Advancing the understanding of brain function with multivariate pattern analysis [Elektronische Ressource] / von Michael Hanke

Sujets

• Psychologie
• Biologie

Informations

• Publié par: otto-von-guericke-universitat_magdeburg
• Publié le: 01 janvier 2009
• Nombre de lectures: 6
• Langue: English
• Poids de l'ouvrage: 13 Mo

Extrait

erAd2009vGabrieleancinghaeltheProf.UnderstrandingonoforenBrainfeldeFunctionowithtMulerteidigttivMagdeburgariah.teeP20.ainttehrnStefanAnalPDysiseingereicDissertation24.zur2009Erlangung25.desersit?takvademiscDipl.-PsychenMicGradesHankdogebctoramrerumOktobnaturalium1978(Dr.Leinerer.Gutacnat.)ter:genehmigtDr.durcPhllmanndieDr.FLohmannakult?thf?ram:NaturwissenscM?haftenzdervOtto-vam:on-GuericJunike-UnivHere,AbstracttheDecoofdingilypatternstoofthisneurallactivittycalledonotoealthciognitivareeshostates.isresponeopofoftheduced.cenlibrariestraltgoalsages.ofusage,functionalanbrainWhileimaging.oStandardonuniveariateanfMRIeminenanalysisbmethoeds,el,whicwhA,copatternrtsreeslateabilitcognitivlargeecomput-aterfacenhine-dispeerceptualyfunctionthiswithoftheAblo-oidvoprimaryxtecygeonwhicaabletMithison-levmoelfeaturedepofendentotresearc(BOLD)nosig-cross-platform,nal,sourcehaframewvMetheproultivvtecendatassuccessfulininAidenoftifying'anatomicaltoregionsinbasedarietonlanguagessignalenincreasestoduringthecoexistinggnitivpaceframewandtedpanderceptualontasks.andRecenprotlyIn,proresearcvhersstrategieshaofvneuroimagingearbpegunetoedexploreprnewouslymstudies,ultivcusariatesensitivitytecthathniquestothatinhaadditionalvareevpropartvtenA-basedtovbhniqueesuitedmorey-indepexible,analysis,moreisreliable,ananduniformmoreectsensi-cognitivtivneuroscienceeh.thanastandardvunivPython-based,ariateandanalysis.en-Drasoftwingareonork,thePyeldVPofforstatisticalapplicationlearningmtheoryariate,analysisthesehniquesnewfMRImeultivisariatetro-patternPyMVPanalysismak(MVPuseA)PythtecnhniquesspyossessaccessexplanatorywrittenpaovwyerprogrammingthatandcouldingprvironmenosvideinnewwithinsighwtsofinmactolearningthekfunctionalTheproporkertiespresenofinthethesis,brain.illustrativHoexampleswitsevfeatures,er,programmabilitunlikareevtheded.waddition,ealththesisofvidessoftowerviewarepromisingpacforkapplicationagesMVPfortounivdatasets.ariatevanal-iyses,ousthereossibilitiesarerfewvpacewkbasedagesnthatefacilitateimpubultivishedariatethepatternfoclassicationliesanalysestheofanalysisfMRIhniquedata.isThiswninprturnvideprevtenrestingtsinformationthehadoptionread-oafailtheseasmethoofdsybypicalyVPastudylargeMoreoner,umtecbisertlyofforresearcdalithendengroupsdatatoafullythatassessdemonstratedtheirypexampleotenatialiwithAanalysisofofthedatasetstialfromanalysisfourtidierenelltrbrainaimagingcedure,domains.tatiTheevthesisasconcludespwithinaterpretation.discussionasabstandardoutpro-theincludingcshallengessthatcalhaaluationvresults,ewtoasbotenepitfallsfacedtheitoinestablishiiMVPthanAbcplanet,knoticwledgemenotsoceAanddissertationaisfunaandbigdasteporineaforscientheticwcareeyrP,abandprotheypathokingto-comingwoutardsonderfulitthankisypandepponeeredowithu-cstraighhallenges.ISomeaareypurelyhiking,inhourstellectual,freesomeasidesaisrtimeeorkjustamlabhaotorertheless,ihasousonestasks.withoutEitherkonewcanherbveMangiexhaustivouteoutwhenr?ndfacedtroalone.orld:Fesidesortunatelyurkish,linearIehadoutmanMoreyucpcompaneopleBaumgartnerthatfromhelponedtoandalguidedcounmefruitfulonresearcthiswjourneyam.life.Firstthatofaallsix,actuallyItowhoursould,likleardtoethanktStefanforPears.ollpmannwhoforfriends.hisygenerouswandsad,unconditionalysuppwortouldfromburden.thetvMaertenserynbteginningalmost(actuallyAngeleveendiscussionsblifesteforeonesthat).heWithoutIngohisasattitudeorginallytomeriskthonsomething,!ander,reacEnglish,hofoutisfortnewwhicgroundathishelpproningjectrwfewouldtlyneveryeren-hadvofeandgotteniiiinthehoneshapmoeininariswhicRussianhbutitsisthenotlessw.ofHediscussionsnevouterh,seemedsofttoare,hathevusingeoflostAfterthevingfaithearthinindeedprogress,smalldespiteandsomehourstracesdierenceofhelpwhatoucouldwb24eainyterpretedIasreallyanooforwccasionaltoprovcrastinationhisofrighminenext(onlymineattherstysighNevtaofocourse).guyTonlyenvirtualyWith-earsmago,realYlifearoslaouldveHalandcmhenkwocolleaguesandorIwcouldbhaavIeanservtoedMarianneasforthefriemaindshipcencouragemenharactersoineratenlears,ighat-heartednmolliviecasualababoutItaliantyle,wtoughoabguystthatscienmettopics.onFthewintheternet.whoHinaducedvingtovPyerywsimilarThanksbMore-eliefsvabboutGerman,hoandwbititTshouldhebalsoeendoneinrrightalgebra,,handassharingtramendousslighinttetendencthingsymotoewaatimes.rrecends,unsolicitedvpreacmhing,hIjocaneonlythesayyRekthatDanielitFlorianwandasallgirlhopfacet.eFtheyitdon'tymwhoindthatmtoyKarolinefrequenandthaappteforarances.aIIwbouldecaforlsopanion,likinegtokidsthanklifemtic,yIfelloofwproPyMVPextra-freedom,AcessarydevanelopmersvPandereSederboscergeingandestEmansuppueleancOlivyettimore,forthaevandihnaglytheseessenonallycreallycasionalsuppdis-grandparencussionsathatusextendtmeyashorizonivwithinwmintutes,thankandyforelotheedjofriendyprospoftivwritingwifepapMakershwithbthem.mAlobngcom-thatunselshline,orter,Ithealsohorwmanlife.turther-toI'mthankladJimoHaxbvyourandCharlotteStevMoritz,eenricHansonourforwiththeircomplesustainedesuppnon-scienort,butandtialevAnden,moreamforgratefulthethememorableortptheirots,salwt-barbysecuevideddiscussions.withMostbiimpofortanwhenevtlyr,whonew.ever,.Con.ten.ts3.1.1.WAhigh-ldv.ancing3.1.2.thensComprehensi.onproofaBrain.F.unctiondissect1.1.1..Exploring.the.brain33v.o.xel.b.y.v.oCommxel..p...resp...hligh.More.v...a..3.1.2..Considering.brain-resp2.5.2.onse.p34atternsterface.......full-brain...eature.....dev7.1.3.3.Linkingsystemsbrain-respClassifyonses.to.cognitivveonenstates.with.MVP3.1.3.A.9.1.3.1.vTanalysisec.hnical.adv32anwtages..........Dataset.........eature.......Demonstrating9el1.3.2..Theoretical.adv2.6.1.ances........2.6.2........2.6.3..........37.y-driv11t1.3.3..Practical.adv.anStrategiestageseling.y.44.n.........ving.xel-space:.com-........13.1.4..Theariateneed.for.transparency..47.is.S.tivit.............2.5..orko.abstr.ction..14.1.5..Practical.diculties:.Where.is.the.soft.w2.5.1.aresamplingr.e.?..........3317F2.selectionTheceduresP.yMVP.A.F.ramew.ork2.6.21the2.1.evPython:inlingua.f.r.an.c35aLoadingofdatasetcomputational.(neuro)science.2.1.2.2..Bridging.fMRI.data36andSimplemacanalysishine.learning.soft.w.are..3623F2.3.selectionDataset.handling..............2.7..unit.en.elopmen.............40.Analysis.43.Lab25functional2.4.bMacpredictionhineerformancelearning3.1.1.alagorithmsd...............44.Lea.the.o.Brain.onse.p30ts2.4.1..Classier.abstr.a.ction............46.Multiv.searc.t..........30.2.4.2.3.2.Finformationeatureameasuresailable:.e.i.y...52..t
3.5.2.ten.tsP3.2.1.ofEv.e.ntiont-relatedeatureexample.dataset..97.Self-organizing.81.ject-pro.....-..54athom3.2.2..Spatio-tempeoralaMVPCurriculumA....oking.v.83.........ey......55Conclusions3.3.otenSensitivit.y.analysismethoas.ablueprimo.dalit5.y-indepvienden.t.tec.h-.niqueExample:.categorical.t.tral.stream.Statistical.......3.5.1..stabilit.......measures.accuracies.............4..4.1..full.al....64.3.3.1.94EEGw.for.e...Oering.t.e.....103.11.3.4.1..maps...............3.4.2..Lo.for.informa-65in3.3.2.heMEGen.ob.cessing...3.5..safeguarding.results...........85.F.selection.y...........88.Statistical69b3.3.3.ondExtracellularandConrecordingstests.......................9071General3.3.4.93fMRIF.the.p.ti.................4.2..o.erful.ds.cognitiv.neuroscienc.......4.3..the.n74of3.4.mind-rAadermetho.d.to.in99vReferencesestigateA.informationvitaerepresen5tations81t1.asAwdvtoancingexample,thetComprehensionpofOrban,BrainhighlighFandunctionandwithtoMultinformation-bivtheariate(VPoaMVPtternariateAninalysistempRecenferentlythey,Inneiurong,oGosDetre,c(MVPienmmtistsotell,haanvonlyeitsrepvortedofsurprisingacresultsaswherepresennvtheycortexappliedrepresenmaccategorieshineandlearningateischahniquesofbasedhniquesondingsestati(sBandettini,tie.g.cal2006).learningmultivariatetheoryeinds.theirdiameteanalysisTofhoups,fMRIconsequendataindividual1car-.smallFinformationorwn.example,studiestfwtheomstudiesappremploFyingAmusedultivestigateariaoftjectsetrpatternranalysessbthatyofHaobynesspatially&vReesealed(2005)vandrKamitanistructure&toTliterature,onge(2005)apwhineeregablneuralecto&predictHatheal.,ordiKriegeskenel,tationorofanalysisvisualPstimHaxbuthislgeneraliatternfromwillfMRItodatatheserecorded0.5iinnrhanduel,umanoprimaryScvisual&cortex.2002),Thesetlystuydivesxelaggregatedriesinformationaconamountaineddiscriminatinginonvoariable,OthersubtleA-basedresphaonseebiasesurtherintedlargestrengthnaumultivbanalysisersoofh.voroMVPxelswwhicrsthtowvouldneuralnottationsbfaceseobdetectedinbenyalunivoarialateandanalysis.hoTheedstheetationssmalldif-signaltbiasesjectwareeredistributedsucienoterlappingtorevdisamthatbiguatehastimeulussimilaorienitationsydespitethattherelatedfact1thatthetheirauthorsfMRIvdatareferredwtheeplicationremacreclearnordednattec3tommdataspatialderesolution.oThis(KamitaniisTesp2005;eciallynyetnotable2007),baseecauseanalysisthee.g.organizationorte,ofebthe&primary2006)visualmulti-voxelcortexatterni(nNorman,monkolyn,eys&indicatesy,thatThroughoutthethesisoriemorentermtation-selecptanalysisiA)vbeusedcolumnsreferareallonlymethoappro1ximatelyhstimNiculusoproplertieshapterandthesemanexibleticwithrelatioprontshipsnecessarily(Haxbtialyanalysisetwal.,v2001;ceHanson,neurosciMatsukthea,b&tlyHaxbadvy,the2004;aO'Thniquesoofole,areJiang,regression,Abe.g.di,and&ousHaxbdality,These2005).elopAlthoughythealuablerstneuralusetheiroftheirMVPiAanalysisongeneralneuroimagingindataanddatesarebacthiskerviewinfMRItoreferredthe&earlyultiv1990shniques.(Moypicallyelle.g.erp&fulStrother,ultiv1991;andKippapplicableenhahn,rBarkmoer,norPecicascal,iNagel,hniques,&ereDuara,outside1992,commwithcbanotherothhstudiesysisusingThisdatatofromotenpconositronimpemissionetomographwithyail-,ceduresPET),ointhelimitations,stages.eductionstrikingedevitedelopmenwhictsrelatedhaevFeojustinfererecennon-inferentlyysisattractedreaderconsiderablePinPterest(1999a,b).throughoutmtheariateneuroscienceteccommTheyunittygeneric,(see(Normanclassication,etclustering),al.,o2006;er-Ha(ynesm&ariate,Rees,inear2006,non-linear)foroftenreviews).toWhileaclassical,imdataulti