Close-range Feature-based Head Motion Tracking for MRI and PET-MRI

Abstract

Optical motion tracking systems are effective in measuring head motion during MRI and PET scans. However, most systems rely on tracking attached markers which can slip or move relative to the head. In this study, we aimed to validate a methodology which uses a stereo-optical camera system to track small feature patches on the forehead. This approach has the advantage that tracked features `native' to the skin can be tracked at very close range (<5cm), making it ideal for use inside an MR scanner bore. 15 volunteers were instructed to perform 6 degree of freedom head motions while simultaneously being tracked by two systems - our feature-based tracking system, and a ground-truth multi-view optical system relying on passive IR-reflective markers attached to the back of the head. Sub-millimeter agreement between the two systems was achieved when head motion was purely rigid. In the case of non-rigid movement of the skin with respect to the head, large spikes in the motion traces derived from the feature-based algorithm were observed. This experiment provides a valuable dataset to benchmark future improvements and optimizations of the feature-based tracking algorithm, such as techniques to handle non-rigid motion.