Structural analysis of the cerebral cortex using blind source separation

Blind Source Separation (BSS) methods have become ubiquitous, but their performance varies greatly depending on how well their assumptions are satisfied by the data. Cortical thickness and sulcal folding patterns are ideal datasets for BSS analysis because there is limited prior knowledge on how they are affected by brain development and pathologies of the central nervous system. However, to date there are no studies exploring these datasets with BSS methods. We propose a novel spatial BSS method based on the Second Order Blind Identification (SOBI) method, but tailored for data on the cerebral cortex. Simulations show our method outperforms the regular SOBI and popular FastICA methods. Experimental data reveal underlying patterns in cortical maps of curvature variance.