Optimized statistical parametric mapping for partial-volume-corrected amyloid positron emission tomography in patients with Alzheimer’s disease and Lewy body dementia

We present an optimized voxelwise statistical parametric mapping (SPM) of partial-volume (PV)-corrected positron emission tomography (PET) of ¹¹C Pittsburgh Compound B (PiB), incorporating the anatomical precision of magnetic resonance image (MRI) and amyloid β (Aβ) burden-specificity of PiB PET. First, we applied region-based partial-volume correction (PVC), termed the geometric transfer matrix (GTM) method, to PiB PET, creating MRI-based lobar parcels filled with mean PiB uptakes. Then, we conducted a voxelwise PVC by multiplying the original PET by the ratio of a GTM-based PV-corrected PET to a 6-mm-smoothed PV-corrected PET. Finally, we conducted spatial normalizations of the PV-corrected PETs onto the study-specific template. As such, we increased the accuracy of the SPM normalization and the tissue specificity of SPM results. Moreover, lobar smoothing (instead of whole-brain smoothing) was applied to increase the signal-to-noise ratio in the image without degrading the tissue specificity. Thereby, we could optimize a voxelwise group comparison between subjects with high and normal Aβ burdens (from 10 patients with Alzheimer’s disease, 30 patients with Lewy body dementia, and 9 normal controls). Our SPM framework outperformed than the conventional one in terms of the accuracy of the spatial normalization (85% of maximum likelihood tissue classification volume) and the tissue specificity (larger gray matter, and smaller cerebrospinal fluid volume fraction from the SPM results). Our SPM framework optimized the SPM of a PV-corrected Aβ PET in terms of anatomical precision, normalization accuracy, and tissue specificity, resulting in better detection and localization of Aβ burdens in patients with Alzheimer’s disease and Lewy body dementia.