can we get around the crossing fiber problem by … word - crossing_vs_resolution_abstract_v2.docx...
TRANSCRIPT
Canwegetaroundthecrossingfiberproblembyincreasingresolution?KurtSchilling,VaibhavJanve,YuruiGao,IwonaStepniewska,BennettALandman,AdamWAnderson
VanderbiltUniversityInstituteofImagingScience,VanderbiltUniversity,Nashville,TN
PURPOSE It is nowwidely recognized that voxels with crossing fibers and complex configurations present achallenge fordiffusionMRI (dMRI) reconstructionand fiber trackingalgorithms.This “crossing fiber”problemhasbeenestimatedtoaffectanywherefrom30%[1]to90%[2]ofwhitematter(WM)voxels,anditisgenerallyassumedthatincreasingthespatialresolutionasmuchaspossiblewillminimizethenumberofvoxelscontainingmultiple fiber populations. Here, using both dMRI and histology from the same brain, we investigate theprevalenceofthiscrossingfiberproblemacrossarangeofspatialresolutions.METHODSdMRI:anexvivomacaquehemispherewasscannedona9.4TAgilentscanner(b-value6,000s/mm2;101 gradient directions) at isotropic resolutions ranging from 800um to 300um, in increments of 100um.Crossing fibers were detected following [2], using constrained spherical deconvolution [3]. Histology: 16histological sectionswere stained using Gallyasmyelin stain and imaged on a Leica Brightfieldmicroscope at0.5um/pixel.Structure tensor (ST)analysis [4]wasused toextractorientationestimates foreverypixel in theimage.Theimagewasthendownsampledtoresolutionsrangingfrom32umto1024um,andtheFODwasderivedfor each downsampled block. FOD’s were fit to single and mixture vonMises distributions to determine thepresenceofcrossingfibers.RESULTSFigure1showstheresultsofSTanalysisinaregionwithin-planefibercrossings.Wefindthat,inbothMRI and histology, thepercentofvoxelswithcrossingfibersactuallyincreasesasweincreasetheresolution. Themiddleandbottomrowshighlight “single” fiberregions inredand“crossing” fibersasgreen,andqualitativelyshow thatmany regionswith a “single” fiberpopulation at a coarse resolution exhibit crossingpatternswhenprocessedatafinerresolution.Thispatternwasobservedacrossthewholebrain,inbothMRIandhistology,andinbothgrayandwhitematter.Ananalysisofthecrossingangleinhistologyshowsthatthefiberstendtocrossatmoreacuteanglesinthehigherresolutiondatasets.DISCUSSION The problem of crossing fibers appears to be a fundamental limitation associated with theanatomical geometry, rather than a technical problem that can be overcome with higher fields, strongergradients, and technological advances thatmay increase spatial resolution. This increase in crossing fibers athigherresolutionislikelyaresultofadecreaseinpartialvolumeeffectsbetweenbundlesthatsharesimilar,butnotexactlythesame,orientations.Thesecomplexconfigurationsalsotakeplaceinregionstypicallyassumedtocontain single fiberpopulations (i.e. corpus callosum),whereheterogeneous fiberorientationsareapparent inhistologicalsections.CONCLUSIONBecausevoxelswithcomplexfiberconfigurationswillalwaysexistindatasets,evenatresolutionsfarbeyondcurrentdMRIcapabilities,itiscriticaltofocusontissuemodelsfordescribingthefibergeometryinvoxels, rather than focus on pushing resolution (and sacrificing SNR)where gains in fiber reconstruction andfibertrackingwillbeminimal.REFERENCES 1. Behrens, T. Neuroimage, 2007. 2. Jeurissen, B. Hum Brain Mapp, 2013. 3. Tournier, JD.Neuroimage,2008.4.Budde,MD,Frank,JA.Neuroimage,2012.