基于超声图像的前列腺病变计算机辅助诊断

《基于超声图像的前列腺病变计算机辅助诊断》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、AbstractABSTRACTProstatecaIlcerisoneofthecommondiagIlosedmali印anciesi11middle．agedaIldeldermeIl，a11dthesuⅣivaIrateofthepatientscaIloIllybeenctl；眦cedbydetectioninthee训ystageofcancer．Nowadays，TransRectalUltrasound(，nWS)puIl咖etecⅢqueisoneofthemost毹quentlyuSedmethod

2、sfortheedydetection0fshowthattheaCcuracyofthepatholog)rdetectionprostatecancer．MaIlyresearchescanbe缸恤ere11llancedbyusingcompmeraideddiagnosis(CAD)technolo黟1heprostatec觚cerCADsystemllsuallyperfbmsthediagnosisbaSed0nt：hemorphologyalldgrayscaletexturestatisticalfea

3、turesofthepatllologicalareaS．Ho、ⅣeVer，theCADsystem011Iyutilizest】heaforementioned佗aturescan’tdi行．erentiatethepathoIogyareasefrectivelybecauseofthepoorcontrastintheTl≈USimages．Theref．ore，nlisdissertationstudiesmedi虢rentcharacteristicsof也epathologicala∞ashmedi仃ere

4、mscalesa11d舭qucncydomainst0assistthediagnosisofprostatecancerSe位ctiVely．Themainresearchworkandcontributionsofthisdissertationcanbes岫m撕zedaSfollows：(1)Preprocessingthe1RUSimages．Aimingatthecharacteristicsmatmeultrasoul】dimageshavea10tofspecklenoises，趾hproVedaIlis

5、otropicdif如sionwasproposed．’nlealgorithmusedthcmediansadaptivespecklerIeductionalgorithmofregionstoreplacetheoriginalpixelsandaddedthediagonalgradienttenn，wllichwouldeIlllancethedetailinfomlation．111esuppressthespecklenoisese髓ctively肌dalgoritb吼combinedafideli锣te

6、rm，、7v_llichmadethealgomhmsuppresstllespecklemaintainthenoiseeffectivelyaIldmorerobust，moreover，thcalgoritllIncouldbouIldar)rgeomenyinfomatione虢ctiVelyandsolvemeproblemthattheresultsoftheconventionalalgorit№swere甜rectedbytheiteratiVeImmbers铲eatly．(2)Segmentingth

7、eTRUSimagestogettheprostatebo邶da够T11isdissertationbasedontheproposedanautomaticprostatebo眦d2urysegmentationalgoritllIncoarsecontourcouldbegottenbyiⅡlprovedgeodesicactiveconto比Firstly，aradialbas．reliefandregionfiUalgorithmcontourcombiningtothe1RUSimages，thecouldb

8、eusedastllei11itialcontourfortheactivecontourmodel．Thentheproblemof、Ⅳeakedgescouldalsobee伍cientlysolvedbypresentinganewregion．basedsignedforces‰ctiont0fhnctionandiIlc