Page 220 - Demo

218Chapter12Figure12ExampleofAIobjecttrackingAdeepneuralnetworkisusedtomatchthevirtualmodelofamotorcycle(whitelines)including the tagsonarealphysicalobjectofamotorcycle(https//wwwptccom/en/products/vuforia)TRACKINGAfterregistrationtrackingensuresareal-timelocationupdateofthepatientandinstrumentsTheHoloLensusesabuild-inmethodfortrackingandstableplacementsofhologramsThis‘spatialmapping’reliesonflatandnon-movingobjectsforoptimalperformanceHowever,intheoperationtheatrethemovementsofthesurgicalteamandinstrumentsaroundthepatientresultinaconstantly altering space whichhampers aproper working of this spatialmappingandintroducesdriftintheholograms44InChapter6anexternaltrackingsystemwasaddedtotheHoloLenswhichsignificantlyimproveditsaccuracyHowever,adownsideisthatanexternaltrackingsystemwasrequiredtotrackatleastthreeobjects(theHoloLensthepatientandaninstrument)andallobjectsneededtobeinthelineofsightoftheexternalcameraLater,duetosoftwareupdatesandmoreresearchwediscoveredthatQRmarkertrackingisabletoproducesimilaraccuracyresultswithintheworkingdistanceofasurgeonInadditionthereisalowerrisk thatobjectsareoutofsightbecauseQR trackinguses thebuilt-in(color)cameraoftheHoloLensandisthereforemorepracticalForthesereasonsQRmarkertrackingwillbepartoffurtherresearchSincethebestfittingtrackingmethoddependsontheclinicalapplicationitislikelythatfuturesolutionsforARguidancewillswitchtoothertrackingmethodsFortrackingobjectsinsidethepatient(suchasorgansinsidethepatient’s