Database-Driven Identification of Functional Modules in the Cerebral Cortex

Abstract

The organization of the cerebral cortex into distinct modules may be described along several dimensions, most importantly structure, connectivity and function. Functional neuroimaging provides a powerful tool for the localization of function, which allows testing hypotheses about structure-function relationships. This method is, however, intrinsically less well suited to delineate the organization of a particular brain region. While neuroimaging studies may thus test hypotheses about a functional differentiation between cortical modules, their potential for delineating those in a particular region of interest is limited. Identification of cortical modules by differences in whole-brain connectivity profiles derived from diffusion tensor imaging or resting state correlations have therefore raised much recent interest. As these approaches, however, do not carry task-related information, the functional relevance of the obtained parcellation have so far remained largely elusive.

The emergence of comprehensive databases for functional neuroimaging results provides a novel basis for delineating cortical modules by co-activation networks. Importantly, such approaches are data-driven in that they do not rely on a classification of tasks or paradigms, but merely rely on the spatial pattern of whole brain co-activation profiles. The key idea behind database-informed cortical parcellation is computing the whole-brain co-activation patterns of each individual voxel within a seed region, regardless of the ontological classification of the original experiments. Recording the co-activation likelihoods of all grey-matter voxels outside the region of interest then yields a functional co-activation matrix. This connectivity matrix may then be used to group the seed voxels in such manner, that voxels showing similar co-activation are clustered together and separated from those showing different co-activation profiles. Hereby functional modules may be identified in a data-driven fashion using task-based neuroimaging information. By assessing the functional characteristics and spatial response patterns of those experiments associated with the ensuing clusters, the derived parcellation may be directly related to network properties and task properties.