资源描述：
《Statisticalshapemodelsfor3Dmedicalimagesegmentation,Areview.pdf》由会员上传分享，免费在线阅读，更多相关内容在学术论文-天天文库。
1、MedicalImageAnalysis13(2009)543–563ContentslistsavailableatScienceDirectMedicalImageAnalysisjournalhomepage:www.elsevier.com/locate/mediaStatisticalshapemodelsfor3Dmedicalimagesegmentation:AreviewTobiasHeimann*,Hans-PeterMeinzerDiv.ofMedicalandBiologicalInformatics,GermanCancerResearchCent
2、er,ImNeuenheimerFeld280,D-69120Heidelberg,GermanyarticleinfoabstractArticlehistory:Statisticalshapemodels(SSMs)havebynowbeenfirmlyestablishedasarobusttoolforsegmentationofReceived27August2007medicalimages.While2Dmodelshavebeeninusesincetheearly1990s,wide-spreadutilizationofReceivedinrevised
3、form24February2009three-dimensionalmodelsappearedonlyinrecentyears,primarilymadepossiblebybreakthroughsinAccepted18May2009automaticdetectionofshapecorrespondences.Inthisarticle,wereviewthetechniquesrequiredtocre-Availableonline27May2009ateandemploythese3DSSMs.Whileweconcentrateonlandmark-b
4、asedshaperepresentationsandthoroughlyexaminethemostpopularvariantsofActiveShapeandActiveAppearancemodels,wealsoKeywords:describeseveralalternativeapproachestostatisticalshapemodeling.StructuredintothetopicsofshapeStatisticalshapemodelrepresentation,modelconstruction,shapecorrespondence,loc
5、alappearancemodelsandsearchalgo-DeformablesurfaceActiveShapemodelrithms,wepresentanoverviewofthecurrentstateoftheartinthefield.WeconcludewithasurveyActiveAppearancemodelofapplicationsinthemedicalfieldandadiscussionoffuturedevelopments.Ó2009ElsevierB.V.Allrightsreserved.1.Introductionconcentr
6、ateonmethodsforvolumetricimages.However,manymodelingmethodshaveonlybeenappliedto2Dsofar.TotheInthelasttwodecades,model-basedsegmentationapproachesextentthatsuchmethodscanbegeneralizedorextendedtothehavebeenestablishedasoneofthemostsuccessfulmethodsfor3Dcase,wehaveincludedtheminthepresentre
7、view.Wehaveimageanalysis.Bymatchingamodelwhichcontainsinformationfurthermoreincludedsomemethodswhichhaveshowntobeveryabouttheexpectedshapeandappearanceofthestructureofinter-successfulin2D,butwhicharetechnicallynotfeasiblein3D–esttonewimages,thesegmentationisco
