RuffExperimentalDesign认知神经科学实验设计.ppt

ExperimentalDesign ChristianRuffWiththanksto RikHensonDanielGlaser Realignment Smoothing Normalisation Generallinearmodel Statisticalparametricmap SPM Imagetime series Parameterestimates Designmatrix Template Kernel Gaussianfieldtheory p 0 05 Statisticalinference Overview CategoricaldesignsSubtraction Pureinsertion evoked differentialresponsesConjunction TestingmultiplehypothesesParametricdesignsLinear Adaptation cognitivedimensionsNonlinear Polynomialexpansions neurometricfunctionsFactorialdesignsCategorical InteractionsandpureinsertionParametric Linearandnonlinearinteractions PsychophysiologicalInteractions Aim BrainstructuresunderlyingprocessP Procedure Contrast TaskwithP controltaskwithoutP Pthecriticalassumptionof pureinsertion CognitiveSubtraction CognitiveSubtraction Baseline problems Severalcomponentsdiffer Distant stimuli Differentialevent relatedfMRI Inferotemporalresponsetofaces SPM F testingforevokedresponses Evokedresponses Rest baseline Baseline herecorrespondstosessionmean andthusprocessingduring rest NulleventsorlongSOAsessentialforestimation Cognitive interpretationhardlypossible butusefultodefineregionsgenerallyinvolvedinthetask ExperimentaldesignWordgenerationGWordrepetitionRRGRGRGRGRGRG G R Intrinsicwordgeneration underassumptionofpureinsertion Acategoricalanalysis Overview CategoricaldesignsSubtraction Pureinsertion evoked differentialresponsesConjunction TestingmultiplehypothesesParametricdesignsLinear Adaptation cognitivedimensionsNonlinear Polynomialexpansions neurometricfunctionsFactorialdesignsCategorical InteractionsandpureinsertionParametric Linearandnonlinearinteractions PsychophysiologicalInteractions Onewaytominimisethebaseline pureinsertionproblemistoisolatethesameprocessbytwoormoreseparatecomparisons andinspecttheresultingsimpleeffectsforcommonalitiesAtestforsuchactivationcommontoseveralindependentcontrastsiscalled Conjunction Conjunctionscanbeconductedacrossawholevarietyofdifferentcontexts tasksstimulisenses vision audition etc Butthecontrastsenteringaconjunctionhavetobetrulyindependent Conjunctions Example Whichneuralstructuressupportobjectrecognition independentoftask namingvsviewing VisualProcessingVObjectRecognitionRPhonologicalRetrievalP Object Colourviewing seelater Conjunctions Priceetal 1997 Commonobjectrecognitionresponse R Conjunctions SPM8offerstwogeneralwaystotestthesignificanceofconjunctions Testofglobalnullhypothesis Significantsetofconsistenteffects whichvoxelsshoweffectsofsimilardirection butnotnecessarilyindividualsignificance acrosscontrasts Testofconjunctionnullhypothesis Setofconsistentlysignificanteffects whichvoxelsshow foreachspecifiedcontrast effects threshold Choiceoftestdependsonhypothesisandcongruenceofcontrasts theglobalnulltestismoresensitive i e whendirectionofeffectshypothesised Twoflavoursofinferenceaboutconjunctions Fristonetal 2005 Neuroimage 25 661 7 Nicholsetal 2005 Neuroimage 25 653 60 Overview CategoricaldesignsSubtraction Pureinsertion evoked differentialresponsesConjunction TestingmultiplehypothesesParametricdesignsLinear Adaptation cognitivedimensionsNonlinear Polynomialexpansions neurometricfunctionsFactorialdesignsCategorical InteractionsandpureinsertionParametric Linearandnonlinearinteractions PsychophysiologicalInteractions ParametricDesigns GeneralApproach Parametricdesignsapproachthebaselineproblemby Varyingastimulus parameterofinterestonacontinuum inmultiple n 2 steps andrelatingblood flowtothisparameterFlexiblechoiceoftestsforsuchrelations LinearNonlinear Quadratic cubic etc Data driven e g neurometricfunctions Model based Adaptation Lineareffectsoftime Alinearparametriccontrast Adaptation Nonlineareffectsoftime Anonlinearparametriccontrast Inverted U responsetoincreasingwordpresentationrateintheDLPFC SPM F Polynomialexpansion f x b1x b2x2 upto N 1 thorderforNlevels E g F contrast 010 onQuadraticParameter Nonlinearparametricdesignmatrix SPM8GUIofferspolynomialexpansionasoptionduringcreationofparametricmodulationregressors Linearchange Quadraticchange Meanresponse ParametricDesigns Neurometricfunctions Rees G etal 1997 Neuroimage 6 27 78 versus Inverted U responsetoincreasingwordpresentationrateintheDLPFC Rees G etal 1997 Neuroimage 6 27 78 ParametricDesigns Neurometricfunctions CodingoftactilestimuliinAnteriorCingulateCortex Stimulus a presence b intensity and c painintensityVariationofintensityofaheatstimulusappliedtotherighthand 300 400 500 and600mJ B cheletal 2002 TheJournalofNeuroscience 22 970 6 Assumptions ParametricDesigns Neurometricfunctions B cheletal 2002 TheJournalofNeuroscience 22 970 6 Stimuluspresence Painintensity Stimulusintensity ParametricDesigns Model basedregressors Seymour O Doherty etal 2004 Nature Overview CategoricaldesignsSubtraction Pureinsertion evoked differentialresponsesConjunction TestingmultiplehypothesesParametricdesignsLinear Adaptation cognitivedimensionsNonlinear Polynomialexpansions neurometricfunctionsFactorialdesignsCategorical InteractionsandpureinsertionParametric Linearandnonlinearinteractions PsychophysiologicalInteractions ColoursObjectsColoursObjects interactioneffect TaskxStimuli Viewing Naming Factorialdesigns MaineffectsandInteractions Maineffectoftask A1 B1 A2 B2 Maineffectofstimuli A1 A2 B1 B2 Interactionoftaskandstimuli Canshowafailureofpureinsertion A1 B1 A2 B2 Interactionsandpureinsertion Object naming specificactivations Context nonamingnaming adjustedrCBF ComponentsVisualprocessingVObjectrecognitionRPhonologicalretrievalPInteractionRxPInteraction nameobject colour viewobject colour 1 100 001 1 P R V RxP P V R V V RxP Interactionsandpureinsertion Interactions simpleandcross overWecanselectivelyinspectourdataforoneortheotherbymaskingduringinference A1A2B1B2 A1A2B1B2 Overview CategoricaldesignsSubtraction Pureinsertion evoked differentialresponsesConjunction TestingmultiplehypothesesParametricdesignsLinear Adaptation cognitivedimensionsNonlinear Polynomialexpansions neurometricfunctionsFactorialdesignsCategorical InteractionsandpureinsertionParametric Linearandnonlinearinteractions PsychophysiologicalInteractions AlinearTime by ConditionInteraction differenceinadaptationforrepeatingvsgenerating Contrast 531 1 3 5 11 55 33 111 13 35 5 LinearParametricInteraction FactorialDesignwith2factors Gen Rep Categorical 2levels Time Parametric 6levels Timeeffectsmodelledwithbothlinearandquadraticcomponents NonlinearParametricInteraction Overview CategoricaldesignsSubtraction Pureinsertion evoked differentialresponsesConjunction TestingmultiplehypothesesParametricdesignsLinear Adaptation cognitivedimensionsNonlinear Polynomialexpansions neurometricfunctionsFactorialdesignsCategorical InteractionsandpureinsertionParametric Linearandnonlinearinteractions PsychophysiologicalInteractions Parametric factorialdesign inwhichonefactorisapsychologicalcontext andtheotherisaphysiologicalsource activityextractedfromabrainregionofinterest Psycho physiologicalInteraction PPI Parametric factorialdesign inwhichonefactorisapsychologicalcontext andtheotherisaphysiologicalsource activityextractedfromabrainregionofinterest Psycho physiologicalInteraction PPI AttentionalmodulationofV1 V5contribution Psycho physiologicalInteraction PPI 001 SPM Z AttentionalmodulationofV1 V5contribution attention noattention V1activity V5activity time V1activity Psycho physiologicalInteraction PPI Stimuli Facesorobjects PPC IT Psycho physiologicalInteraction PPI 001 adjustedrCBF medialparietalactivity Faces Objects Stimuli Facesorobjects PPC IT SPM Z Psycho physiologicalInteraction PPI PPIstestedbyaGLMwithform y V1 A b1 V1 b2 A b3 ec 100 However theinteractiontermofinterest V1 A istheproductofV1activityandAttentionblockAFTERconvolutionwithHRFWearereallyinterestedininteractionatneurallevel but HRF V1 HRF A HRF V1 A unlessAlowfrequency e g blocked mainlyproblemforevent relatedPPIs SPM8caneffectadeconvolutionofphysiologicalregressors V1 beforecalculatinginteractiontermandreconvolvingwiththeHRF the PPIbutton Psycho physiologicalInteraction PPI Overview CategoricaldesignsSubtraction Pureinsertion evoked differentialresponsesConjunction TestingmultiplehypothesesParametricdesignsLinear Adaptation cognitivedimensionsNonlinear Polynomialexpansions neurometricfunctionsFactorialdesignsCategorical InteractionsandpureinsertionParametric Linearandnonlinearinteractions PsychophysiologicalInteractions