基于多任意形状的ROI编码框架的数字化乳腺X线摄影新诊断Loseless压缩方法的(IJMECS-V

I.J.ModernEducationandComputerScience,2011,5,33-39PublishedOnlineAugust2011inMECS()Copyright©2011MECSI.J.ModernEducationandComputerScience,2011,5,33-39ANewDiagnosisLoselessCompressionMethodforDigitalMammographyBasedonMultipleArbitraryShapeROIsCodingFrameworkPingXu1,YanZuo2,Wei-DongXu1,andHua-JieChen21:CollegeofLifeInformationScience&InstrumentEngineering,HangzhouDianziUniversityHangzhou,Zhejiang,ChinaE-MAIL：[e-mail],[e-mail]2:CollegeofAutomation,HangzhouDianziUniversity,Hangzhou,Zhejiang,ChinaE-MAIL：[e-mail],[e-mail]—Withtherapidlygrowinguseofdigitalimagesinmedicalarchivalandcommunication,imagecompressiontechnology,especiallydiagnosislosslesscompressiontechnology,playsamoreandmoreimportantroleformedicalapplications.Inthisthesis,anoveldiagnosisloselesscompressionalgorithmispresentedfordigitalmammography.Themammogramisdividedintobreastregion,pectoralmuscleandbackgroundusingtheCADtechnology.ThenmutiplearbitraryshapeROIscodingframeworkisusedtocompressthemammograminwhichthebreastregionandpectoralmusclearecompressedlosslesslyandlossilyrespectively,andthebackgroundcanbediscardedorcompressedlossilyasuser’swill.Experimentalresultsshowthattheproposedmethodofferpotentialadvantageinmedicalapplicationsofdigitalmammographycompression.IndexTerms—digitalmammography;diagnosislosslesscompression;CAD;ROII.INTRODUCTIONInthelatesttwentyyears,themammogramcancerhasbecomeoneofthemostdangerousmalignanttumorsofwomenwhosedeathratiohasbeenmorethanfortypercent.Withtherapiddevelopmentoflivingstandardandrequirementsofhealthcare,doctorsadvisewomentotakemamogramstwiceeveryyear.Itisestimatedthatapproximately10%-30%ofbreastcancercasesaremissedbyradiologists[1][2].Computer-aiddiagnosis(CAD)systemshavebeenwidelyusedtohelpimprovethedetectionprecision[5][6].DiagnosisinformationofdigitalmammographyneedbesavedlosslesslyinCAD.Withoutcompression,thesizeofeachdigitalmammogramcanbemorethan4MB.Thisbringsahugechallengetothecurrentmedicalsystem.Severallosslessandlossycompressionmethodhavebeenpresentedtoresolvethisproblem[7-18].However,losslesscompressionhasbroughtaboutatmostonly4:1compressionratio.Mostlossycompressionalgorithmsneedtakeintoaccountthespecialclinicalissuetobedealtwith[8-17].Receiveroperationcharacteristic(ROC)analysisonlossycompressionshowthatitispromisingtouselossytechniquesinmedicalimagecompres-sion[8].Inrecentyears,digitalmammographycompressionhasbecomearesearchfocusinthefieldofmedicalimageprocessing.Previousstudieshaveevaluated,withCADsystemsorobserver’sperformancestudies,lossycompressionindigitalmammography.GoodandZhengassessedthedetectionofmassesandclusteredmicrocalcificationsinJPEGmammogramsbymeansofROCstudy[9][10].Kocsisfoundthatcompressedmammogramswithwavelettransformalgorithmat40:1compressionratioprovidedperceptuallylosslesscompression[11].Perlmutteretal.foundnosignificantdifferencesbetweenoriginalandcompressedimagesat80:1compressionratioof57digitalmammogramscompressedwiththesetpartitioninginhierarchicaltrees(SPIHT)algorithm[12].Sung,SuryanarayananandPenedocomparedtheperformanceofdetectionofprimarysignsofbreastcancerusingoriginalimagesandimagesreconstructedat20:1compressionratioafterJPEG2000compression[13-15].Inasimilarstudy,Suryanarayananetal.obtainedthatJPEG2000doesnotaffecttheCADschemefordetectingmasses,butdetectionclusterofmicrocalcificationsisaffectedwithcompressionat30:1compressionratio[16].Idrisfoundthatfalsepositiverateofmicrocalcificationswasashighas66percentandsomedetailswithlessonsignoredat100:1compressionratioafterJPEG2000compression,whichmayleadtounnecessarymisdiagnosis[17].PenedofoundthatdetectionofmicrocalcificationclustersandmasseswasnotfailurewhensegmentedbreastregioniscompressedbyJPEG2000andOB-SPIHT(object-basedsetpartitioninginhierarchicaltrees)atthecompressionratioof40:1and80:1[18].Inthispaper,anewdiagnosislosslesscompressionalgorithmbasedonCADtechnologyandmutiplearbitraryshaperegionofinterests(ROIs)codingframeworkisproposedfordigitalmammographytoThisworkwassupportedbyTheNationalNaturalScienceFoundationofChina(61004119,60705016,30800248),andTheNaturalScienceFoundationofZhejiangProvince(Y1080674)34ANewDiagnosisLoselessCompressionMethodforDigitalMammographyBasedonMultipleArbitraryShapeROIsCodingFrameworkCopyright©2011MECSI.J.ModernEducationandComputerScience,2011,5,33-39dividedΩΩ1ΩM-1ΩMISA-DWTISA-DWTISA-DWTΛ1ΛMModifiedSPIHTModifiedSPIHTCodingStream{}MiiDL1max_=RΛM-1RM-1R1׃׃׃׃׃׃׃׃׃׃ModifiedSPIHTFig.1FlowdiagramofmutiplearbitraryshapeROIscodingframeworkhighcompressionratiowhilemaintainingthediagnosisinformationlosslessly.ThispaperisorganizedwithSectionIIpresentingtheproposedcompressionmethod.SectionIIIprovidestheexperimentalresults.SectionIVdrawstheconclusion.II.PROPOSEDMETHODA.Multiple,arbitraryshapeROIscodingframeworkMutiple,arbitraryshapeROIscodingframeworkisgivenasshowninFigure.1.Thecodingframeworkcanbesummarizedasfollows:Firstly,theimageplaneispartitionedintomultipleROIswithdifferentprioritieswhicharemarkedbyagraymask,inwhichthehighergrayvalueo