A comparison of five partial volume correction methods for Tau and Amyloid PET imaging with [F]THK5351 and [C]PIB.

Purpose: To suppress partial volume effect (PVE) in brain PET, there have been many algorithms proposed. However, each methodology has different property due to its assumption and algorithms. Our aim of this study was to investigate the difference among partial volume correction (PVC) method for tau and amyloid PET study.

Methods: We investigated two of the most commonly used PVC methods, Müller-Gärtner (MG) and geometric transfer matrix (GTM) and also other three methods for clinical tau and amyloid PET imaging. One healthy control (HC) and one Alzheimer's disease (AD) PET studies of both [F]THK5351 and [C]PIB were performed using a Eminence STARGATE scanner (Shimadzu Inc., Kyoto, Japan). All PET images were corrected for PVE by MG, GTM, Labbé (LABBE), Regional voxel-based (RBV), and Iterative Yang (IY) methods, with segmented or parcellated anatomical information processed by FreeSurfer, derived from individual MR images. PVC results of 5 algorithms were compared with the uncorrected data.

Results: In regions of high uptake of [F]THK5351 and [C]PIB, different PVCs demonstrated different SUVRs. The degree of difference between PVE uncorrected and corrected depends on not only PVC algorithm but also type of tracer and subject condition.

Conclusion: Presented PVC methods are straight-forward to implement but the corrected images require careful interpretation as different methods result in different levels of recovery.