Regional effects of an MR-based brain PET partial volume correction algorithm: a Zubal phantom study

Magnetic resonance (MR) based partial volume correction (PVC) compensates for the limited spatial resolution of positron emission tomography (PET) by using higher-resolution MR images to determine the fractional contributions of small anatomic structures to observed voxel activities. The efficacy of a PVC algorithm depends on correct estimation of imager resolution and accurate PET-MR image co-registration. The purpose of this study was to investigate the regional effects of these errors. Simulated PET and segmented MR data were generated using the Zubal MR brain phantom. Structures were classified as gray matter (GM), white matter (WM) or cerebral spinal fluid (CSF) to form a three-compartment segmented MR image. PET images were simulated by assigning known activities to each segmented MR tissue (10:3:1 GM:WM:CSF) and blurring (4 to 12 mm FWHM). With the above two datasets as inputs, the effects of resolution mismatch and misregistration on the corrected GM activity returned by a PVC algorithm were observed. Mean regional gray matter recovery (GMR) was calculated for each GM structure defined by the Zubal phantom. GMR values were then related to structure dimension and location characteristics. Regional GMR variations in the original simulated PET data were nearly eliminated after ideal PVC but resurfaced, although to a lesser extent, when resolution mismatch and misregistration were introduced. Regional GMR error was found to be related to the surface area to volume ratio of the region and the fraction of gray matter of adjacent regions.