Brain State Entropy as a Marker of Injury Severity and Sedation Effects Following Traumatic Brain Injury

Traumatic brain injury disrupts large-scale brain networks. Dynamic functional connectivity captures time-varying network interactions from functional MRI (fMRI) and provides insights into the brain's dynamic patterns of integration and segregation. Here, we investigate dynamic functional connectivity in subacute moderate-severe traumatic brain injury patients (10 days to 6 weeks post-injury) and explore the relationships with blood and imaging biomarkers of injury and propofol sedation. We hypothesized that traumatic brain injury patients would show less complex brain state dynamics that would be associated with greater injury severity measured by white matter integrity and blood biomarkers. Sixty-five subacute traumatic brain injury patients and 48 healthy controls underwent structural and resting-state fMRI. Patients were followed-up at 6 and 12 months post-injury. Plasma concentrations of neurofilament-light chain, microtubule-associated protein, glial fibrillary acidic protein, ubiquitin carboxyl-terminal hydrolase L1, and serum S100 calcium-binding protein B were measured. Fractional anisotropy (FA), a measure of white matter integrity, was derived from diffusion-weighted imaging for a set of white matter tracts. Dynamic functional connectivity analysis was performed using a sliding-window approach. Correlations between time courses of 19 regions of interest representing the default mode network, bilateral frontaloparietal networks, and the salience network were calculated, and k-means clustering was applied to these connectivity matrices. Temporal characteristics of the resulting brain states, including fraction time, dwell time, number of transitions, and entropy of state transitions, were calculated. Four distinct brain states were identified. Brief periods of anticorrelation between key large-scale networks that support cognitive control were a dominant feature. Traumatic brain injury resulted in reduced temporal flexibility, less anticorrelated activity, and fewer transitions. Reduced entropy of state transitions was significantly associated with elevated blood-based biomarkers and reduced white matter integrity. Propofol sedation markedly reduced entropy. Dominance analysis identified glial fibrillary acidic protein, an astroglial plasma marker, as the strongest predictor of entropy. Preserved entropy during the subacute period was a significant predictor of 12-month functional outcomes. Entropy normalization at 6 months was associated with changes in glial fibrillary acidic protein, ubiquitin carboxyl-terminal hydrolase L1, and microtubule-associated protein over the same time period. We show that dynamic functional connectivity is disrupted following moderate-to-severe traumatic brain injury and these effects are exacerbated by sedation. The observed reductions in brain state entropy indicate a loss of network segregation and a shift toward less complex and more predictable brain activity, with important implications for prognosis.