Principal component analysis of FDG PET in amnestic MCI

Purpose. To evaluate the combined accuracy of episodic memory performance and 18F-FDG-PET in identifying patients with amnestic Mild Cognitive Impairment (aMCI) converting to Alzheimer’s disease (AD), aMCI non-converters, and controls. Methods. Thirty-three patients with aMCI and 15 controls (CTR) were followed-up for a mean of 21 months. Eleven patients developed AD (MCI/AD) and 22 remained aMCI (MCI/MCI). 18F-FDG-PET volumetric Regions of Interest underwent Principal Component Analysis (PCA), that identified 12 principal components (PC), expressed by coarse component scores (CCS). Discriminant analysis was performed using the significant PCs and episodic memory scores. Results. PCA highlighted relative hypometabolism in PC5, including bilateral posterior cingulate and left temporal pole, and in PC7, including the bilateral orbitofrontal cortex, both in MCI/MCI and MCI/AD versus CTR. PC5 itself plus PC12, including the left lateral frontal cortex (LFC: BAs 44,45,46,47), were significantly different between MCI/AD and MCI/MCI. By a 3-group discriminant analysis, CTR were more accurately identified by CCS+delayed recall score (100%), MCI/MCI by CCS+ either immediate or delayed recall scores (91%), while MCI/AD by CCS alone (82%). PET increased by 25% the correct allocations achieved by memory scores, while memory scores increased by 15% the correct allocations achieved by PET. Conclusion. Combining memory performance and 18F-FDG-PET yielded a higher accuracy than each single tool in identifying CTR and MCI/MCI. The PC containing bilateral posterior cingulate and left temporal pole was the hallmark of MCI/MCI patients, while the PC including the left LFC was the hallmark of conversion to AD.