AI EEG prognostication after cardiac arrest

DeepCRI: real-time EEG-based prognostication after cardiac arrest
Resuscitation · Added 27 July 2026

Accurate prediction of neurological outcome after cardiac arrest is essential for guiding intensive care decisions.

Study type: Other
Design: Deep learning prognostic model developed and validated on multicentre EEG data
Participants: 522 comatose post-arrest patients for development; internal validation 219 and external validation 167
Main finding: Internal validation: lock-in classification in 81.7%; poor outcome sensitivity 49.5% with 100.0% specificity; good outcome sensitivity 94.7%, specificity 81.9%. External validation: lock-in in 59.9%; poor outcome sensitivity 36.8%, specificity 98.4%; 18.3% (internal) remained grey zone.
Caveat: Performance degraded in external validation and residual EMG artefact contributed to false poor-outcome predictions; ICU-based in-hospital cohorts
Bottom line: Hypothesis-generating

Learning goal

Weigh how emerging EEG-based prognostic tools should inform neurological outcome prediction in comatose post-arrest patients.