Multimodal imaging in mild cognitive impairment: Metabolism, morphometry and diffusion of the temporal-parietal memory network

This study compared sensitivity of FDG-PET, MR morphometry, and diffusion tensor imaging (DTI) derived fractional anisotropy (FA) measures to diagnosis and memory function in mild cognitive impairment (MCI). Patients (n=44) and normal controls (NC, n=22) underwent FDG-PET and MRI scanning yielding measures of metabolism, morphometry and FA in nine temporal and parietal areas affected by Alzheimer's disease and involved in the episodic memory network. Patients also underwent memory testing (RAVLT). Logistic regression analysis yielded 100% diagnostic accuracy when all methods and ROIs were combined, but none of the variables then served as unique predictors. Within separate ROIs, diagnostic accuracy for the methods combined ranged from 65.6% (parahippocampal gyrus) to 73.4 (inferior parietal cortex). Morphometry predicted diagnostic group for most ROIs. PET and FA did not uniquely predict group, but a trend was seen for the precuneus metabolism. For the MCI group, stepwise regression analyses predicting memory scores were performed with the same methods and ROIs. Hippocampal volume and FA of the retrosplenial WM predicted learning, and hippocampal metabolism and parahippocampal cortical thickness predicted 5 minute recall. No variable predicted 30 minute recall independently of learning. In conclusion, higher diagnostic accuracy was achieved when multiple methods and ROIs were combined, but morphometry showed superior diagnostic sensitivity. Metabolism, morphometry and FA all uniquely explained memory performance, making a multi-modal approach superior. Memory variation in MCI is likely related to conversion risk, and the results indicate potential for improved predictive power by the use of multimodal imaging.