Propofol anesthesia destabilizes neural dynamics across cortex

Anesthesia is ubiquitous in hospitals, yet a mechanistic understanding of how anesthetic drugs induce unconsciousness is lacking. A prominent hypothesis suggests that dynamic stability is critical to cortical function: awake brains are poised at a state that is sufficiently excitable for activity generation and propagation, yet controllable and stable. Measuring dynamic stability during the transition from consciousness to anesthetic unconsciousness could identify the neural mechanisms disrupted by anesthesia, and thus provide insight into the neural basis of consciousness. Related work suggests that anesthesia could either destabilize or excessively stabilize neural dynamics—the question remains unresolved. This is likely due to the lack of rigorous approaches to estimating stability in neural data and the paucity of studies performing this analysis using electrophysiology. Harnessing results from dynamical systems theory, we develop a novel and principled approach, Delayed Linear Analysis for Stability Estimation (DeLASE), to quantifying population-level dynamic stability that is scalable to large volumes of neural data. DeLASE constructs linear delay dynamical systems models of activity and discounts the impact of further back states to estimate stability. We validate DeLASE by verifying it can accurately infer changes in stability in simulated networks from partial observation of activity. We then apply DeLASE to local field potentials from four areas across the macaque cortex during transitions between awake and anesthetized states and find that neural dynamics are destabilized in anesthetic unconsciousness relative to wakefulness. This suggests some degree of stability is necessary for conscious processing. Accordingly, we find that stimulus-evoked trajectories diverge in anesthesia - whereas in wakefulness they quickly stabilize - presenting a neural mechanism for a lack of stimulus information integration in unconsciousness. Given the hypothesized changes of stability in neuropsychiatric conditions like depression, anxiety, and schizophrenia, the impact of our rigorous stability estimation approach could provide a quantitative link across multiple fields of study.