Comparison of 22 partial volume correction methods for amyloid PET imaging with 11C-PiB

1732 Objectives: Various strategies have been proposed to correct partial volume effects (PVE) in brain PET images. The present study aimed to determine the correction abilities of 22 partial volume correction (PVC) methods for amyloid 11C-PiB PET. Methods: We retrospectively analyzed 11C-PiB PET/CT and 3T MRI data derived from nine healthy individuals (controls) and nine patients with Alzheimer disease (AD). Magnetic resonance images were parcellated using FreeSurfer and merged into 20 anatomical regions. We compared PVC by geometric transfer matrix (GTM), Labbe (LAB), Richardson-Lucy (RL), Van-Cittert (VC), region-based voxel-wise correction (RBV), multi target correction (MTC), iterative Yang (IY), Muller-Gartner (MG) and 14 combinations of these methods using PETPVC toolbox. Both uncorrected and PVC images were normalized with averaged ROI values of cerebellar grey matter to calculate standardized uptake value ratios (SUVR). The inter-subject coefficient of variation (COV) and intra-regional variability (COVr) were calculated based on data from all patients. Results: The SUVR was almost identical in white matter regions with and without PVC. The SUVR and COV after PVC in images of grey matter regions increased by 8% and 22%, respectively, among the controls, and by 43% and 69%, respectively, among patients with AD. The RBV increased SUVR by 45% and decreased COV by 9% in these patients. The COVr values were more prominent when PVC methods included RL. Conclusions: The SUVR and image quality obviously differed according to the PVC Methods: Thus, PVC should be cautiously applied to amyloid 11C-PiB PET images. Our results suggest that RBV or RBV+LAB was the most reliable means of qualifyingAβ plaque deposition and thus diagnosing AD. The present study provides useful information about how to select PVC methods for amyloid 11C-PiB PET images.