Garment compensation in marker-less motion tracking for medical imaging

Abstract

To mitigate image artifacts resulting from patient motion, detection and tracking of motion is necessary. Existing techniques to track patient motion are either image-derived or utilize systems external to the imaging modality such as visual tracking systems (VTS). VTS methods adopted in clinical research often use optical stereo cameras to track retro-reflective markers placed on the patient to provide a measure of motion for a target structure. However, these systems have important limitations including: high equipment costs, hindrance to patient workflow, marker slippage, and frequent calibration and maintenance. An alternative is to use marker-less VTS, which employ small low-cost depth-sensing optical cameras that can be placed inside the imaging bore and track 1000's of points without markers using disturbances in reflected light patterns. Despite the advantages of size and not needing markers, marker-less VTS cannot directly see the patient's body surface, only the surface of the garments covering their body. In lieu of asking the patients to disrobe, we have developed a method to enable tracking of the body surface of patients despite being fully clothed. Our method is based on physical simulation of Newtonian forces acting on the clothing's fabric to accurately predict how a garment will deform from a given a body shape and motion. This work leverages recent advances in the computer graphics field that now enable robust methods for capturing and modeling the shape, motion, elasticity, and other properties of garments. In simulations, we show the ability to track points on the surface of the torso despite loose and wrinkled garments and our findings confirm that garment compensation will be a necessary and important part of the overall motion tracking methods which utilize marker-less VTS.