Quantitative EEG Analyses for Outcome Prediction in Comatose Patients After Cardiac Arrest

Conventional visual interpretation of EEG has several limitations, including inter- and intraobserver variability, a substantial review burden, and the need to reduce a continuously evolving spatiotemporal signal to categorical descriptions. Quantitative analysis of the EEG may help overcome some of these limitations by capturing signal characteristics and temporal dynamics that are difficult to summarize consistently in words alone. In this review, we briefly summarize the main approaches that have been used to support EEG-based outcome prediction in comatose patients after cardiac arrest. These range from explicit feature-based indices, such as the Cerebral Recovery Index, to machine-learning classifiers using larger EEG feature sets, and, more recently, to deep learning models trained directly on raw EEG. Importantly, these methods are relevant not only for the early identification of patients with a poor prognosis but also for identifying patients with a realistic chance of good neurologic recovery. Furthermore, qEEG increases objectivity, reducing dependence on categorical wording and supporting continuous bedside assessment of neurologic recovery. These approaches can reach sensitivities of 50% to 60% at specificities of > 99% for poor and > 95% for good neurologic outcome, respectively, particularly for early EEG (recorded < 24 hour after cardiac arrest). Within a multimodal prognostic framework, qEEG can serve as a valuable decision-support tool to assist final assessment by a trained electroencephalographer.