Relationship of critical dynamics, functional connectivity, and states of consciousness in large-scale human brain networks

Abstract

Recent modeling and empirical studies support the hypothesis that large-scale brain networks function near a critical state. Similar functional connectivity patterns derived from resting state empirical data and brain network models at criticality provide further support. However, despite the strong implication of a relationship, there has been no principled explanation of how criticality shapes the characteristic functional connectivity in large-scale brain networks. Here, we hypothesized that the network science concept of partial phase locking is the underlying mechanism of optimal functional connectivity in the resting state. We further hypothesized that the characteristic connectivity of the critical state provides a theoretical boundary to quantify how far pharmacologically or pathologically perturbed brain connectivity deviates from its critical state, which could enable the differentiation of various states of consciousness with a theory-based metric. To test the hypothesis, we used a neuroanatomically informed brain network model with the resulting source signals projected to electroencephalogram (EEG)-like sensor signals with a forward model. Phase lag entropy (PLE), a measure of phase relation diversity, was estimated and the topography of PLE was analyzed. To measure the distance from criticality, the PLE topography at a critical state was compared with those of the EEG data from baseline consciousness, isoflurane anesthesia, ketamine anesthesia, vegetative state/unresponsive wakefulness syndrome, and minimally conscious state. We demonstrate that the partial phase locking at criticality shapes the functional connectivity and asymmetric anterior-posterior PLE topography, with low (high) PLE for high (low) degree nodes. The topographical similarity and the strength of PLE differentiates various pharmacologic and pathologic states of consciousness. Moreover, this model-based EEG network analysis provides a novel metric to quantify how far a pharmacologically or pathologically perturbed brain network is away from critical state, rather than merely determining whether it is in a critical or non-critical state.