TITLE:IdentificationoftheventraloccipitalvisualfieldmapsinthehumanbrainAUTHORS:Winawer,JonathanPsychologyandCenterforNeuralScienceNewYorkUniversityNewYork,NY,USAjonathan.winawer@nyu.eduWitthoft,NathanPsychologyStanfordUniversityStanford,CA,USAwitthoft@stanford.eduCORRESPONDINGAUTHOR:JonathanWinawerKEYWORDS: visualcortex;functionalmagneticresonanceimaging;retinotopicmap;visualfieldmap;HV4;populationreceptivefield;collateralsulcus;fusiformgyrus;occipitallobe.ABSTRACT:Thelocationandtopographyofthefirstthreevisualfieldmapsinthehumanbrain,V1-V3,arewellagreeduponand routinelymeasuredacrossmost laboratories.Thepositionof4thvisualfield map, ‘hV4’, is identified with less consistency in the neuroimaging literature. Usingmagnetic resonance imaging (MRI) data,wedescribe landmarks tohelp identify thepositionandbordersof ‘hV4’.Thedataconsistofanatomical images, visualizedas corticalmeshes tohighlightthesulcalandgyralpatterns,andfunctionaldataobtainedfromretinotopicmappingexperiments,visualizedaseccentricityandanglemapsonthecorticalsurface.Severalfeaturesofthefunctionalandanatomicaldatacanbefoundacrossnearlyallsubjectsandarehelpful for identifying the locationandextentof thehV4map.ThemedialborderofhV4 is shared with the posterior, ventral portion of V3, and is marked by a retinotopicrepresentationof theupper verticalmeridian. Theanteriorborderof hV4 is sharedwith theVO-1 map, and falls on a retinotopic representation of the peripheral visual field, usuallycoincidentwith theposterior transversecollateral sulcus.Theventro-lateraledgeof themaptypically fallson the inferioroccipitalgyrus,where functionalMRIartifactsoftenobscure theretinotopic data. Finally, we demonstrate the continuity of retinotopic parameters betweenhV4anditsneighbors;hV4andV3vcontainiso-eccentricitylinesinregister,whereashV4andVO-1containiso-polaranglelinesinregister.Together, themultipleconstraintsallowforaconsistent identificationof thehV4mapacrossmosthumansubjects.

.CC-BY-NC-ND 4.0 International licensecertified by peer review) is the author/funder. It is made available under aThe copyright holder for this preprint (which was notthis version posted January 5, 2016. . https://doi.org/10.1101/035980doi: bioRxiv preprint

INTRODUCTION:Thehumanbraincontainswelloveradozenvisualfieldmaps1-3. Identificationofthesemapshasbeenamajorsuccessinthehistoryofvisualneuroscience.Becauseresearcherscanidentifythe same brain region across multiple measurements and diverse populations, scientificfindingscanbeaggregatedacrossstudies toarriveatabetterunderstandingofhumanbrainfunction. The success of such aggregation, however, is limited by the accuracywithwhich agivenregioncanbeidentifiedacrossindividuals.Thereislittletonodoubtaboutthepositionandbordersofseveralmaps.Primaryvisualcortex(V1)alwaysliesontheCalcarinesulcus4-7,anditsborderscanbeidentifiedbasedondatafromasinglefMRIscanningsessionwithhighprecision 8,9. V2 andV3 can alsobe identifiedquite accurately and routinely, and in fact theretinotopyparametersofallthreemapscanbereasonablyestimatedfromtheanatomyalone10,11.In contrast, the fourth visual field map has proven more difficult to characterize, withconsiderably less consistency in map definition across laboratories, compared to V1-V312-19.Thereareseveralreasons.ComparedtoV1-V3,inhV4thefMRIsignalsarelessreliablydrivenby simple contrast patterns20, the homology with animal models is less certain21, imagingartifacts canaffect somepartsof themap inmany subjects13, andanatomical landmarksarelessfrequentlyusedinmapdelineation.Recentwork,however,hasexaminedthehV4mapinsome detail13,18, suggesting that one can identify hV4 and its neighbors (Figure 1) withreasonablygoodconsistencyacrossindividuals.In our view, successful map identification of the hV4 maps is aided by combining severalsourcesofdata.First,weconsidertheanatomy,especiallythepatternofsulciandgyri.Secondweusetheanglemapandeccentricitymapderivedfromretinotopymeasurements.Third,weuseametricofsignalqualitysuchasmeanBOLDsignal(orsignaltonoiseratio,orthequalityofthe retinotopic fit) to identify possible imperfections in the measurements that should beignored in identifyingthemap. Inthesimplestscenarios,allconstraintsagreeandthemapiseasilyidentifiable.Insomecases,themeasurementsdonotfullyagreebutwecannonethelessidentifythegenerallocationofhV4basedonthemutualconstraints,andatleastoneormoreofthemapboundaries.TheprotocolforidentifyingthehV4retinotopicmapisdescribedinthefollowingtextandinastep-by-stepvideodescription(SupplementaryMovie1).

.CC-BY-NC-ND 4.0 International licensecertified by peer review) is the author/funder. It is made available under aThe copyright holder for this preprint (which was notthis version posted January 5, 2016. . https://doi.org/10.1101/035980doi: bioRxiv preprint

Figure1.Locationofventraloccipitalvisualfieldmaps.Thelocationsof6medialandventralvisualfieldmapsareshownonarenderingofasubject’srighthemisphere.Themeshunderlayisaslightlysmoothedrenderingofthecorticalsurface.Sulciareindicatedbydarkgrayandgyribylightgray.Adaptedfrom1.

PROTOCOL:Our procedure for identifying visual field maps involves two kinds of MRI data and severalprocessingsteps.Therequireddataareananatomical imageofthesubject’sbrain(structuralMRI) and functional data from a retinotopic mapping experiment. Typically, all data can becollectedinlessthanonehourofscantimeona3Teslascannerusingastandardgradientechopulsesequence for functionaldataandT1-weighted images foranatomicaldata.Theanalysissteps include segmenting gray andwhitematter to generate a corticalmesh, identifying thesulcal and gyral patterns in ventral occipito-temporal cortex, extracting the eccentricity andangledatafromtheretinotopicexperiments,andthenusingthesemeasurementstoguidethemarkingofvisualfieldboundaries.Thereisconsiderablevariabilityinthespecificproceduresusedbydifferentresearchgroupsforacquisition of functional and anatomical MRI data and for the derivation of retinotopyparametersfromthefMRIscans.Anyoftheretinotopymethodsincommonusageissuitable;

.CC-BY-NC-ND 4.0 International licensecertified by peer review) is the author/funder. It is made available under aThe copyright holder for this preprint (which was notthis version posted January 5, 2016. . https://doi.org/10.1101/035980doi: bioRxiv preprint

theonlyrequirementsarethatonecanvisualizeeccentricityandangledataonacorticalmesh.The methods to produce the images shown in this paper are described in more detailelsewhere18. We summarize thesemethods briefly and then focus on the specific steps foridentifyingthehV4mapanditsneighbors.1.Dataacquisition1.1) Anatomical data. Anatomical data are needed in order to render retinotopic data onsurface representations.For subjects1-3depicted in thevideopresentation,2-4wholebrainSPGRT1-weightedMRIimageswereacquiredona3TGEscannerat1mmisotopicresolutionusing an 8-channel whole-brain coil. Themultiple images for each subject were aligned andaveraged.Averagingmultipleanatomicalimagesisdesirableasitincreasesthecontrastoftheboundary between the grey and white matter and therefore aids segmentation and thecreationofanaccuratecorticalsurface.1.2)Functionaldata.FunctionaldatawereacquiredasT2*-weightedgradientechoimageswitheitheraspiral22orrectilinear(EPI)trajectorythroughk-space,with2.5mmisotropicvoxelsandaTRofeither1.5sor2s.(Fordetailsseeref18.)2.Stimuli.Retinotopy was measured with stimuli designed either for travelling wave analysis23 orpopulation receptive modeling24. For traveling wave analysis, alternate scans containedcontrastpatternsviewedthroughexpandingringaperturesorclockwiserotatingwedgeswithavisualfieldextentof14degreesfromfixation.(Fordetailsseeref18.)Forpopulationreceptivefieldmodeling,stimuliwerecontrastpatternsviewedthoughmovingbaraperturesthatslowlysweptacrossthevisualfield,withamaximumeccentricityof6degreesfromfixation.3.FunctionalMRIanalysis3.1)Preprocessing. Several standard fMRIpreprocessing stepswereusedprior to retinotopicanalysis, including slice timingcorrection,motioncorrection,andhighpass temporal filteringwitha1/20Hzcutoff.3.2)Travellingwaveanalysis.Datafromtravelingwavescans(wedgesandrings)wereanalyzedvoxelbyvoxelbyapplyingtheFouriertransformtothetimeseries.Thephaseatthestimulusfrequencymeasuresthepolarangleoreccentricityofthestimulusthatmosteffectivelydrivesthe BOLD response in that voxel. The coherence of the signal at the stimulus frequencyindicatesthegoodnessoffitoftheresponsetotheperiodicdesign.Coherenceisdefinedasthepoweratthestimulusfrequencydividedbythepoweracrossallfrequencies.Fordetailssee25.3.4)Populationreceptivefields.ApopulationReceptiveField(pRF)modelwassolvedforeachvoxelusingmethodsdescribedpreviously24. Inbrief,eachpRFwasdefinedasa2-D isotropicGaussian,parameterizedbyitscenter(x,y),size(sdofGaussian),andamplitude(scalefactor).ThemodelpredictionisthedotproductofthepRFandthestimulusaperture,convolvedwithahemodynamic response function. Themodelwas fit throughaminimizationof least squarederrorbetweenthepredictedandobservedtimeseriesusingacoarse-to-fineapproach.ThepRF

.CC-BY-NC-ND 4.0 International licensecertified by peer review) is the author/funder. It is made available under aThe copyright holder for this preprint (which was notthis version posted January 5, 2016. . https://doi.org/10.1101/035980doi: bioRxiv preprint

centercanbeconvertedtopolarcoordinatestoyieldanangleandeccentricitymap,similartothat obtained from traveling wave analysis. For the purposes of retinotopicmapping in thispaper,otherpRFparameterswerenotused(sizeandamplitude).4.SegmentationandvisualizationBecause the retinotopic maps in visual cortex are organized on the 2D surface, not the 3Dvolume, map data is best visualized on a surface representation, typically rendered as aninflated(smoothed)orun-inflated3Dsurfacemesh,orasaflattened(2D)surface26-28.Thedatausedforthesurfacerenderingisusuallytheborderofthegrayandwhitematter,derivedfroma segmentation procedure that labels anatomical voxels from the T1 image as gray orwhitematter,andfindsthenfindstheboundarythatseparatesthetwotissues.Forthedatashowninthis paper, we segment the gray and white matter using the Freesurfer auto-segmentationtools 29 followed by manual correction using ITK-GRAY software(http://white.stanford.edu/software/, modified from ITK-SNAP http://www.itksnap.org)30.Functional data were then aligned to the whole brain anatomical data and rendered on asmoothed surface mesh using custom software (“vistasoft”;http://vistalab.stanford.edu/software/;https://github.com/vistalab/vistasoft).5.AnatomicallandmarksWefind ithelpful to identify several sulci andgyribeforedrawing the retinotopicmaps,andfollowthenamingconventionsusedbyDuvernoy31.ThesulciandgyrilabeledinFigure2canallbe seen in post-mortem brains in Duvernoy‘s text (e.g., his figure 17). Their locations aredescribedbrieflybelow.

.CC-BY-NC-ND 4.0 International licensecertified by peer review) is the author/funder. It is made available under aThe copyright holder for this preprint (which was notthis version posted January 5, 2016. . https://doi.org/10.1101/035980doi: bioRxiv preprint

Figure 2. Anatomical landmarks in ventral occipitotemporal cortex. Several sulci and gyri are labeled on theventralandmedialsurfaceofahighlysmoothedrighthemisphere.Thesesulciandgyriprovideusefullandmarksintheidentificationofthevisualfieldmaps.Calc:Calcarinesulcus;Ling:Lingualsulcus/gyrus;CoS:Collateralsulcus;Fus:Fusiformgyrus;ptCoS:Posteriortransversecollateralsulcus;IOG:Inferioroccipitalgyrus.5.1) Calcarine sulcus. The most useful landmark is the calcarine sulcus (sometimes calledcalcarine fissure), which locates primary visual cortex (V1). It is the large sulcus in medialoccipitalcortexseparatingthelingualgyrusfromthecuneus.Itcanbefoundunambiguouslyinevery subjectand isalways the locationofV1. It runsonananterior-posterioraxis,with theanteriorendmarkedby theparietaloccipital fissure,and theposteriorendapproximatelyattheoccipitalpole,thoughthereisvariability intheposteriorextent,withthesulcuswrappingaroundtothelateralsurfaceinsomesubjects,andterminatingontheinferiorsurfaceinothers32,33.5.2)Lingualgyrusandlingualsulcus.ThelingualgyrusisinferiortotheCalcarinesulcusandisthelocationoftheventralportionofV2andV3.Itsinferiorboundaryisthecollateralsulcus.Itstretches approximates from the posterior pole to the parahippocampal gyrus. The lingualgyruscontainsoneormoresulcitogetherconsideredthelingualsulcus.5.3)Collateralsulcus.Thecollateralsulcusseparatesthelingualandparahippocampalgyruson

.CC-BY-NC-ND 4.0 International licensecertified by peer review) is the author/funder. It is made available under aThe copyright holder for this preprint (which was notthis version posted January 5, 2016. . https://doi.org/10.1101/035980doi: bioRxiv preprint

theone side from the fusiformgyruson theother. Its posterior end is usually denotedby atransverseportionof the sulcus, called theposterior transversecollateral sulcus (ptCoS).TheptCoSisanimportantlandmarkbecauseitusuallymarksthedivisionbetweenthehV4mapandtheVO-1map18.5.4)Fusiformgyrus. The fusiform gyrus is inferior to the lingual and parahippocampal gyrus,and is bounded medially by the collateral sulcus. The foveal representation of the ventral-occipitalmaps(VO-1/2)liesintheposteriorportionofthefusiformgyrus.5.5)Inferioroccipitalgyrus.Theinferioroccipitalgyrusliesposteriortothefusiformgyrusandcollateralsulcusand inferiortothe lingualgyrus. It isgenerallyseparatedfromtheposteriorportionoftheposteriorfusiformgyrusbytheposteriortransversecollateralsulcus.Thelateralportion of the hV4 map often lies here. This gyrus is typically in close proximity to thetransversevenoussinus,whichcanleadtofMRIartifactsthatobscurethisportionofthehV4map13.6.Drawingthemaps6.1)V1-V3maps.Identificationofthefirstthreevisualfieldmapsinhumanswasoneofthefirstaccomplishments in visual neuroscience following the advent of fMRI23,25,34. Delineation ofthesemapboundariesisnowroutine,althoughtracingthemapsthroughthefovealconfluencecanbedifficultintheabsenceofmethodsdevelopedspecificallyforthispurpose9.Inbrief,theV1 map is centered on the calcarine sulcus, and its borders with V2 lie on retinotopicrepresentationsoftheverticalmeridian.TheperipheralboundaryisusuallynotidentifiedwithfunctionalMRIbecausethefieldofviewofthatcanbeachievedduringscanningislessthanthefieldofviewrepresentedinthemaps.Hencetheperipheralboundaryofaregionofinterestisusuallychosenasthemostanteriorregionofthemapswherethesignalisgood,accordingtoathreshold incoherence (travelingwaveanalysis)orvarianceexplained (pRFanalysis). If thereare clear reversals in the angle map all the way back to the occipital pole, then it may bepossible to trace V2 and V3 in concentric ‘V’ shapes around V1; otherwise themost fovealportionofthemaps is typically labeledasanon-specific ‘fovealconfluence’,andeachmap istracedfarintothefoveaastheresolutionintheanglemapsallow,usuallyoneortwodegreesfromfixation.TheV2/V3bordersaremarkedbyarepresentationofthehorizontalmeridianintheanglemap.6.2)ThehV4map.ThehV4mapliesontheventralsurfaceoftheoccipitallobeandbordersatleasttwoothervisualfieldmaps,theventralpartofV3(V3v)andVO-1.6.2.1.) The anterior boundary: hV4 / VO-1 The boundary between hV4 and VO-1 can beidentified by both anatomy and retinotopic data. The retinotopic feature that defines thisborderisareversalintheeccentricitymap(Figure3).ThisperipheraleccentricitybanddividinghV4fromVO-1usuallycoincideswithaparticularanatomicalfeature,theposteriortransversecollateralsulcus(ptCoS)18.

.CC-BY-NC-ND 4.0 International licensecertified by peer review) is the author/funder. It is made available under aThe copyright holder for this preprint (which was notthis version posted January 5, 2016. . https://doi.org/10.1101/035980doi: bioRxiv preprint

Figure3.Eccentricitymap.Thecoloroverlayonthesmoothedcorticalmeshinthemainpanelshowsthestimuluseccentricity thatmost effectively drives each cortical location. The eccentricitywas derived from solving a pRFmodel, and the stimulus extent was limited to 6 degrees. The asterisk indicates the confluent fovealrepresentation.Blackandwhitelinesmarktheboundariesbetweenvisualfieldmaps.ThewhitelinedivideshV4from VO-1 and is the only boundary line in this figure which is derivable from the eccentricity map. This linecoincideswiththeptCoS(inset)andwiththeeccentricityreversal(blueincoloroverlay).Hencethewhitelinealsodividestheventraloccipitalmaps intotwoclusters,onewhich includesV1-hV4,andonewhich includesVO-1/2.DataarelimitedtovoxelsinwhichthepRFpredictionaccountedforatleast10%ofthevarianceexplainedinthetimeseriesandtoaposteriormaskthatincludesthe6labeledvisualfieldmaps:V1,V2andV3ventral,hV4,VO-1,andVO-2.Theinsetisthesamecorticalmeshwithlabelsindicatingthemajorsulcalandgyralpatterns,asinfigure2.6.2.2.) Themedial boundary: hV4/V3vOne of the borders of hV4 is sharedwith the ventralportionofV3 (V3v).Thisborder isdefinedbyanupperverticalmeridianpolarangle reversal(Figure 4). Because the V1-V3 maps are typically well defined by the retinotopic data, it isusuallyalsoclearwherethisboundaryisfound.

.CC-BY-NC-ND 4.0 International licensecertified by peer review) is the author/funder. It is made available under aThe copyright holder for this preprint (which was notthis version posted January 5, 2016. . https://doi.org/10.1101/035980doi: bioRxiv preprint

Figure 4. Angle map. The color overlay in the main panel indicates the angle in the visual field that mosteffectivelydrives responses ineach cortical location. Thevisual fieldmapboundaries inwhite canbe identifiedfromtheanglemap.Otherwiseasfigure3.6.2.3.) The ventral/lateral boundary The ventral/lateral boundary of hV4 is more difficult toidentifythantheotherboundariesbecausethereisnounambiguousfeatureoftheretinotopicdatatodefineit.Asdescribedin6.2.1and6.2.2,weknowwhatisontheothersideoftwoofthehV4boundaries,makingthoseboundarieswelldefined:TheanteriorboundaryisdefinedbyaneccentricityreversalandissharedwithVO-1,andthemedialboundaryisdefinedbyapolarangle reversal and is shared with V3v. In contrast, there is not a well-established map orretinotopicfeatureabuttingtheventralsideofhV4.Furthermore,thereisoftensignaldropoutin the fMRI measure on the ventral / lateral aspect of the hV4 map due to the transversevenoussinus,whichtypicallyrunsneartheinferioroccipitalgyrus13.Insomecases,thesignaldropout defines themost ventral/lateral extent of themap that can be identified (Figure 5;Figure6).

.CC-BY-NC-ND 4.0 International licensecertified by peer review) is the author/funder. It is made available under aThe copyright holder for this preprint (which was notthis version posted January 5, 2016. . https://doi.org/10.1101/035980doi: bioRxiv preprint

Figure5.Varianceexplainedmap.Thecoloroverlayinthemainpanelindicatesvarianceinthevoxeltimeseriesexplainedby thepRFmodelpredictions.Otherwiseas figure3. Some locationswith lowvarianceexplainedarelikelydue to fMRI artifacts, suchas the region indicatedby thewhite line,where the lowvarianceexplained iscaused by dropout from the transverse sinus. Other regionswith low variance explainedmay have visual fieldrepresentationsoutsidethestimulusextent,suchastheanterioredgeofV1,V2,andV3.

.CC-BY-NC-ND 4.0 International licensecertified by peer review) is the author/funder. It is made available under aThe copyright holder for this preprint (which was notthis version posted January 5, 2016. . https://doi.org/10.1101/035980doi: bioRxiv preprint

Figure6.Meansignalmap.ThecoloroverlayinthemainpanelindicatesthemeanfMRIsignal.Thesignalisnotuniformacrosscortex.Certain regionshave lowmeansignalduescanningartifacts.The region indicatedby thewhitelineliesnearthetransversevenoussinuswhichcausessignaldropout.Retinotopicdatafromtheselocationsmustbeinterpretedwithcaution.Otherwiseasfigure3.6.2.4.) Iso-eccentricity linessharedbyhV4andV3v Thegeneralorganizationof thehV4mapand its neighbors can be understood by tracing iso-eccentricity and iso-angle lines. The hV4eccentricity map is in register with the V1-V3 maps, so that iso-eccentricity lines span thehV4/V3vborder(Figure7).UnliketheV1-V3maps,thehV4mapappearstocontainlittletonorepresentationofthefarperiphery.Alongtheposterior-to-anterioraxis,hV4isthereforemuchshorterthanV1-V3,andthemostanterioreccentricitybandinhV4crossedV3vnearthemiddleofthelengthofV3v.

.CC-BY-NC-ND 4.0 International licensecertified by peer review) is the author/funder. It is made available under aThe copyright holder for this preprint (which was notthis version posted January 5, 2016. . https://doi.org/10.1101/035980doi: bioRxiv preprint

Figure 7. Iso-eccentricity lines. Iso-eccentricity lines in visual cortex are continuouswithin clusters, such as theposteriorclustercontainingV1,V2,V3,andhV4,andtheventraloccipitalclustercontainingtheVO-1/2maps.Theeccentricityoverlayandthemapboundariesareidenticaltothoseinfigure3.6.2.5.) Iso-angle linessharedbyhV4andVO-1The iso-angle lines inhV4arecontinuouswiththeVO-1mapsandnottheV3vmap.ThelinesusuallybendalongthehV4/V0-1border,withthe lower meridian representation being the most ventral and the shortest iso-angle line(Figure8).

.CC-BY-NC-ND 4.0 International licensecertified by peer review) is the author/funder. It is made available under aThe copyright holder for this preprint (which was notthis version posted January 5, 2016. . https://doi.org/10.1101/035980doi: bioRxiv preprint

Figure8.Iso-anglelines.Theiso-anglelinesinhV4andVO-1,indicatedbywhitelines,arecontinuousacrossthetwomapclusters.Otherwiseasfigure4.

RESULTS:The sulci and gyri associated on the ventral occipital cortical surface from a representativesubject(S1)areshowninFigure2.Themostusefulanatomicallandmarksforlocatingthevisualfield maps are the calcarine sulcus (V1); the lingual gyrus and lingual sulcus (V2v/V3v); theposteriorcollateralsulcusandinferioroccipitalgyrus,whichboundhV4;andthefusiformgyrusandcollateralsulcus,wheretheVOmapsarefound.TheeccentricitymapforsubjectS1showsthelarge-scaleorganizationofthemaps(Figure3).ThekeyfeatureontheeccentricitymapforidentifyingthehV4istheperipheralrepresentationontheptCoS,whichmarksthehV4/VO-1boundary.Theanglemap(Figure4)isusedtodefinetheV3v/hV4boundaryaswellastheVO-1/VO-2boundary.FMRIsignalqualityisnotuniformacrossthecorticalsurface.Somelocationshavepoorsignalduetoknownmeasurementartifacts,suchasthosewhichariseinregionsnearlargesinuses.In

.CC-BY-NC-ND 4.0 International licensecertified by peer review) is the author/funder. It is made available under aThe copyright holder for this preprint (which was notthis version posted January 5, 2016. . https://doi.org/10.1101/035980doi: bioRxiv preprint

subjectS1,thetransversevenoussinuscorruptsthefMRIsignalontheinferioroccipitalgyrus,resultinginlowvarianceexplainedbythepRFmodel(Figure5)andsignaldropout(Figure6)inthisregion.Iso-eccentricityand iso-angle lines in subject S1 clarify the internal structureof thehV4mapand the relationshipbetweenhV4and its neighbors. The iso-eccentricity lines are in registeracross hV4 and V1-V3 (Figure 7). The iso-angle lines are continuous across the hV4 / VO-1boundary(Figure8).AnglemapsandeccentricitymapsfromadditionalsubjectsshowsimilarpatternstoS1.TheV1-V3,hV4andVO-1/2mapsareshownforarepresentativerighthemisphere(S2;Figure9;Figure10)andlefthemisphere(S3;Figure11;Figure12).Thelargescaleorganizationissimilarforallsubjects. For example, there is an eccentricity reversal dividing hV4 and VO-1 and an anglereversaldividinghV4andV3.Inallsubjects,theeccentricityreversalfallsonorneartheptCoS.Somedetailsdifferbetweensubjects.Forexample,thefovealrepresentationinhV4isclearinS1andS2butnotS3(Figure11), likelyduetocorruptedsignal fromdrainingveins.Aseconddifferenceacrosssubjects isthattheuppermeridiananglereversaldividingVO-1andVO-2 isclear in S1 and S3 but not S2 (Figure 10). Nonetheless, there is sufficient regularity acrosssubjectstoidentifytheprincipalfeaturesdefiningthehV4mapanditsneighbors,V3vandVO-1(Figure13).

.CC-BY-NC-ND 4.0 International licensecertified by peer review) is the author/funder. It is made available under aThe copyright holder for this preprint (which was notthis version posted January 5, 2016. . https://doi.org/10.1101/035980doi: bioRxiv preprint

Figure9:Eccentricitymap,S2.Aneccentricitymapisshownonthepartiallyinflatedcorticalsurfaceofsubject2’sright hemisphere. The stimulus extentwas 14 deg and the data are from travelingwave analysis of expandingrings.Otherwiseasfigure3.

.CC-BY-NC-ND 4.0 International licensecertified by peer review) is the author/funder. It is made available under aThe copyright holder for this preprint (which was notthis version posted January 5, 2016. . https://doi.org/10.1101/035980doi: bioRxiv preprint

Figure10:Anglemap,S2.Ananglemapisshownforsubject2’srighthemisphere.Otherwiseasfigure9andfigure3.

.CC-BY-NC-ND 4.0 International licensecertified by peer review) is the author/funder. It is made available under aThe copyright holder for this preprint (which was notthis version posted January 5, 2016. . https://doi.org/10.1101/035980doi: bioRxiv preprint

Figure11:Eccentricitymap,S3.AnexampleofalefthemisphereeccentricitymapderivedfrompRFmodelfitting.OtherwiseasFigure3.

.CC-BY-NC-ND 4.0 International licensecertified by peer review) is the author/funder. It is made available under aThe copyright holder for this preprint (which was notthis version posted January 5, 2016. . https://doi.org/10.1101/035980doi: bioRxiv preprint

Figure12:Anglemap,S3.AnexampleofalefthemisphereanglemapderivedfrompRFmodelfitting.OtherwiseasFigure4.

.CC-BY-NC-ND 4.0 International licensecertified by peer review) is the author/funder. It is made available under aThe copyright holder for this preprint (which was notthis version posted January 5, 2016. . https://doi.org/10.1101/035980doi: bioRxiv preprint

Figure13:Mapcomparison,S1-S3. Four typesofdataare shown for three subjects: sulcalandgyral landmarks(column 1), eccentricity (column 2), angle (column 3), and mean fMRI signal (column 4). Comparisons of thedatasets reveal regularities across subjects, such as the fact that the ptCoS is well aligned with the hV4/VO-1boundarydefinedbyaneccentricityreversal(white line incolumn1).Therearealsodifferencesacrosssubjects.ForexamplethefovealrepresentationofhV4islessclearinS3thaninS1andS2.

DISCUSSION:RegularityofmapsandanatomyIdentifyingvisualfieldmapsisanimportantcomponentofcharacterizingtheorganizationandfunctionofvisualcortex.Itisimportanttohavewell-justifiedandreproduciblemethodstodefinethemaps.InthecaseoftheV1-V3maps,thefunctionalandanatomicalorganizationissufficientlyregularandwellunderstoodthatthesemapscanbeidentifiedusingautomatedprocedures(nohumanintervention)10,11.ThehV4andVOmapsarenotyetincludedinautomatedfittingprocedures;however,recentprogresssuggeststhatthesemaps,too,haveahighdegreeofregularity.Twoboundariesarewelldefinedbyretinotopicfeatures(sections6.2.2,6.2.3),andoneofthesealsocoincideswithananatomicallandmark,theptCoS.Moreover,theinternalstructureofthehV4mapanditsneighborsarewellunderstood,suchthattheiso-eccentricityandiso-anglelinesderivedfromretinotopicmappingfollowregularpatterns.

.CC-BY-NC-ND 4.0 International licensecertified by peer review) is the author/funder. It is made available under aThe copyright holder for this preprint (which was notthis version posted January 5, 2016. . https://doi.org/10.1101/035980doi: bioRxiv preprint

VariabilityacrosssubjectsWhilethemostofthelargescalestructuresintheventraloccipitalretinotopicmapsaresimilaracross subjects, there are also individual differences. Someof these differences likely reflectquantitativedifferencesbetweensubjectsinthesizeandlayoutofthemaps.Otherdifferencesreflectvarioussourcesofmeasurementnoise.Innocasewillameasuredmapexactlymatchatemplate, and in some cases map boundaries will be ambiguous. We believe that the bestapproach offered is to simultaneously satisfy asmany of the constraints from the anatomy,eccentricity,andanglemapsaspossible.

ACKNOWLEDGMENTS:Theauthorsacknowledge two funding sources,NIHGrantR00-EY022116 (JW)andNIHGrantR01-EY023915toKalanitGrillSpector(supportingNW).DISCLOSURES:Theauthorshavenothingtodisclose.

.CC-BY-NC-ND 4.0 International licensecertified by peer review) is the author/funder. It is made available under aThe copyright holder for this preprint (which was notthis version posted January 5, 2016. . https://doi.org/10.1101/035980doi: bioRxiv preprint

REFERENCES1 Wandell,B.A.&Winawer,J.Imagingretinotopicmapsinthehumanbrain.Vision

research51,718-737,doi:10.1016/j.visres.2010.08.004(2011).2 Wandell,B.A.,Dumoulin,S.O.&Brewer,A.A.Visualfieldmapsinhumancortex.

Neuron56,366-383(2007).3 Wang,L.,Mruczek,R.E.,Arcaro,M.J.&Kastner,S.ProbabilisticMapsofVisual

TopographyinHumanCortex.CerebCortex,doi:10.1093/cercor/bhu277(2014).4 Henschen,S.E.Onthevisualpathandcentre.Brain16,170-180(1893).5 Inouye,T.DieSehstörungenbeiSchussverletzungenderkortikalenSehsphäre:nach

BeobachtungenanVerwundetenderletztenjapanischenKriege.(W.Engelmann,1909).

6 Holmes,G.DisturbancesofVisionbyCerebralLesions.TheBritishjournalofophthalmology2,353-384(1918).

7 Horton,J.C.&Hoyt,W.F.Therepresentationofthevisualfieldinhumanstriatecortex.ArevisionoftheclassicHolmesmap.Archivesofophthalmology109,816-824(1991).

8 Kirson,D.,Huk,A.C.&Cormack,L.K.Quantifyingspatialuncertaintyofvisualareaboundariesinneuroimagingdata.JVis8,1011-15,doi:10.1167/8.10.10(2008).

9 Schira,M.M.,Tyler,C.W.,Breakspear,M.&Spehar,B.Thefovealconfluenceinhumanvisualcortex.JNeurosci29,9050-9058(2009).

10 Benson,N.C.,Butt,O.H.,Brainard,D.H.&Aguirre,G.K.CorrectionofdistortioninflattenedrepresentationsofthecorticalsurfaceallowspredictionofV1-V3functionalorganizationfromanatomy.PLoSComputBiol10,e1003538,doi:10.1371/journal.pcbi.1003538(2014).

11 Benson,N.C.etal.Theretinotopicorganizationofstriatecortexiswellpredictedbysurfacetopology.CurrBiol22,2081-2085,doi:10.1016/j.cub.2012.09.014(2012).

12 Hansen,K.A.,Kay,K.N.&Gallant,J.L.TopographicorganizationinandnearhumanvisualareaV4.JNeurosci27,11896-11911(2007).

13 Winawer,J.,Horiguchi,H.,Sayres,R.A.,Amano,K.&Wandell,B.A.MappinghV4andventraloccipitalcortex:thevenouseclipse.Journalofvision10,1,doi:10.1167/10.5.1(2010).

14 Hadjikhani,N.,Liu,A.K.,Dale,A.M.,Cavanagh,P.&Tootell,R.B.RetinotopyandcolorsensitivityinhumanvisualcorticalareaV8.NatNeurosci1,235-241(1998).

15 Brewer,A.A.,Liu,J.,Wade,A.R.&Wandell,B.A.Visualfieldmapsandstimulusselectivityinhumanventraloccipitalcortex.NatNeurosci8,1102-1109(2005).

16 Tootell,R.B.&Hadjikhani,N.Whereis'dorsalV4'inhumanvisualcortex?Retinotopic,topographicandfunctionalevidence.CerebCortex11,298-311(2001).

17 Larsson,J.&Heeger,D.J.Tworetinotopicvisualareasinhumanlateraloccipitalcortex.JNeurosci26,13128-13142(2006).

18 Witthoft,N.etal.WhereIsHumanV4?PredictingtheLocationofhV4andVO1fromCorticalFolding.CerebCortex,doi:10.1093/cercor/bht092(2013).

.CC-BY-NC-ND 4.0 International licensecertified by peer review) is the author/funder. It is made available under aThe copyright holder for this preprint (which was notthis version posted January 5, 2016. . https://doi.org/10.1101/035980doi: bioRxiv preprint

19 Zeki,S.etal.Adirectdemonstrationoffunctionalspecializationinhumanvisualcortex.JNeurosci11,641-649(1991).

20 Kay,K.N.,Naselaris,T.,Prenger,R.J.&Gallant,J.L.Identifyingnaturalimagesfromhumanbrainactivity.Nature452,352-355,doi:10.1038/nature06713(2008).

21 Meadows,J.C.Disturbedperceptionofcoloursassociatedwithlocalizedcerebrallesions.Brain97,615-632(1974).

22 Glover,G.H.&Lai,S.Self-navigatedspiralfMRI:interleavedversussingle-shot.MagnResonMed39,361-368(1998).

23 Engel,S.A.etal.fMRIofhumanvisualcortex.Nature369,525,doi:10.1038/369525a0(1994).

24 Dumoulin,S.O.&Wandell,B.A.Populationreceptivefieldestimatesinhumanvisualcortex.Neuroimage39,647-660,doi:10.1016/j.neuroimage.2007.09.034(2008).

25 Engel,S.A.,Glover,G.H.&Wandell,B.A.RetinotopicorganizationinhumanvisualcortexandthespatialprecisionoffunctionalMRI.CerebralCortex7,181-192(1997).

26 Teo,P.C.,Sapiro,G.&Wandell,B.A.CreatingconnectedrepresentationsofcorticalgraymatterforfunctionalMRIvisualization.IEEETransMedImaging16,852-863(1997).

27 Wandell,B.A.,Chial,S.&Backus,B.T.Visualizationandmeasurementofthecorticalsurface.JCognNeurosci12,739-752(2000).

28 Carman,G.J.,Drury,H.A.&VanEssen,D.C.Computationalmethodsforreconstructingandunfoldingthecerebralcortex.CerebCortex5,506-517(1995).

29 Dale,A.M.,Fischl,B.&Sereno,M.I.Corticalsurface-basedanalysis.I.Segmentationandsurfacereconstruction.NeuroImage9,179-194,doi:10.1006/nimg.1998.0395(1999).

30 Yushkevich,P.A.etal.User-guided3Dactivecontoursegmentationofanatomicalstructures:Significantlyimprovedefficiencyandreliability.Neuroimage31,1116-1128(2006).

31 Duvernoy,H.M.&Bourgouin,P.Thehumanbrain:surface,three-dimensionalsectionalanatomywithMRI,andbloodsupply.2ndcompletelyrev.andenl.edn,(Springer,1999).

32 Dumoulin,S.O.etal.Automaticvolumetricsegmentationofhumanvisualretinotopiccortex.Neuroimage18,576-587(2003).

33 Stensaas,S.S.,Eddington,D.K.&Dobelle,W.H.Thetopographyandvariabilityoftheprimaryvisualcortexinman.JNeurosurg40,747-755,doi:10.3171/jns.1974.40.6.0747(1974).

34 Sereno,M.I.etal.Bordersofmultiplevisualareasinhumansrevealedbyfunctionalmagneticresonanceimaging.Science268,889-893(1995).

.CC-BY-NC-ND 4.0 International licensecertified by peer review) is the author/funder. It is made available under aThe copyright holder for this preprint (which was notthis version posted January 5, 2016. . https://doi.org/10.1101/035980doi: bioRxiv preprint