肿瘤计算机辅助诊断系统关键技术的研究

《肿瘤计算机辅助诊断系统关键技术的研究》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、东南大学博士学位论文肿瘤计算机辅助诊断系统关键技术的研究姓名：姜晓彤申请学位级别：博士专业：生物医学工程指导教师：罗立民20040501东南人学博士研究生论文AbstractThecancerharmsthepeople’shealthgreatly，howeverearlydetectanddiagnosiscanimprovethesurvivalrateofcancerpatients．MedicalDevicehasbeengreatlyimprovedinrecentyears．Theimageresolvingpowerandimagi

2、ngspeedbecomehigherandhigher，theimagecontainsmorgandmorevaluableinformationofpatient’sorganortissue．Newimagingdevices,suchasPET、fMRIetal，whichcandetectthemetabolismchangeofthebedy，aremoreandmoreperfect．Thesebringgoodchancestocancercomputeraided—diagnosissystem．Medicalquantitat

3、iveimaginganalysisisoneofthemostchallengingresearchtopicsinimagingsciencerecenfly．Generally，thistechniqueconstructsphysics、mathematicsmodelbasedoncfinicexperience，abstractstumorvaluablefeaturesbycomputerimageprocessing、artificialintelligenceetal，thenfusesallvaluablefeaturesand

4、physiologicaldataofpatientstoanalyseanddiagnosistumors．ThisdissertationmakesaresearchonthesystemframeofTumorComoputerAidded-diagnosisSystemandsomekeytechniques，suchasthesegmentationofsoakagetumors，theanalysisofthree·dimensionaltumorshape、theanalysisofthree-dimensionaltumortext

5、ure、three·dimensionaltumorsurfacereconstructandtumourgrowthmodel，arediscussedthroughly．Anautomaticsegmentationmethodofsoakagetumor，whichfusesCTandPETimages，ispresented．Thismethodcanconfirmthesegmentationresultincludethewholetumorandhigherspaceresolution．Makeathroughresearchonm

6、ultiparametersoptimizationanddesigntwonewoptimizemethods：(a)Thehybridoptimizationbasedonchaodcsimulatedannealingandsimplexsearch．(b)Thesimulatedannealingoptimizationbasedonsimplexsearch．Anewmethodoflungtumoridentificationbasedonquantitativethree-dimensionalshapeanalysisispropo

7、sed．Wedefinedthreetwo—dimensionalshapedescriptorscorrelatingwithtumortypefromdifferentpointsofview；thenfourthree-dimensionalshapedescriptorsweredefinedasinputwdueofRadialBasisFunctionneuralnettorealizetheautomaticclassificationoflungtumot．Anapproachwaspmposedforsurfacereconstr

8、uctionfromcontoursbasedonstringmatchingalgorithm．Theexperimen