Prediction of conversion from amnestic MCI to Alzheimer's disease using principal component analysis of FDG PET

Objectives “Study on diagnosis of early Alzheimer's disease -Japan (SEAD-J)” is a prospective multi-center cohort study that aims to establish a scientific evidence for usefulness of 18F-FDG-PET in early diagnosis of Alzheimer disease (AD) at the stage of amnestic mild cognitive impairment (aMCI).The purpose of this study was to evaluate scores by principal component analysis (PCA) of the FDG PET image for predicting conversion from aMCI to AD.

Methods Subjects were 88 patients with aMCI who were followed for up to three years. Forty three out of 88 patients developed AD. FDG PET images at the baseline were processed with 3-dimensional stereotactic surface projections (3D-SSP), and their values were normalized with individual global mean. Seventy eight ROIs of Brodmann’s areas (BA) were applied to 3D-SSP FDG PET using automated ROI software, SEE2 (Japan MediPhysics). ROI values underwent PCA that identified 14 principal components (PCs), expressed by PC scores. Discriminant analysis was performed using significant PCs.

Results PCA detected 14 PCs accounting for 84% of the total variance. The scores of PC6 and PC7 showed significant difference between the converters and the non-converters (p<0.005). PC6 included the temporoparietal association area, the posterior cingulate, and the precunes (BA 7L, 39L, 40L, 21L, 22L, 23L, 31L). PC7 included temporoparietal association area and the retrosplenial area (BA 40R, 29L, 39R, 20R, 21R). Two-group (converter/non-converter) discriminate analysis represented that the diagnostic sensitivity, specificity, and accuracy were 0.73, 0.74, and 0.74, respectively. Cases having lower PC6 and higher PC7 scores showed more typical AD metabolic pattern. Cases, even though converters, having higher PC6 and lower PC7 scores demonstrated the glucose metabolic patterns were unlike AD.

Conclusions The results suggest the PCA analysis of FDG PET is helpful to predict conversion from aMCI to AD. PCA is also a useful tool to evaluate glucose metabolic variation in the patients with aMCI