Bias-Free Multi-dimensional MRI Analysis to Resolve Age-related Microstructural Differences in the Ex-Vivo Female Bonnet Macaque Brain

Motivation: Age-related changes in the brain involve multiple biological processes, thus requiring advanced analysis approaches of multi-variate MRI maps are necessary. Goal(s): We applied a support vector machine classifier to multivariate MRI maps in order to identify groups of metrics most predictive of age-related change. Approach: High-resolution MRI was acquired in seven female bonnet macaque post-mortem brain specimens ranging in age from 10-34 years. Voxel-wise binary classification was performed with all-metrics, diffusion-only, and relaxometry-only co-registered datasets. Results: Age classification using all-metrics dataset achieved high accuracy (>90%). The insula had significant age accuracy classification (59%) and were explained by R2* and RD maps. Impact: Multi-variate approaches, such as a support vector machine classifier are advantageous when assessing aging and show potential in being applied to neurodegenerative disease studies.