Evaluation of the Benefit of Partial Volume Correction for High Resolution PET Scanners

Abstract

The benefit of partial volume correction (PVC) within and post image reconstruction for high resolution PET scanners was examined. Listmode data from a realistic simulation of the HRRT that represented radioactivity and matter distributions for amyloid negative, amyloid positive and FDG scans in the head were conducted and reconstructed using the HRRT community software. PVC methods of Lucy-Richardson deconvolution and Rousset were applied both within and post reconstruction. Hammers’ atlas brain regions that were predominantly both grey and white matter were applied to the images and the ground truth. For all three simulated scans (FDG, amyloid negative and amyloid positive) there was very little benefit of Lucy-Richardson deconvolution both within and post reconstruction. For Rousset PVC, both within and post image reconstruction, there was a reduction on average in the bias. However, a large region-specific bias was observed which was larger for the small regions and when PVC was applied within image reconstruction. This work casts doubt on the benefit of PVC for high resolution PET scanners with further work needed to evaluate whether PVC is of value.