Characterizing the functional connectivity diversity of the insula cortex: Subregions, diversity curves and behavior

Abstract

The connectivity of the insula cortex is diverse. We present new models to characterize the resting-state connectional diversity of the human insula cortex and perform model selection using high-quality fMRI data from the Human Connectome Project. We first attempt to parcellate the insula into distinct subregions using traditional clustering methods, but find that the resulting subregions are not homogeneous and that the optimal number of subregions is substantially influenced by data smoothness. We then introduce the concept of a diversity curve, which we use to continuously parameterize the insula's Laplacian eigenmap with respect to streamlines propagated through the eigenmap's gradient field. To perform model selection, we compare the insula's diversity curve to benchmark diversity curves for: i) two distinct regions; ii) a continuum of gradual change; and, iii) an absence of any connectional diversity (i.e. homogenous region). Of the three benchmarks tested, we find that the insula's connectional diversity is most parsimoniously modeled as continuum of gradual change, from dorsal-posterior to ventral-anterior. We find that individuals who score high on measures of positive affect, self-efficacy, emotion recognition, motor dexterity and gustation show greater diversity within the anterior insula. Our findings are replicated using data from a second fMRI session. We conclude that the functional connectivity diversity of the insula can be characterized parsimoniously as a continuum, avoiding the vexed task of determining an optimal number of insula subregions, and that inter-individual variation in this continuum can explain significant variation in behavior.

Keywords: Behavior; Clustering; Diversity; Insula; Parcellation; Resting-state fMRI.