Clinical applications of EEG connectivity in acute brain injuries: A systematic review.

OBJECTIVE: While connectivity methods have been widely studied as predictors of recovery in chronic disorders of consciousness (DoC), evidence for EEG-based connectivity measures in the acute phase of acute brain injuries (ABI) is limited. This study systematically reviews the features and models used to derive prognostic connectivity markers. METHODS: A systematic literature search was conducted in accordance with PRISMA guidelines, using the keywords: 'connectivity', 'EEG', 'traumatic brain injury', 'cardiac arrest', 'stroke', 'acute brain injury', 'encephalitis', 'subarachnoid hemorrhage', 'intracranial hemorrhage', 'consciousness disorders', and 'coma'. RESULTS: Of 3107 screened studies, 36 were included. Fifteen studies demonstrated good performance, with AUC-ROC exceeding 80%. The best results were achieved using multimodal machine learning models integrating EEG with other quantitative EEG (qEEG) and clinical parameters. EEG connectivity and qEEG provide complementary information, with the alpha band most frequently identified as the discriminative frequency in connectivity analyses. The lack of standardized methodologies and external validation may limit generalizability. CONCLUSIONS: EEG connectivity is a promising tool for predicting outcomes and detecting consciousness in acute care. Multimodal models combining EEG, qEEG, and clinical data could enhance prediction accuracy. SIGNIFICANCE: This study highlights the need for further research with standardized methodologies and external validation strategies.