Abstract
Alzheimer's disease (AD) is the most common neurodegenerative disorder among the elderly, and early detection is of great importance if new therapies are to be effectively administered. We have used multivariate data analysis (orthogonal partial least squares to latent structures (OPLS) analysis) to investigate whether the discrimination between AD and elderly healthy control subjects can be improved by adding magnetic resonance spectroscopy (MRS) measures to magnetic resonance imaging (MRI). In this study, 30 AD patients and 36 control subjects were included (mean (SD) age=77(5) and 77(5) years, MMSE=23(4) and 29(1) respectively). High resolution T1-weighted axial magnetic resonance images were obtained from each subject. Automated regional volume segmentation and cortical thickness measures were determined for the images. 1H MRS was acquired from the hippocampus and LCModel was used for metabolite quantification. Altogether, this yielded 54 different volumetric, cortical thickness and metabolite ratio variables which were used for multivariate analysis. All analyses were performed using seven-fold-cross-validation. Combining MRI and MRS measures resulted in a sensitivity of 97% and a specificity of 94% compared to using MRI or MRS measures alone (sensitivity: 93%, 76%, specificity: 86%, 83% respectively). Adding the MRS measures to the MRI measures more than doubled the positive likelihood ratio from 7 to 17. Adding MRS measures to a multivariate analysis of MRI measures resulted in significantly better classification than using MRI measures alone. The OPLS method shows strong potential for discriminating between Alzheimer's disease and controls.