Graph Representation of Brain Images for Classification and Correlation Discovery

Correlations between the structure of the brain and aspects of personality, health, and well-being are known to exist. For example, the lateral ventricles are known to enlarge due to Alzheimer’s Disease. Since the advent of MR imaging, it has been possible to examine the structure of the brain in a living individual. We present a method of producing a graph-based representation of the shape of the brain, discovering correlations using this representation, and classifying scans based on various criteria including level of education, age, gender, socioeconomic status, and level of impairment due to Alzheimer’s Disease. Correlations are verified by the ability to accurately classify test data, and displayed as an overlay over MR images.