251 Extending directed brain connectivity beyond the mean: in-variance causality

Abstract

Purpose Granger Causality (GC) approaches have recently been employed to estimate the directionality of the influence exerted by a brain region on another. Despite the fact that fluctuations in the BOLD signal at rest contain important information about the physiologic al processes that underlie neurovascular coupling, so far associations between brain signals have focused on central tendencies (i.e. mean or median) and have modeled this compounded variability as noise. A possible causative structure in the variability of brain activity remains completely unexplored. Methods and materials In this contribution, we develop a theoretical framework for simultaneous estimation of both in-mean and in-variance causality in complex networks. We validate our approach in synthetically–generated signals from complex networks of coupled nonlinear Kuramoto oscillators and employ it on fMRI Human Connectome Project (HCP) data in order to estimate of in-variance connectome of the human brain. Results Fig. 1 shows ROC curves obtained for both in-mean and in-variance causal network reconstruction in synthetic validation. Applying the framework to in vivo data, structured and distinct in-mean and in-variance causal connectomes emerge. Fig. 2 shows circular plots highlighting the top 1% connections belonging to the matrices derived from HCP data. Conclusion Our results serve as proof of principle for demonstrating the relevance of targeted experimental investigation about the origin of coupled fluctuations in variance in BOLD signals measured using fMRI.