Evaluation of voxel-based methods for the statistical analysis of PIB PET amyloid imaging studies in Alzheimer's disease

Deposition of amyloid plaques is believed to be a central event in the development of Alzheimer's disease (AD). The present study was undertaken to evaluate statistical methods for the assessment of group differences in retention of an amyloid imaging agent, PIB, throughout the brain and to compare these results to FDG studies of glucose metabolism performed in the same subjects on the same day. PET studies were performed in 10 mild to moderate AD and 11 control subjects. Parametric images of PIB retention (over 90 min post-injection) were generated using the Logan graphical analysis with cerebellar (CER, reference region) data as input. FDG parametric images were created by summing the uptake over 40-60 min post-injection and normalizing that to the CER to give a standardized uptake value ratio. Data were compared using parametric (SPM) and non-parametric (SnPM) statistical methods with familywise error (FWE) and false discovery rate (FDR) corrections. PIB results were consistent with previous regional results as AD subjects showed highly significant retention in frontal, parietal, temporal, and posterior cingulate cortices (FDR-corrected p<1.4e-10). FDG results showed regions of marginally significant decreases in uptake in AD subjects (frontal, parietal, temporal, posterior cingulate cortices: FDR-corrected p<0.1) consistent with previous studies. Relative to FDG, the PIB analyses were of greater statistical significance and larger spatial extent. Additionally, the PIB analyses retained significance after both FWE and FDR corrections. These results indicate that voxel-based methods will be useful for future larger longitudinal studies of amyloid deposition that could improve AD diagnosis and anti-amyloid therapy assessment.