资源描述：
《Hongju Zhao- ECM-aware cell-graph mining for bone tissue modeling and classification》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、DataMinKnowlDisc(2010)20:416438DOI10.1007/s10618-009-0153-2ECM-awarecell-graphminingforbonetissuemodelingandclassificationCemalCagatayBilgin·PeterBullough·GeorgeE.Plopper·BülentYenerReceived:20January2009/Accepted:24September2009/Publishedonline:22October2009TheAuthor(s)2009AbstractPathologicalexa
2、minationofabiopsyisthemostreliableandwidelyusedtechniquetodiagnosebonecancer.However,itsuffersfrombothinter-andintra-observersubjectivity.Techniquesforautomatedtissuemodelingandclassificationcanreducethissubjectivityandincreasestheaccuracyofbonecancerdiagnosis.Thispaperpresentsagraphtheoreticalmet
3、hod,calledextracellularmatrix(ECM)-awarecell-graphmining,thatcombinestheECMformationwiththedistributionofcellsinhematoxylinandeosinstainedhistopathologicalimagesofbonetissuessamples.Thismethodcanidentifydifferenttypesofcellsthatcoexistinthesametissueasaresultofitsfunctionalstate.Thus,itmodelsthes
4、tructure-functionrelationshipsmorepreciselyandclassifiesbonetissuesamplesaccuratelyforcancerdiagnosis.Thetissueimagesaresegmented,usingtheeigenvaluesoftheHessianmatrix,tocom-putespatialcoordinatesofcellnucleiasthenodesofcorrespondingcell-graph.Uponsegmentationacolorcodeisassignedtoeachnodebasedont
5、hecompositionofitssurroundingECM.Anedgeishypothesized(andestablished)betweenapairofnodesifthecorrespondingcellmembranesareinphysicalcontactandiftheysharetheResponsibleeditor:R.BharatRaoandRomerRosales.C.C.Bilgin(B)·B.YenerDepartmentofComputerScience,RensselaerPolytechnicInstitute,Troy,NY12180,USA
6、e-mail:bilgic@cs.rpi.eduB.Yenere-mail:yener@cs.rpi.eduG.E.PlopperDepartmentofBiology,RensselaerPolytechnicInstitute,Troy,NY12180,USAe-mail:plopg@rpi.eduP.BulloughDepartmentofLaboratoryMedicine,HospitalforSpecialSurgery,NewYork,NY10021,USA123ECM-awarecell-graphmining417samecolor.Hence,multiplecolo
7、red-cell-graphscoexistinatissueeachmodelingadifferentcell-typeorganization.BothtopologicalandspectralfeaturesofECM-awarecell-graphsarecomputedtoquantifythestructuralpropertiesoftissuesamplesandclassifytheirdifferentfun
