Quantification of amyloid deposition with partial volume correction

Objectives Positron emission tomography (PET) with tracers such as the Pittsburgh Compound B (PIB) is widely used to quantify Aβ deposition in vivo for the investigation of the Alzheimer’s disease (AD). Partial volume correction (PVC) techniques are usually applied to compensate to the errors caused by the limited spatial resolution of PET. Typical implementation of PVC adopts a two-tissue (CSF and brain tissue) or a three-tissue model (CSF, gray matter, white matter). However, Aβ deposition varies by region and these models are overly simplistic. In this study, we aim to investigate a more extensive PVC model and its impact on the quantification of PIB image analysis.

Methods Imaging data from 33 subjects were analyzed. Automated segmentation of the brain was performed based on MRI data using FreeSurfer v5.0. Registration of MRI and the PIB PET images were performed using in-house software. Regional time activity curves were extracted for each region with and without PVC using a regional-spread-function technique based on FreeSurfer regions. Binding potentials (BP) were evaluated for each region with and without PVC. Mean cortical binding potential (MCBP) was also calculated using both approaches and compared.

Results The R², slope and y-intercept for each region and MCBP between partial volume corrected and uncorrected BPs were: caudate (0.92, 1.64, 0.14); gyrus rectus (0.96, 1.98, 0.12); precuneus (0.97, 2.09, 0.12); prefrontal (0.99, 2.01, 0.14); occipital (0.71, 1.20, 0.27); rostral anterior cingulate (0.91, 1.67, 0.15); lateral temporal (0.97, 1.80, 0.07); MCBP (0.99, 1.99, 0.11).

Conclusions The variability in the slope and intercept for the linear fitting between corrected and uncorrected BPs for different regions demonstrated the heterogeneous impact of partial volume effects on different regions. The high correlation between corrected and uncorrected MCBP suggests that PVC may not have a large impact on the ranking of subjects in terms of their amyloid load.

Research Support U01AG032438 P01AG003991 P01AG026276 P50AG0568