Page 202 - Demo

200Chapter12SOFTWAREUSABILITYFuturedevelopmentscouldextend theimpactof this workflowOnesuggestionisthat we should aim tomake(new)research datamore easily accessiblein clinicForexamplethecranialdevelopmentofalargegroupofpatientsandcontrolsthatwasstudiedin Chapter2and Chapter10resultedinmanygraphsand3Dgrowthmapsbutsimilarcapabilitieshavenot yetbeenimplementedinthesoftwarethatweuseduringclinicalconsultationFurtherautomationofoursoftwareshouldmakeitpossibletoinstantlycomparea3Dphotoofanindividualpatienttotheaveragegrowthmapsorcompareaskullcircumferencetothereferencevaluesinourdataset ThiswouldmakeclinicalconsultationevenmoreinsightfulandmaketheevaluationofindividualpatientsmorestraightforwardCurrentlytheclinicalworkflowfortheevaluationofcranialshapesstillcontainsmanymanualanduser-dependentstepssuchastheplacementoflandmarksneededforpre-alignmentAlthoughweaimedtoautomatethisstepinpreviouswork we werenot successful with our attempts Gladfully othermethodsbased ontexture3Dgeometryordeeplearningdemonstratethepossibilitiesofautomaticlandmarkplacementon3Dphotos12Thiswillfurtherautomateandsimplifytheworkflowwhichmakesitmoreoperableforotherspecialists(egphysicians) ThisimprovementwillallowotherDutchandinternationalcenterstouseoursoftwareeasilyAnotherveryimportantadvantageisthatwhenthesamemethodsareusedsurgery-relatedoutcomemeasurementsbecomemoreeasilyexchangeableandcomparableThiswillleadtoanobjectivecomparisonoftreatmentresultsandwillhelpachievethebestpossibletreatmentmethodforindividualpatientsNEWMETHODSFORAUTOMATICIDENTIFICATIONANDEVALUATIONTomakedetectionandevaluationofcranialdeformitiesclearer,anewmethodwasdevelopedtoeasilydescribethe3Dshapeoftheskull(Chapter3)Withprincipalcomponentanalysis(PCA)thismethodcanautomaticallyisolatethemostimportantshapevariationsinalargesetof3DphotosPCAusesonlyafewoutcomemeasurementstodescribethecomplete3DmorphologyofthecranialshapesothismethodhasthepotentialtoimprovethecurrentlyusedclinicaldiagnosticmethodsforcranialabnormalitiesThePCAmodelcreatedfromandbased on normal cranial shapes was able to identify all 40 craniosynostosis patientsAtthatmomentthedatasetthatwasusedfortrainingandtestingwassmalltheagerangebetweenpatientsandreferenceswasbetween3and6monthsand