Preservation of anticorrelated brain networks predicts recovery after traumatic brain injury

Some patients with moderate to severe traumatic brain injury (TBI) make a full recovery, while others remain severely disabled. Accurate prognostication is important, because withdrawal of life-sustaining therapy based on perceived poor prognosis is the leading cause of death after TBI. Synchronized activity between brain regions, measurable with resting-state functional MRI (rs-fMRI), may underlie neurological recovery. However, which functional connections are critical for recovery, and whether functional connectivity measured shortly after brain injury predicts long-term recovery, is unknown. Here, we analyzed data from three prospective cohorts of patients with moderate or severe TBI (N = 116 patients; 134 controls) who underwent rs-fMRI shortly after injury. The strongest predictor of 6-mo functional outcomes in the Training Cohort (mean cross validation AUC 0.94) and independent Testing Cohort (AUC 0.78; P = 0.001) was functional connectivity between three pairs of brain regions from functionally distinct networks, two of which were anticorrelated. Results were robust to controlling for sedation (P = 0.02) and level of consciousness at time of MRI (P = 0.02). Finally, preserved anticorrelations improved the leave-one-out outcome prediction accuracy of an established prognostic score (AUC 0.90 vs. 0.80; P = 0.02). Preserved functional anticorrelations in acutely traumatized brains identify patients with the neurological substrate required for recovery. This biomarker can inform prognostic decisions in patients at high risk for death from withdrawal of life-sustaining therapy.