Impact of accurate motion-corrected statistical reconstruction on dynamic PET kinetic parameter estimation

With continuing improvements in spatial resolution of PET scanners, small subject movements become a significant source of resolution degradation. This work assesses the impact of an accurate motion correction (MC) method, modeling the 'motion-induced' interaction of lines-of-response (LORs) inside and outside the field-of-view, on the task of kinetic parametric estimation in high resolution dynamic PET. Following extensive prior phantom and simulation validations, we performed external tracking of motion using the Polaris camera on ten Parkinson's decease (PD) subjects scanned on the high resolution research tomograph (HRRT). The resulting time-activity curves (TACs) generated from dynamic frames were compared in terms of (i) increased binding potential (BP) values as well as (ii) reduced χ² variations in the TAC fitting task, due to hypothesized increased effective-resolution and accuracy achieved via accurate MC. The Wilcoxon non-parametric matched-pair signed-rank test was applied to the BP as well as χ² values before and after MC, and demonstrated significant changes for the overall striatal region as well as a number of individual striatal regions-of-interest (ROIs). For the BP values, the increases were shown to be more significant for the subset of subjects (6 out of 10) which exhibited maximum displacements larger than the resolution (2.5mm) of the HRRT. Additionally, the method was demonstrated as a possible approach to standard clinical imaging with no, or minimal, head restraint.