Researchers applied a pretrained EEG embedding framework without retraining to three independent cohorts, including 575 cardiac arrest patients, 100 patients with subarachnoid hemorrhage, and 141 routine EEG recordings. EEG segments were mapped into a 128-dimensional space anchored by expert-derived reference brain states. Favorable outcomes or physiological recordings lay closer to healthy reference states, whereas poor outcomes and pathological recordings shifted toward pathological regions. Trajectory-derived geometric and temporal features discriminated outcomes after cardiac arrest with ROC-AUC 0.83 and after subarachnoid hemorrhage with ROC-AUC 0.76.
This EEG representation showed potentially useful outcome discrimination after cardiac arrest. The abstract does not establish its incremental clinical value or suitability for individual prognostic decisions.
Study limitation: The abstract does not specify outcome definitions, assessment timing, or uncertainty around discrimination estimates.
AI-generated from the published abstract with automated source checks. For research discovery; consult the original paper before applying findings.