医学图像数据挖掘若干技术研究

《医学图像数据挖掘若干技术研究》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、Y7818铽分类号!￡3l!密级公理一IJDC——学号Qi篮Z2⑨东南大掌博士学位论文医学图像数据挖掘若干技术研究研究生姓名：导师姓名：申请学位级别一工堂。蠼．士．警科专业名称荭薹扭座援燕苤论文提交日期星壁鲢．笺!墨且．垫目．论戈馨辩日期缝§生目旦学位授予单位塞瘟太堂学位授予日翘鞠生县旦答辩委员会主席评阅人2004．年12月25臼Abs仃actMedicalimagediagnosisisoneofthemajormethodstomakenon—injuringdiagnosisinmedic

2、ineTherearemanydatafeatureandILlleinfomaationthatwillbediscoveredandrecognizedinmedicalimage．Nowclusteringandassociationrulemininghasbeenaflom-ishingresearchareaandwidelyrecognizedasakeytoolforminingvaluableinformationandknowledgefromthedatabases．Ima

3、geminingdealswiththeextractionofimplicitknowledge，imagedatarelationship，orotherpatternsnotexplicitlystoredintheimages，Theresearchonmedicalimagemininghasjliststarted．Itissignificanttoresearchandfindpowerfuldataminingalgorithmsthatcanhandlemedicalimage

4、datasetseffectivelyandefficientlyThissituationshapesupthebackgroundofresearchesinthisthesisThegrayanddensityofpixelisamaincontentshowingmedicalimage．Thesemanticrelationshipbetweenthegraydensityofmedicalimageandtheanatomyofhumanorgantissuesisdiscusshe

5、re．Afeatureexpressionanddataprocessingmethod，whichcanbeadaptedtomedicalimagedatamining，ispresented．Theprocessofthemedicalimagedataminingisoffered．Detalledstudyisdonetoimprovetheaccuznaeyofthebinnedkereeldensityestimationmethodsandapproximationkerneld

6、ensityfunctionbasedongrid-based．Clusteringandfeatureextractionalgorithmsarepurposedbasedongrid-baseddensityfunctioncharacterizationofthemedicalimagedataset．Besides，findingarbitraryshapedclustersalsohasbeenproposedsuccessfully．Basedontheapproximatedde

7、nsityconstruction，theexperimentalstudiesforextractingmedicalimagefeaturehavebeenpurposedinthethesisTheprocessingstrategiesandexperimentalresultsprovetherationalityandavailabilityofthenovelmethod．Associationrulemininghasbeenappliedtolargeimagedata．FP—

8、growthalgorithmisproposedforenhancingefficiencyofApfiofialgorithm．Therulethatclassifiesbetweennormalandanomalisticorganimageisdiscoveredusingassociationrulemining．DMFACIAfDiscoveryMaximumFrequentAssociationClassifierltemsetsAlgorithm)ispresentedandap