudate Left.Pallidum CHAPSOC_TOTAL Right.Hippocampus HOPKINS_SOMATIZATION Right.Pallidum CHAPPHY_TOTAL Table1:ThefollowingmeasureswereusedinaBayesian networktoidentifythehierarchicalrelationshipaboutstructuralMRImarkers, demographicfeatures,diagnosis,andworkingmemoryandcognitivecontrol. Genes Right.Caudate Left.Accumbens.area PPC DLPFC Total Gray Volume ... CortexVol Circuits WM CC ADHDSelf-ReportScaleASRS-V1.1 X X AdultADHDInterview(modulefromKSADS-PL) X X StructuredClinicalInterviewforDSM-IV,Axis1,PatientVersion X X Left.Putamen Screening/Diagnosis/ClinicalRatingScales YoungManiaRatingScale SCAP SMNM ... WMS-IV Symbol Span Right.Putamen X ScaleforAssessmentofNegativeSymptoms X ScaleforAssessmentofPositiveSymptoms X BriefPsychiatricRatingScale X Right.Accumbens.area Brain.Stem Personality/Temperament/SymptomQuestionnaires ... Latent Specific Factor k X Akiskal'sBipolarIIScale X BarrattImpulsivityScale X Eysenck'sImpulsivity,VenturesomeandEmpathyInventory X MPQ(Control-Impulsivityitems) X TheDickmanScaleofFunctionalvsDysfunctionalImpulsivity X ChapmanScales(RevisedPhysicalAnhedonia;RevisedSocialAnhedonia;PerceptualAberrations METHODS Right.Cerebellum.White.Matter Left.Cerebellum.Cortex X VerbalMemoryandManipulationTask X SpatialMemoryandManipulationTask X VerbalWorkingMemoryCapacityTasks X SpatialWorkingMemoryCapacityTasks X WMS-IVSymbolSpan X WMS-IVDigitSpan X WMS-IVLetterNumberSequencing(LNS) X Right.Cerebellum.Cortex TotalGrayVol AGE Stop-SignalTask X CNPCPT X ReversalLearning X ColorTrailsTest X TaskSetSwitching X StroopTest X AttentionNetworksTask(ExecCenter) X DelayDiscounting X BalloonAnalogRiskTask X lhCorticalWhiteMatterVol rhCorticalWhiteMatterVol Right.Amygdala NeuroImagingMeasures FunctionalMRI(FunctionalNetworkConnectivityfromICA-derivedbrainnetworks[seetext]) X Physical Anhedonia (Chapman) and Left Cortical White Matter Volume Verbal Working Memory and Right Pallidum Volume Depression and Cortical Volume Somatization and Left Cerebellar Cortical Volume X 90000 300000 StructuralMRIvolumes:(totalbrain,totalcortical,DLPFC,VLPFC,PPC,totalgrey/whitecorticaland subcortical,Ventrofronto-striatal,hippocampal,thalamic,globus pallidus) X X 6e+05 80000 Genotyping OmniExpress X X 2000 Table2:Units ofanalysesavailableintheCNPdata RESULTS DX 5e+05 ADHD BP Control SZ 1500 250000 DX DX ADHD ADHD BP BP Control Control SZ SZ 70000 DX ADHD BP 60000 Control SZ 50000 200000 HomegeneousItemParcels Carelessness Haphazardness BlindActions G F1 F2 F3 F4 F5 Overspending SpendVsSave BuyingSpendingSprees AvoidSnapDecisions AvoidSimpleSnapDecisions QuickDecisions 4e+05 p2 0.52 0.48 0.48 0.44 0.76 0.77 0.23 0.26 0.40 0.69 0.64 0.36 0.25 0.68 0.50 0.50 0.07 0.48 0.27 0.73 0.07 0.61 0.40 0.60 0.05 0.38 0.34 0.28 0.72 0.53 0.56 0.44 0.51 0.49 0.61 0.58 0.41 0.54 0.46 0.60 0.32 0.57 0.44 0.56 0.22 0.35 0.60 0.50 0.50 0.25 0.36 0.85 0.85 0.15 0.15 0.30 0.74 0.64 0.36 0.14 0.33 0.79 0.74 0.26 0.15 0.37 0.35 0.29 0.71 0.48 Table3:Bifactor analysesofself-reportparcels,yieldingasinglecommondimensionand5specificfactors.h2iscommunality,u2isuniquevariance,andp2ispercentofcommonvariance attributabletothegeneralfactor.Loadings<.30notshown. 40000 1000 0 0.0 ThrillSeeking Deliberation2 u2 0.47 Activities2 Organization h2 0.51 SeekFrighteningThings Deliberation1 StructuralMRI,demographicinformation,andmeasuresofworking memory,responseinhibition,anhedonia(Chapmanscales),and otherdimensionsoftheHopkinssymptomchecklistFinally,these measurementswereintegratedusingaBayesiannetwork,where themostlikelyrelationshipspanningalldimensionswereassessed withoutassuminggivenhierarchyorcausality.Thisnetworkwas learnedusingthescore-basedmethod,connecting44nodeswith 133 arcs,fitusingtheBICforconditionalGaussian. Left.Cerebellum.White.Matter NeurocognitiveMeasures TheCNPdatasetconsistsofvolunteers(healthyandpatient)from thegreaterLosAngelesareawhoenrolledandcompleted diagnosticinterviews,self-reports,cognitiveexams,andcontributed geneticdata.SamplecharacteristicsarelistedinTable1. Neuroimagingdata(MRI,fMRI)werecollectedalongwithall personality,neurocognitivemeasuresandgenotypinglistedinTable 2 StructuralMRIdatawasprocessedusingFreeSurfer,extracting corticalandsubcorticalmeasurements. Hierarchicalclusteranalyseswasusedtoidentifyredundant variableswithnumerousoutcomemeasuresgeneratedbyeach neurocognitivetask.Next,exploratorybifactor analysiswasusedto developmeasurementmodelsforeachconstructofinterest, whereinthebifactor approachallowsustoexaminewhatasetof tasksshareincommon(generalfactor)anddata-drivenmodelsfor specificvariancewouldallowustoexaminerelationshipsamong neurocognitivevariablesthatremainafteraccountingforthe generalfactor. Compositemeasuresofcognitivecontroland workingmemorywerecreatedusingbifactor modeling,extracting “g”generalfactormeasurementsofvisualandspatialworking memoryseparately. EckbladandChapman'sHypomanicPersonalityScale Left Cortical White Matter Volume Latent Specific Factor 1 X Right Pallidum Volume Latent General Factor TheTemperamentandCharacterInventory Cortical Volume Self Reports Optic.Chiasm Left Cerebellar Cortex Volume Paradigms Left.Thalamus.Proper DX HOPKINS_DEPRESSION Variable WM.hypointensities GENDER Bayesian network modeling is a method of inference where the joint frequency of data is explained via a product of marginal distributions (9). In this framework, multimodal observations such as behavior, self-report, and paradigms represent observations recorded on a single subject. In the Bayesian networks, nodes represent random variables, and an arrow defines a probabilistic connection (dependence) among the variables. Formally, if the variables are defined as discrete, then the joint probability distribution can be defined as the product of the local probability distributions. Bayesian Network COMT RESULTS METHODS(Continued) 0.5 1.0 Depression (Hopkins) 1.5 2.0 1 2 Somatization (Hopkins) -2 -1 0 1 Verbal Working Memory 2 0 10 20 30 40 Physical Anhedonia (Chapman) CONCLUSIONS REFERENCES OurresultssuggestthatBayesiannetworkscanbeusedtoidentify thedependenciesbetweenstructuralMRImeasurementsand observedchangesincognitionandbehavior.Morebroadly, diagnosisstillmodulatestherelationshipbetweensymptom measuresandstructuralMRImeasures,suggestingthatdiagnosis maybecapturinginformationnotcontainedalreadyinthestructural andsymptombasedmeasures.Futureworkwillincludegenotype andfunctionalMRImeasurestofurtherassessthehierarchical structurespanningtheRDoC framework. 1. CronbachLJ,Meehl PE.Constructvalidityinpsychologicaltests.Psychologicalbulletin1955;52(4):281. 2. Borsboom D,Mellenbergh GJ,vanHeerden J.Theconceptofvalidity.Psychologicalreview2004;111(4):1061. 3. CuthbertBN,Kozak MJ.Constructingconstructsforpsychopathology:TheNIMHresearchdomaincriteria.2013. 4. Insel T,CuthbertB,GarveyM,Heinssen R,PineDS,QuinnK,Sanislow C,WangP.Researchdomaincriteria(RDoC):towardanew classificationframeworkforresearchonmentaldisorders.AmericanJournalofPsychiatry2010;167(7):748-751. 5. CuthbertBN,Insel TR.Towardthefutureofpsychiatricdiagnosis:thesevenpillarsofRDoC.BMCMedicine2013;11(1):126. 6. Asparouhov T,Muthén B.Exploratorystructuralequationmodeling.StructuralEquationModeling:AMultidisciplinaryJournal 2009;16(3):397-438. 7. Kievit RA,vanRooijen H,Wicherts JM,Waldorp LJ,Kan K-J,ScholteHS,Borsboom D.Intelligenceandthebrain:Amodel-based approach.Cognitiveneuroscience2012;3(2):89-97. 8. GreenlandS,Brumback B.Anoverviewofrelationsamongcausalmodellingmethods.Internationaljournalofepidemiology 2002;31(5):1030-1037. 9. JensenFV.AnintroductiontoBayesiannetworks. Vol210:UCLpressLondon;1996. SupportedinpartbytheNationalInstituteofHealth(NIH)–NIMHR03MH106922,NIA K25AG051782.ArianaEAnderson,Ph.D.,holdsaCareerAwardattheScientificInterface fromBWF. 3
© Copyright 2026 Paperzz