Neural Complexity and Prognosis: Predicting Recovery in Pediatric Epilepsy Using EEG Markers

Predicting patient functional outcomes is an indispensable part of clinical care in the Pediatric Intensive Care Unit (PICU), especially for children with epilepsy, a prominent neurological emergency. Electroencephalography (EEG) is a dynamic tool for assessing brain activity, with brain complexity and spectral power features emerging as predictors of consciousness recovery. We investigated whether patients’ EEG activity under anesthesia could predict their recovery, using data from 12 pediatric epilepsy patients (mean age: 11.0±2.2 years). Neural complexity, the intricacy of connectivity between brain regions, is heavily implicated in a patient’s capacity for consciousness. We hypothesize that neural complexity will be a stronger predictor of patient outcomes than spectral power and that higher complexity will be associated with better outcomes. EEG features were analyzed during sedated, baseline (non-sedated), and difference states. Recovery was assessed three months post-injury using the Glasgow Outcome Scale-Extended (GOS-E). The predictive performance of significant EEG markers was evaluated using logistic regression with leave-one-out cross-validation and permutation testing. Baseline EEG features showed minimal prognostic power, whereas sedation and difference states yielded high prognostic accuracy. In the sedated state, the complexity features rate entropy and Lopez-Ruiz-Mancini-Calbet Complexity (HC-LMC) predicted recovery, separating good and poor outcomes with 100% accuracy. These findings demonstrate that EEG markers of complexity can predict the recovery of consciousness in pediatric epilepsy patients under anesthesia. Therefore, EEG analysis could be an accessible, accurate, and powerful prognostic tool in clinical settings. Future research should explore these results in larger samples to validate the findings that rate entropy and HC-LMC are predictive of recovery. Further, these features should be studied in patients of different etiologies to analyze their potential as generalizable markers of consciousness.