Joint representation of cortical folding, structural connectivity and functional networks

Studying the regularity and variability of anatomical, structural and functional architectures of human brains and their relationship has been a longstanding, challenging problem due to the intrinsic complexity and variability of the cerebral cortex. One major technical obstacle is the lack of a joint representation of cortical folding, structural connectivity and functional networks that can be robustly and accurately identified in individual brains. In this paper, we present a novel computational framework that integrates three lines of research efforts including cortical folding representation, structural connectivity based brain parcellation, and sparse dictionary learning derived functional networks into a joint representation of brain anatomy, connectivity and function in individual brains, such that regularity and variability of cortical folding, structural connectivity and functional networks can be quantitatively assessed and interpreted in a comprehensive neuroscientific context. The application of our framework on the Human Connectome Project (HCP) multimodal DTI/fMRI data suggested that structural connectivity based brain parcellations and sparse dictionary learning derived functional networks exhibited deeply rooted regularity across individuals, but cortical folding patterns are substantially more variable. Our work indicates that novel features considering the above findings could be designed in the future for brain image registration and normalization applications.