Human brain parcellation using time courses of instantaneous correlations

Recent functional neuroimaging studies have shown that the brain can be divided into several spatially separate, functionally distinct networks. The investigation of these networks has become an active field of research to understand the brain as a set of interacting subunits. Key to this understanding is the manner in which these network break down into smaller fundamental sub-regions, and how these sub-regions reflect brain function and/or structural markers such as cytoarchitectonic features. In this work we propose a novel top-down functional parcellation strategy, using time courses of instantaneous correlations to subdivide an initial region of interest (ROI) into subregions. We demonstrate that large-scale functional networks can be broken down into biologically meaningful subdivisions that largely follow cytoarchitectonically defined areas. We apply our Instantaneous Correlation Parcellation (ICP) strategy on high-quality resting-state FMRI data, and evaluate against an example system in the form of the thalamus. The thalamic parcellation was compared in detail to the Morel histological atlas. All parcellations were found to show a close correspondence with underlying histology, with Dice overlaps of up to 0.96.