Markerless motion tracking for motion-compensated clinical imaging

Abstract

Motion-compensated brain imaging can dramatically reduce image artifacts and degradation associated with head motion. However, it has been slow to enter routine clinical use, and remained largely a research tool. One possible reason for this is the lack of a practical motion tracking method for the clinical setting. Here we present the initial validation of a highly convenient markerless motion tracking method, previously developed for motion-compensated PET imaging of rats, for human head motion estimation in a mock imaging scenario. Sixteen reclining volunteers followed the projected path of a robotically controlled laser beam during a 2 min experiment in which their resulting head motion was simultaneously tracked at 30 Hz from the anterior and posterior using the markerless system and a marker-based system, respectively. Motion estimates were compared in a common coordinate system and generally showed excellent agreement, especially in early frames (RMSE of 1-3 mm). Larger discrepancies were associated with the unwanted detection of non-rigid features which could be prevented in future studies. These results are a promising first step in translating the markerless tracking system to clinical use.