In a previous study, we found that non-demented patients with Parkinson’s disease (PD) did not differ from normal
controls (NC) in average global PiB binding, but focal cortical PiB retention was observed in many PD subjects
and in all demented patients with PD. In this study, we tested whether the overall burden and spatial distribution
of amyloid deposits differed in non-demented PD compared to healthy controls.
Twenty non-demented PD patients (age 70±7 years) and 70 NC (aged 74±8 years) underwent PiB PET, highresolution
MR imaging, Freesurfer processing, and partial volume correction. Cortical PiB DVR values in ROI were
left-right averaged. A linear stepwise discriminant analysis initially identified a subset of 7 ROIs from a pool of 35, a linear combination of which significantly discriminated the two groups (p<0.0001; 63% of variance of the function accounted for by the group classification). A parallel logistic regression analysis produced a similar linear combination of ROI predictors, which discriminated the PD from NC (p<0.0001); the ROC analysis area under the
curve (AUC) was 0.976. Cross-validation was by the leave-one-out method (AUC=0.953; for 1000 bootstrap resamples
95% CI=0.92-0.98) and by 1000 random split-half re-samples (AUC CI=0.65-0.94).
The discriminant coefficients indicated higher DVR values in parahippocampal, post-central and rostral anterior
cingulate ROIs reflecting PD status, whereas higher DVR values in posterior cingulate, lateral occipital, temporal
pole and amygdala were indicative of NC status. Non-partial volume-corrected PET data yielded slightly weaker
results though still significant. We also found significant relations of discriminant scores to measures of motor
impairment in PD (Hoehn & Yahr, UPDRS). These results suggest that PD patients and NC subjects can be reliably
distinguished by different patterns of amyloid accumulation.
