ISO IEC 03532-2-2024.docx

ISOInternationa1.StandardISozIEC35322Firstedition2024-02Informationtechno1.ogyMedica1.image-basedmode1.1.ingfor3Dprinting一F台的格ntationTechno1.ogiesde!'informationMode1.isationmedica1.edbased'imagespour!'impression3DPartie2:SegmentationReferencenumberISO/IEC3532-22024(en)COPYRIGHTPROTECTEDDOCUMENT©ISO/IEC2024Wintdcddu1.fk*jdOthenviseCbyqwipMrtiW1.eombXttrCni<wkjhrtRkkCTrtWpm9HtUSXnPyiP脾11匕mienmaytheinternetoranintranet,withoutpriorwrittenpermission.PermissiononberequestedfromeitherISOatthebe1.oworISOsmemberbodyinthecountryoftherequester.b,Vyfrfiandonnet8CH-1214Vcmicr,GenevaPhoae:*41227490111聊独te:薪括嬴OQrgPub1.ishedinSwitzer1.andContentsPageForewordIntroductionScopeNormativereferencesTermsanddefinitionsAbbreviatedtermsObjectiveofsegmentation5.1 Background5.2 TypesofsegmentationmethodsOvera1.1.segmentationprocessStep1.jdatapreparationSteP2：耙'瑞再做5桶HgforsegmentationStep4：se1.ectionofsegmentationnehvorkmode1.Step5:performanceeva1.uation7Datapreparation7.1Genera1.667.2Medica1.image67.2.267.3Preparationsteps7怔魂局CqUiWtion787.3.3ImagereconstructionPreprocessingforsegmentation78.1Genera1.789Annotation99.1Data1.abe1.1.ing99.2Preprocessingforannotation910Se1.ectionofnetworkmode1.10101.Genera1.<(.1010.2Inputpatch.11Eva1.uation111.1 1Genera1.111.2 2EVa1.Ua1.1.Onmetricsa*>.>.*.>>*-.a.a.>>.a.s.*><>121213 Dep1.oymentandrunning13Post-processingforsegmentation14Annex A (informative)CTscanningconditionsfororbita1.bonesegmentation15Annex B (informative)Characteristicsoforbita1.bonesegmentationfromCT16Annex C (informative)Deep1.earningtechniques18Annex D (informative)Considerationsforovera1.1.segmentationperformance19Bib1.iography24ForewordISO(theInternationa1.OrganizationforStandardization)andIEC(theInternationa1.E1.ectrotechnica1.GtHmhiEotUamon圈SPadiIHPe出邠tbmfdewvmMidkdh的nd11bahUon.沁Idin旅H岫mghtechniocommitteesestab1.ishedbytherespectiveorganizationtodea1.withparticu1.arfie1.dsoftechnica1.activity.ISOandIECmitteesco1.1.aborateinfie1.dsofmutua1.interest.Otherinternationa1.organizations.governmenta1.andnon-governmenta1.,in1.iaisonwithISOandIEC,a1.sotakepartinthework.Theproceduresusedtodeve1.opthisdocumentandthoseintendedforitsfurthermaintenancearedescribedintheISO/IECDirectives,Part1.Inparticu1.ar,thedifferentapprova1.criterianeededforthedifferenttypes岷雌s,s出斯R削副WjI麻蝴中*略皿4“碳国呻W浦用5科g*MFiiIHoftheISO/ISOandIECdrawattentiontothepossibi1.itythattheimp1.ementationofthisdocumentmayinvo1.vethe暗用d（八）patent配曲"即跳雌林加品依屏RfngPUbI四ioneW掘谑W立以喊9两0躺fR踮anyreceivednoticeo11a)patent(三)whichmayberequiredtoimp1.ementthisdocument.However,Imp1.ementersarecautionedthatthismayn4"叩"皿1.h1.atestinformation,whichmaybeobtainedfromthepatenthft§1.ei1.ab1.eidtifyinW,u'w.1.>¾Vjypateni1.1.t-httpd.ISOandIECsha1.1.notbehe1.dnytradenameusedinthisdocumentisinformationgivenfortheconvenienceofusersanddoesnotconstituteanendorsement.Foranexp1.anationofthevo1.untarynatureofstandards,themeaningofISOspecifictermsandexpressionsre1.atedtoconformityassessment,aswe1.1.asinformationaboutISO*sadherencetotheWor1.dTradeb1.陟6Sfi活nWTrade(TBT)seewwwriofoFewoFdJum1.ThisdocumentwaspreparedbyJointTechnica1.CommitteeISO/IECJTC1,Informationtechno1.ogy.A1.istofa1.1.partsintheISO/IEC3532seriescanbefoundontheISOandIECwebsites.Anyfeedbackorquestionsonthisdocumentshou1.dbedirectedtothebody.Acomp1.ete1.istingofthesebodiescanbefoundatvr¼rw.iso.orgmembers.htm1.andIntroductionThisdocumentwasdeve1.opedinresponsetotheneedforcustomizationof3Dprintingtechno1.ogyinthemedica1.industrythroughtheuseofinformationandcommunicationtechno1.ogy(ICT).Therearemanypointswheretheexistingstandardsforadditivemanufacturing(AM)donotmatchtherequirementsofthemedica1.industry.Frommedica1.imagesto3Dprinting,medica1.devicedeve1.opmentisquiteacomp1.exjourneysvithcomp1.icatedmanagementofmu1.tip1.epiecesofsoftware.Withtheemergingmarkettormedica1.3Dprintedparts,therearemanypointsrequiringstandardization.Meis理抨M9%ndard网蚓s1.W卸岫虫W例桃颂三84va1.idationpr蜘(蔗的他6%Formedica1.3Dprinting,segmentationtechniquesshou1.dbeoptimizedandcombinedaccordingtothecharacteristicsofthemedica1.imagesandcorrespondingbodypartstogetanoptima1.3Dmode1.Inparticu1.ar,duringmedica1.imagesegmentation,identificationofthepixe1.soforgansor1.esionsfromrawdatasuchascomputedtomography(CT)ormagneticresonance(MR)images,isoneofthemostcha1.1.engingana1.ysistasks.Forexamp1.e,segmentationoftheorbita1.boneisnecessaryfororbita1.wa1.1.reconstructionincranio-maxi1.1.ofacia1.surgerytosupporttheeyeg1.obepositionandrestorethevo1.umeandshapeoftheorbit.丽ever,inW制IWn电arPH做玳捆丽诉股腌施他2中第1如附框nsit痴"苦,郎碓则出CortiCa1.bonewithaThehumanboneisde1.ineatedandextractedbysegmentationtechniques,anda3Dske1.eta1.mode1.isbui1.tisfE阴赛盘三t稀JMra除用摘ZatiO晶d囹擀用duringthissegmentation哭硼蛹姗般迎Sin翻痕beforeproceeding.械肥内rates格跟婕BAgXtati相盟曲MXeSP如t同欷,ever,co的所恐&册y(nds掴NruesoAMoUte)PeratOE茯。Ptedminimizationoferrorsduringthisjob,operatorsshou1.dknowwhichsegmentationtechniqueismostusedintheirimagingsoftwareandpossessthenecessaryski1.1.sforthattechnique.QP脚枷I1.edSMk1.掰丽喇卅瞰ft好TB悭媚entatim丽刑砒MyTheprob1.emisusua1.1.yunder-segmentation.However,over-segmentationwi1.1.a1.sobeprob1.ematicforfurtherdesigningprocesses,especia1.1.yforsurgica1.imp1.ants.Varioustechniqueshavebeensuggestedtoreducehumanerrorandimproveperformanceandconsistencyforsegmentationissues同Thisdocumentproposesastandardizedprocessfortheoptimizationofsegmentation.Informationtechno1.ogyMedica1.image-basedmode1.1.ingfor3Dprinting一ggntation1ScopeThisdocumentprovidesanoverviewofthesegmentationprocessformedica1.image-basedmode1.1.ingOfhumanbone.Segmentation.ThisdocumentspecifiesastandardizedprocesstoimprovetheperformanceofhumanboneThisdocumentisa1.soapp1.icab1.etomedica1.3Dprintingsystemsthatinc1.udemedica1.3Dmode1.1.ing2capabi1.ities.NormativereferencesR不眦Htg6entsdtffi1efe11<AMA0机底W四布a郎ifH由由诅Q1.a初制6tutesthe1.atesteditionofthereferenceddocument(inc1.udinganyamendments)app1.ies.ISO15708-1,NondestructivetestingRadiationmethodsforcomputedtomographyPart1:Termino1.ogyISO/IEC2382rInformationtechno1.ogyVocabu1.ary31W6FFHfttiOnsmanufacturingGenera1.princip1.esOvemewofdataprocessingfiWI/hS举ftJf1.Qh1够HCWingdWUKent,thetermsanddefinitionsgiveninISO/ASTM52950,ISO15708-1.rISOandIECmaintainIermino1.ogjrdatabasesforuseinstandardizationatthefo1.1.owingaddresses：ISOOn1.inebrowsingp1.atform：avai1.ab1.eathttps卜WWWiSoQgobpj-1ECE1.ectropedia:avai1.ab1.eathttps:WWW.e1.ectropedia.org/imageacquisitionscanningofthestructureOfinterestusingcomputedtomography(CT),magneticresonance(MR)imagingorgtgerthree-dimensiona1.imagingtechno1.ogyimageannotation股CeSSofattaching1.abe1.stoanimage1.abe1.c1.assifyingphraseornameapp1.iedtoatarget3.41.earning<machine1.eaming>processbywhichabio1.ogica1.oranautomaticsystemgainsknow1.edgeorski1.1.sthatitmayusetoimproveitsperformanceNotestoentryhavebeenremoved.SOURCE:1SOIEC2382:2015,2122966,modifiedsegmentationprocessOfseparatingtheobjectsOfinterestfromtheirSunoundingsNote1toentry:Segmentationcanbeapp1.icab1.eto2D.3D,rasterorvectordata.3.6ground-tnith1.abe1.§丐ree1.answerofthetrainingsetforsegmentationbasedonsupervised1.earningregionOfinterestRO1.兆Cifiodboundaryasdefinedintheimagemachine1.earningM1.processofoptimizingmode1.parametersthroughcomputationa1.techniques,suchthatthernodcsbehaviourref1.ectsthedataorexperiencepgURCE:ISO!EC22989:2022,3.3.5oneormore1.abe1.s1.abe1.1.eddata8POfdatathathavebeentaggedwithIiyperparanietercharacteristicofamachine1.earninga1.gorithmthataffectsitsIeaniingprocessNote1toparameters.entr)r：Hyperparametersarese1.ectedpriortotrainingandcanbeusedinprocessestohe1.pestimatemode1.Note2toentry:Examp1.esofhyperparametersinc1.udenumberofnetwork1.ayers,widthofeachIaycrztypeofactivation的眠砒BIIW廊qfty总艇性a三ksd1.svtM肃取圃NftiHi愉厢时811Aseisexpectationmaximizationa1.gorithm：thenumberotGaussiansinaGaussianmixture.依QWRCE:ISO/IEC22989:2022,3.3.4Jmedica1.imagetypeofimagesgeneratedbymedica1.imagingdevices4AbbreviatedtermsA1.artificia1.inte1.1.igenceAIMannotationandimagemarkupCDcomputer-aideddiagnosisCC1.CNNworkCRFCTconditiona1.randomfie1.dcomputedtomographyDICOMD1.digita1.imagingandcommunicationsinmedicinedeep1.earningDSCFCNworkFOVFNfie1.dofviewfa1.senegativeFPHUfa1.sepositiveHounsfie1.dunitIoUMIoUintersectionoverunionmeanintersectionoverunionM1.MRmachine1.earningmagneticresonanceMRINIfTImagneticresonanceimagingneuroimaginginformaticstechno1.ogrinitiativePETSPECTpositionemissiontomographysing1.e-photonemissioncomputerizedtomographyTNTPtruenegativetruepositive5sObjective2e三址帆qo*5.1BackgroundThepurposeofsegmentationistoextractaspecificregionororganfromapatient'sCT/MRmedica1.imageanduseittocreatea3Dmode1.Segmentationistheprocessofpartitioninganimageintodifferentmeaningfu1.segments.Formedica1.image¾qyre11tntmodatahojuwibbdyIdpartieoptji3wiiden1.(5U)YrrtmUrttion.ac(fDhdinHutanichartoienbtcdofandextractedbysegmentationtechniques,anda3Dske1.eta1.mode1.isbui1.tfromthissegmentation.Theminimizationoferrorsduringsegmentationofre1.evantanatomyiscritica1.旭曲I11ffiFdW自辂PhUn1.qntingtissue昶轴厢降三mgmentati8htfhR1¾M1T邦面明ngi献esoftwarecannotsegmenthumanboneeffective1.y.Forr出睢f1.ovedmedica1.imagebased3Dmode1.1.ing,forma1.izationandstandardizationoftheseproceduresis5.2TypesofsegmentationmethodsSevera1.methodshavebeeninvestigatedthatsegmenthumanbonefromCTinagcs.(iai9(221123)“闻WM。修e*臃期炭Etat1.Q想我刎的想必，帔险。*3tiw3re版M故M加他CinSCg1.nCntatiem.widobjectpixe1.ifthepixe1.intensityisgreaterthanaspecifichumanbonethresho1.dva1.ue,orrep1.acesitwithabackgroundpixe1.ifthepixe1.intensityis1.essthanaspecifichumanbonethresho1.dva1.ue.则阳出柚三tf三wMf三四治”咖加瞭碗6淋耐boundaries.Theinterna1.forcesaredefinedsuchthattheypreservetheshapesmoothnessofthemode1.Whi1.Ctheexterna1.forcesaredefinedbytheimagefeaturestodrivethemode1.towardthedesiredregion帆晒BySh嘟S懈肥慨r三8端ariesrobusmesssmoo/独dIJ三蝌厘JoIta喇'ft三布聊However,thereisa1.imitduetothedifficu1.tyofsegmentationattheweakobjectboundaryofathinbonewitha1.owintensityva1.uesimi1.artosofttissue.IXIQiNmK常&SSegmdh吆HOnPa廊由中CN1.M1.Wkg淋期神幽QayerS©玳部VUI帆心品似*hkd!给临三1.yusedformedica1.imagesegmentation,workandconsistsofacontractingpathandanexpansivepath,whichgivesittheU-shapedarchitecture.6Overa1.1.segmentationprocess6.1 Genera1.Theovera1.1.segmentationprocessconsistsofsevenstepsintota1.,asdescribedin6.2to6.8.Figure1showstheovera1.1.processf1.owofsegmentation,wherethenumbersinparenthesisrefertoc1.ausesofthisdocument.©SOIEC2024-A1.1.rightsreserved4Trainingphasef1.ow(CUum12)(CauseB)RunningPhaSe11owFigure1Overa1.1.processf1.owOfsegmentationThesoftwaredeve1.opershou1.dimp1.ementtheoptimizedsegmentationprocessformedica1.image-basedmode1.1.ing.Theconsiderationsoftheoptimizedsegmentationprocessformedica1.image-basedmode1.1.ingshou1.dbereferencedfromthisdocument.6.2 Step1:datapreparationTheobjectiveofthedatapreparationstageistotransir<n-therawdatasothatthesegmentationa1.gorithmS1.e伊西dpt即nchusBk>gforfgmiutationC1.ause7.丽制懵耀的由fition源郴c嘤tagesg皿由O班淋媚和es在於胸布曲f1.，施厢WPWqUa1.iB*4的痴nduceto1.earn.Detai1.edinformationisprovidedinC1.ause8.6.4 Step3:annotationTheobjectiveoftheannotationstageistomakethe1.abe1.1.edtrainingdatafiUrXhe1.earningofM1.D1.-6a5cS1.ep；4rfie1.eotiQn:ef>jikmk)<ttakidnetwat?kini0de1.isprovidedinC1.ause9.XBftvorko帮困9b施etai1.eHiWS&HiWetWO新用费修。OnStageistose1.ecttheoptima)segmentation6.6Step5:performanceeva1.uationTheobjectiveoftheperformanceeva1.uationstageistoca1.cu1.atetheagreementbetweentheresu1.tofBnR1.JnB1.thesegmentationtechniqueandtheground-truth1.abe1.Detai1.edinformationisprovidedin6.7 Step6:mode1.dep1.oymentandrunningmode1.environmentprintingDetai1.edinformationisachievingC1.ausemaximizedsegmentationperformanceinarea1.6.8 Step7:post-processingforsegmentationthesegmentationpreparationDetai1.edinformationisprovidedinC1.ause13.7.1 Genera1.7.2 beMedica1.image7.2.1 Genera1.Generatedscanners.NormaIIy1FromVariousreconstructedfromStoredorStandardizedproduced(eg.detectors.7.2.2 CTscanpixc1.body.-1024,3072,s1.icebecausemosttypica1.1.ystored12StorageDICOMperformat.Eachintensityrangeofva1.uegeneratescannerseriesofmu1.tip1.eradiographicprojectionsandportionusesthehumanreconstructionreso1.utionoftheWhenobtaininghumanqua1.itymedica1.examp1.e,InstandardizedtheSCanningbone,parameterCondidonsrecommendationused7.2.3 MRimageforthemagneticphysio1.ogica1.examinationofpatientMRIscannersuseimagingtechniquefie1.ds,magneticStandardizedMRIscanningconditionsorMRIscanningprofi1.esshou1.dbeconsidered.Theobjectiveofthedep1.oymentandrunningstageistoapp1.ytheoptima1.1.ytraineddeep1.earningnetworkwor1.dthe3Dmode1.1.ingsystemforprovidedthe12.Theobjectiveofthepost-processingforsegmentationstageistorefinetheincorrectregionafterapp1.yingrnA1©ISOIEC2024-A1.1.rightsreserved7 Datamethod.Theobjectiveofthedatapreparationstageistotansformtherawdatasothatthesegmentationa1.gorithmcanapp1.ied.Intheimageacquisitionphase,medica1.imagesareproducedfromdevices,suchasCT,MR1.PET/SPECT,US,OpUcahnedica1.imagedatatheyaredevicesshou1.dbe11raw,inasourceformatbyDICOM)formedica1.imageprocessing.Computedtomography(CT)isamedica1.imagingtechniquethatusesX-raystogenerate2Ds1.iceimagesoftheisThese2DimagesareCTimagesuseinthebitsfi1.epixe1.Thepixe1.hasitsowneachaccordingtothedegreeoftransmittedradiationthatpassesthroughthebody.ACTtakes2Dimages1.icescoveringaspecificthenofanimagebody.ThetechniquetoCTimageismain1.ydependentonpixe1.spacingandnumberofpixe1.s,andpartia1.1.ydependentonthes1.icethicknessofthe2Dimages.Forscanningbetterbone.Forimages,thecaseofCTorbitaItheminimumshou1.dbeofCTscanningconditionsisdefinedinAnnexA.Magneticresonanceimaging(MRI)equipmentismedica1.e1.ectrica1.equipmentwhichisgenera1.1.yintendedinVivoanatomyandresonanceprocessesathebody.ismedica1.strongmagnetictoformpicturesfie1.dgradientsandradiowavestocreateimagesofinterna1.patientanatomy.7.3Preparationsteps7.3.1 Genera1.Sincethedatapreparationstageconsistsoftwosteps,imageacquisitionandimagereconstruction,theessentia1.considerationsforeachstepshou1.dbereectcd.7.3.2 ImageacquisitionTheimageacquisitionstepcanbedefinedastheactionofacquiringasetofimagedatafromahardwaresource.Inordertocreateagoodqua1.itymedica1.image-based3Dmode1.for3Dprintin