Automatic statistical identification of neuroanatomical abnormalities between different populations

Abstract

We present a completely automatic method to identify abnormal anatomical configurations of the brain resulting from various pathologies. The statistical framework developed here is applied to identify regions that significant differ from normal anatomy in two groups of patients, namely subjects who subsequently converted to Alzheimer’s Disease (AD) and subjects with mild AD. The regions identified are consistent with post-mortem pathological findings in AD.