A systematic performance evaluation of head motion correction techniques for3 commercial PET scanners using a reproducible experimental acquisition protocol.

Abstract

Subject's motion during brain PET scan degrades spatial resolution and quantification of PET images. To suppress these effects, rigid-body motion correction systems have been installed in commercial PET scanners. In this study, we systematically compare the accuracy of motion correction among 3 commercial PET scanners using a reproducible experimental acquisition protocol. A cylindrical phantom with two 22Na point sources was placed on a customized base to enable two types of motion, 5° yaw and 15° pitch rotations. Repetitive PET scans (5min × 5 times) were performed at rest and under 2 motion conditions using 3 clinical PET scanners: the Eminence STARGATE G/L PET/CT (STARGATE) (Shimadzu Corp.), the SET-3000 B/X PET (SET-3000) (Shimadzu Corp.), and the Biograph mMR PET/MR (mMR) (Siemens Healthcare) systems. For STARGATE and SET-3000, the Polaris Vicra (Northern Digital Inc.) optical tracking system was used for frame-by-frame motion correction. For Biograph mMR, sequential MR images were simultaneously acquired with PET and used for LOR-based motion correction. All PET images were reconstructed by FBP algorithm with 1 × 1mm pixel size. To evaluate the accuracy of motion correction, FWHMs and spherical ROI values were analyzed. The percent differences (%diff) in averaged FWHMs of point sources at 4cm off-center between motion-corrected and static images were 0.77 ± 0.16 (STARGATE), 2.4 ± 0.34 (SET-3000), and 11 ± 1.0% (mMR) for a 5° yaw and 2.3 ± 0.37 (STARGATE) and 1.1 ± 0.60 (SET-3000) for a 15° pitch respectively. The averaged %diff between ROI values of motion-corrected images and static images were less than 2.0% for all conditions. In this study, we proposed a reproducible experimental framework to allow the systematic validation and comparison of multiple motion tracking and correction methodologies among different PET/CT and PET/MR commercial systems. Our proposed validation platform may be useful for future studies evaluating state-of-the-art motion correction strategies in clinical PET imaging.