Automatically learning cortical folding patterns

A data-driven technique is presented for automatically learn-ing cortical folding patterns from MR brain images of dif-ferent subjects. Cortical patterns are represented in terms of generic scale-invariant image features. Learning automati-cally identifies a set of features that occur with statistical reg-ularity in appearance and geometry from a large set of MR volume renderings, based on a predescribed anatomical re-gion of interest. A filtering technique is presented for distin-guishing between valid cortical features and those likely to arise from incorrect correspondences, based on feature geom-etry. Expert validation of 100 feature instances shows that 77 % correctly identify the same underlying cortical structure in different brains despite high inter-subject variability, and filtering improves the ability to identify the most meaningful patterns. 1