Registration of Cortical Connectivity Matrices

Abstract

This paper presents a novel formulation of the comparison of the neuroanatomy of different individuals. We propose to take into consideration and compare cortical connectivities, i.e. connections between different areas of the cortical surface through neural axon fiber bundles, as inferred from diffusion magnetic resonance volumes. Matching of cortical connectivity is a novel concept that could potentially open the way to many applications and research interests that motivate this work, such as the possible link between cortical connectivity and the appearance of sulci during morphogenesis, the study of causal relationships between functional areas in neurofunctional studies, or the relationship between cortical connectivity and different kinds of populations or diseases. This novel approach still lacks good understanding and mathematical modeling due to poor knowledge of human axon fiber anatomy. Our approach is a conservative one and consists in minimizing a robust distance between cortical connectivity matrices, while at the same time drastically simplifying the problem by imposing matrix entry pairings to stem from realistic spatial configurations, restricted to lie in a finite dimensional deformation space. The whole process is embedded in a multiresolution framework to improve both robustness of the match and computational load.