hibrid segmentation of mass in mammograms using template matching amd dynamic programming

《hibrid segmentation of mass in mammograms using template matching amd dynamic programming》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、HybridSegmentationofMassinMammogramsUsingTemplateMatchingandDynamicProgrammingEnminSong,PhD,ShengzhouXu,PhD,XiangyangXu,PhD,JianyeZeng,MD,YihuaLan,PhD,ShenyiZhang,MSc,Chih-ChengHung,PhDRationaleandObjectives:Accurateimagesegmentationforbreastlesionsi

2、sacriticalstepincomputer-aideddiagnosissystems.Theobjectiveofthisstudywastodeveloparobustmethodfortheautomaticsegmentationofbreastmassesonmammogramstoextractfeasiblefeaturesforcomputer-aideddiagnosissystems.MaterialsandMethods:Thedatasetusedinthisstu

3、dyconsistedof483regionsofinterestextractedfrom328patients.Ahybridmethodforsegmentingbreastmasseswasproposedonthebasisofthetemplate-matchinganddynamicprogrammingtechniques.First,atemplate-matchingtechniquewasusedtolocateandobtaintheroughregionofmasses

4、.Then,onthebasisofthisroughregion,alocalcostfunctionfordynamicprogrammingwasdefined.Finally,theoptimalcontourwasderivedbyapplyingdynamicprogrammingasanoptimizationtechnique.Theperformanceofthisproposedsegmentationmethodwasevaluatedusingarea-basedandbo

5、undarydistance–basedsimilaritymeasuresbasedonradiologists’manuallymarkedannotations.Acomparisonwiththreedifferentsegmentationalgorithmsonthedatasetwasprovided.Results:Themeanoverlappercentageforourproposedhybridmethodwas0.7270.127,whereasthoseforTim

6、pandKarssemeijer’sdynamicprogrammingmethod,Songetal’splane-fittinganddynamicprogrammingmethod,andthenormalizedcutsegmentationmethodwere0.6570.216,0.6360.190,and0.5620.199,respectively.AllPvaluesforthemeasuredistributionofourproposedmethodandtheothe

7、rthreealgorithmswere<.001.Conclusions:Ahybridmethodbasedonthetemplate-matchinganddynamicprogrammingtechniqueswasproposedtosegmentbreastmassesonmammograms.Evaluationresultsindicatethattheproposedsegmentationmethodcanimprovetheaccuracyofmasssegmen-tati

8、oncomparedtothreeotheralgorithms.Theproposedsegmentationmethodshowsbetterperformanceandhasgreatpotentialinimprovingtheaccuracyofcomputer-aideddiagnosissystemsininterpretingmammograms.KeyWords:Computer-aideddiagnosis;masssegmentation;mammogram;templat