Methodology for Positron Emission Tomography Imaging of Amyloid-­‐β Plaques: Cross-­‐Sectional and Longitudinal Validation

The aim of this work was to identify optimal methodological techniques for quantifying 11C-­‐Pittsburgh Compound B (PiB) PET imaging measures of Amyloid-­‐β protein deposition, a neuropathological hallmark of Alzheimer’s disease (AD), in both cross-­‐sectional and longitudinal studies. Simulated phantoms based on existing MR and PiB PET images were generated to characterize and validate the implementation of three partial volume correction (PVC) techniques that account for the limited spatial resolution of PET: the Meltzer, the Müller-­‐Gärtner, and the Region-­‐Based Voxel-­‐Wise methods. The impact of these partial volume correction methods on correlations between region-­‐matched antemortem PiB PET standardized uptake value ratios (SUVR) and postmortem measures of amyloid load was then examined in a unique cohort of 12 subjects. Results indicate that the impact of PVC is not only dependent on the technique utilized, but also highly specific to region placement and subject anatomy. While the Region-­‐Based Voxel-­‐ Wise method outperformed other PVC methods, no PVC method improved correlations between antemortem PiB PET SUVR and postmortem measures of amyloid load, likely due to the noise propagation properties of the techniques. Several PET image analysis methods were also evaluated to assess relative performance in measuring change of amyloid deposition over time: the Simplified Reference Tissue, the Logan graphical, and the SUVR methods. This evaluation was performed in a unique cohort of 48 healthy elderly control, mild cognitive impairment, and AD subjects imaged with PiB PET at baseline and follow-­‐up. The potential impact of non-­‐specific reference region selection, changes in relative PiB delivery between target and reference tissues, and the partial volume effect were examined. The SUVR measure over 40-­‐60 minutes post-­‐injection was found to be optimal in tracking longitudinal changes across diagnostic groups. Results also indicate that the partial volume effects from non-­‐specific binding in white matter can influence measured changes of amyloid load. In conclusion, PVC techniques did not significantly improve correlations between PiB PET and postmortem measures of amyloid load. However, the Region-­‐ Based Voxel-­‐Wise PVC method should be considered for use in longitudinal PiB PET studies.