Time-varying synergy/redundancy dominance in the human cerebral cortex

Abstract

Recent work has emphasized the ubiquity of higher-order interactions in brain function. These interactions can be characterized as being either redundancy or synergy-dominated by the heuristic O-information [1]. Though the O-information can be decomposed into local values to measure the synergy-redundancy dominance at each point in a time series [2] no such analysis of fMRI dynamics has been carried out. Here we analyze the moment-to-moment synergy and redundancy dominance of the fMRI BOLD signal during rest for 95 unrelated subjects. We present results from several interaction sizes. The whole brain is rarely synergy-dominated, with some subjects never experiencing a whole-brain synergistic moment. Randomly sampled subsets of many sizes reveal that subsets that are the most redundancy dominated on average exhibit both the most synergistic and most redundant time points. Exhaustive calculation of the optimally synergistic and optimally redundant triads further emphasizes this finding, with recurrent nodes frequently belonging to a single coherent functional system. We find that when a triad is momentarily synergistic, it is often split between two instantaneously co-fluctuating communities, but is collectively co-fluctuating when it is momentarily redundant. After optimizing for synergy and redundancy in subsets of size five to seventy-five, we show that this effect is consistent across interaction sizes. Additionally, we find notable temporal structure in all optimized redundant and synergistic subsets: higher order redundant and synergistic interactions change smoothly in time and recur more than expected by chance.