Dynamic assessment of head motion compensation for the HRRT

Motion correction in PET has become more important as system resolution has improved. Our previous simulation study based on measured subject motions showed that event-by-event motion correction has the best accuracy in static acquisitions. The purpose of this study was to evaluate the accuracy of both event-by-event and frame-based motion correction methods in simulated dynamic studies. Image reconstructions with various motion correction methods were carried out on simulated HRRT list-mode data that contained a wide range of motion based on actual motion data. The time-activity curves (TAC) of all regions of interest were fitted to a one-tissue compartment model to estimate the distribution volume (V_T) and the binding potential (BP_ND). In addition, we explored the figure-of-merit sum of squared difference (SSD) between the raw TAC and the fitted curve as a potential tool to assess the success of motion compensation in dynamic PET studies. This study showed that event-by-event motion correction method works reliably for a wide range of human head motion, generating V_T and BP_ND estimates that are comparable to the motion-free case. Frame-based methods introduce 5-10% bias in these parameters, as intra-frame motion gets larger than 5mm. For event-by-event motion correction, the SSD between the raw TAC and the fitted curve is consistent with the static motion-free reference for a wide range of motions. For the frame-based methods, the ratio of the SSD to the motion-free reference rises to 2-4 when the intra-frame motion exceeds 5mm, suggesting that this may be a useful measure of quality of motion correction.