Page 132 - Demo

130Chapter7Thetimedifferencesofthemethodscouldbeexplainedbecausethesurgicalguidescouldbeplacedwithinsecondsonthesubjects’craniaduetotheiruniquefitwhiletheHoloLensworkflowrequiredafewminutesforcalibrationandregistrationInadditiondemarcatingtheplanningpointsthroughthegapsintheguideswasfasterthanmovingapencilfreelyinthreedimensionsintheHoloLensworkflowWedeterminedthatanerrormarginof2mmisacceptablewhenapatient’splanis transferredsince thehandlingof thecraniotome willalwaysinduceasmallerrorandtheactualthicknessoftheosteotomiesisdifficulttopredict Withanaverageerror of19 mm and 23 mm for observers1 and 2 respectively the HoloLens transfermethodwasalmostaccurateenoughtomeettheclinicalthresholdof2mm TheoutlierswithrelativelylargeerrorshadthegreatesteffectsontheresultsFigure4showsthatthemedianerroroftheHoloLensmethodforboththeobserverswassmallerthanthemeanerror(16and18versus19and23respectively) ThismeansthatifoutliersandlargeinaccuratemeasurementscanbeeliminatedtheaverageerrorandthestandarddeviationoftheHoloLenstransfermethodwillfundamentallyimproveErrorslargerthan3mmwereparticularlyseenintheresultsofobserver2 Takingintoconsiderationthatobserver1systematicallyusedmoretimetotransfertheVSPonthecranialshapes(1025minutesversus65minutes)alogicalexplanationisthatobserver1wasmoremeticulousthanobserver 2 whichresultedin fewerplanningpoints withlarger errorsNeverthelessoutliersshouldbepreventedbecausetheycanhaveatremendousimpactonthesurgicaloutcome Thelinearregressionconfirmedthattheobserveristhelargestvariable ThereforeweadvisetheHoloLensusertobepreciseandtaketheirtimeto transfer VSP on thepatientIn futurestudiesmore observersshouldbeincludedtoinvestigateinter-observervariabilityWiththeARmethoddevelopedinthisstudynoexternaltrackingsystemswasneededbecauseonlytheintegratedcameraoftheHoloLenswasusedmakingforacompactandlow-costsurgicalguidancesystemOtherstudiesthatusedimagetrackingfortheHoloLensinvolvedrelativelylargemarkersthatlimitedthemanoeuvrabilityofthesurgeonorcoveredasubstantialpartofthesurgicalworkingarea2425Aftertestingseveralmarkersandsizes(unpublisheddata)werecommendusinga5x5cmmarkerwithmanyfeaturepointsbecausethisistheoptimalcompromisebetweenmarkersizeandaccuracyataworkingdistanceof