Signature of consciousness in the dynamics of resting-state brain activity

Abstract

At rest, the brain is traversed by spontaneous functional connectivity patterns. Two hypotheses have been proposed for their origins: they may reflect a continuous stream of ongoing cognitive processes as well as random fluctuations shaped by a fixed anatomical connectivity matrix. Here we show that both sources contribute to the shaping of resting-state networks, yet with distinct contributions during consciousness and anesthesia. We measured dynamical functional connectivity with functional MRI during the resting state in awake and anesthetized monkeys. Under anesthesia, the more frequent functional connectivity patterns inherit the structure of anatomical connectivity, exhibit fewer small-world properties, and lack negative correlations. Conversely, wakefulness is characterized by the sequential exploration of a richer repertoire of functional configurations, often dissimilar to anatomical structure, and comprising positive and negative correlations among brain regions. These results reconcile theories of consciousness with observations of long-range correlation in the anesthetized brain and show that a rich functional dynamics might constitute a signature of consciousness, with potential clinical implications for the detection of awareness in anesthesia and brain-lesioned patients.

Keywords: anesthesia; consciousness; dynamics; functional connectivity; structural connectivity.

Fig. 1.

Stationary functional connectivity as a function of vigilance condition. (A) Average connectivity matrices for all vigilance conditions. (B–D) Brain renders displaying all significant connections (P < 0.001, FDR corrected) for wake (B), moderate (C), and deep (D) sedation conditions. Red lines represent positive connections between ROIs; blue lines represent negative connections. (E) Average positive z-values within each sedation conditions. In all plots, error bars represent 1 SEM. (F) Ratio of negative to positive z-values. Ratios are calculated for each scanning session, and averaged within each sedation condition. (G) Time course of the L1 norm of the matrix Z_c,s,w for w = 1–464, of a sample fMRI session (condition awake, session 14 of monkey J). Inserted matrices show whole brain connectivity patterns at different time points.

Fig. 2.

Dynamical connectivity and brain states for all vigilance conditions. (A) Seven brain states, obtained by unsupervised clustering of the Z_c,s,w matrix. Brain states are sorted according to their similarity to the structural connectivity matrix. (B) Structural matrix derived from the CoCoMac atlas of anatomical macaque cortical connectivity. Colors represents the four grades of connection intensity (black = 0; white = 1; blue = 2; and red = 3). (C) Brain renders displaying the 400 strongest links, for each brain state. Red lines represent positive connections between ROIs; blue lines represent negative connections. Brain render for the CoCoMac structural matrix displays all links with a value of 3, the maximum value of structural connectivity. (D) Probability distributions of brain states for all vigilance condition. Each bar represents the within-condition probability of occurrence of a state. Error bars stand for SEM. (E–G) Probability of occurrence of each brain state as a function of the similarity between functional and structural connectivity for awake (E), moderate (F), and deep (G) conditions. Each point in the figure corresponds to a brain state, characterized both by a similarity score and a probability of occurrence within each vigilance condition. Error bars show SEM. (H) Probability distribution of all z-values for brain states 1 (the least similar to structure brain state) and 7 (the most similar to structure brain state). (I and J) 2D probability distribution of all z-values for brain state 1 (I) and brain state 7 (J), as a function of the distance of every pair of ROIs. Distance is normalized by the largest distance within the Kotter and Wanke brain atlas. (K) Community decomposition for the Kotter and Wanke brain atlas, taken from Shen et al. (31). Four nonoverlapping communities are defined: frontopolar (community 1), fronto-temporal (community 2), fronto-parietal (community 3), and occipito-temporal (community 4). (L) R² value for the regressions between absolute correlation value between ROIs of every pair of communities and similarity score. Asterisks mark significant regression, at P < 0.05 (Bonferroni corrected for multiple comparisons). (M) Ratio between intercommunity z-values and intracommunity z-values for each brain state, as a function of the similarity score. (N) Small world index of all brain states, as a function of the similarity score. Each point represents the small world value of a given brain state. (O) Average life time of brain states for all sedation conditions. Error bars stand for SEM.