From First-order to Higher-order Interactions: Enhanced Representation of Homotopic Functional Connectivity through Control of Intervening Variables

Abstract

The brain's complex functionality emerges from network interactions that go beyond dyadic connections, with higher-order interactions significantly contributing to this complexity. One method of capturing higher-order interactions is through traversing the brain network using random walks. The efficacy of these random walks depends on the defined mutual interactions between two brain entities. More precise capture of higher-order interactions enables a better reflection of the brain's intrinsic neurophysiological characteristics. One well-established neurophysiological concept is Homotopic Functional Connectivity (HoFC), which illustrates the synchronized spontaneous activity between corresponding regions in the brain's left and right hemispheres. We employ node2vec, a random walk node embedding approach, alongside resting-state fMRI from the Human Connectome Project (HCP) to obtain higher-order feature vectors. We assess the efficacy of different functional connectivity parameterizations using HoFC. The results indicates that the quality of capturing higher-order interactions largely depends on the statistical dependency measure between brain regions. Higher-order interactions defined by partial correlation, better reflects HoFC compare to other statistical associations. In this case of first-order interactions, tangent space embedding more effectively demonstrates HoFC. The findings validate HoFC and underscore the importance of functional connectivity construction method in capturing intrinsic characteristics of the human brain.