Refraction-compensated motion tracking of unrestrained animals in PET

Stereo optical motion tracking has been shown to be a feasible and accurate way of measuring head pose in PET studies of minimally-restrained awake animals. It may be convenient and/or necessary to contain the animal within a transparent enclosure to perform such studies. However, the presence of transparent interfaces introduces refraction error in the motion tracking. Here, we extend previous work looking at the refraction error in individual target points to error in ensembles of target points representing a rigid marker for pose estimation. We used our previously validated simulation technique to look at the effect of refraction on absolute and relative pose measurements as a function of the marker size. We show that for sufficiently small markers (≤ 50 mm) the error in absolute pose is a shift equivalent to the triangulation error. For movements (relative pose measurements) the error depends on the shift in each pose. By simulating refraction effects in real rat head tracking data and using these data for motion correction, we demonstrate quantitative errors of ~10% in a motion-corrected image. In situations with more severe refraction conditions, the necessity to correct pose errors is likely to be greater.