Acute cardiac care & resuscitation
A continuous convolutional neural network: very early EEG most predictive for poor neurological outcome in postanoxic coma
Authors' conclusion
The CNN seems valuable for poor neurological outcome prediction in postanoxic coma patients, especially for very early (6-14 h after CA) EEG. A reduced electrode configuration is equally sufficient as the full set.
Study context
- Population
- Postanoxic coma patients after cardiac arrest, EEG from start of recording up to 72 h
- Participants
- 366 patients
- Comparison
- 9-electrode vs 4-electrode convolutional neural network (CNN) model for EEG-based prediction
- Outcome
- Poor neurological outcome: best Cerebral Performance Category (CPC) 3-5 (good = CPC 1-2)
- Follow-up
- Within 6 months after CA
Performance
As reported in the abstract
Prediction of poor neurological outcome:
- Cumulative model
- All available hours
| Test or model | AUC | Sensitivity at 0 % false-positive rate (FPR) |
|---|---|---|
| 9-electrode model | 82 % | 71 % |
9-electrode model
AUC82 %
Sensitivity at 0 % false-positive rate (FPR)71 %
Further findings
Predictive performance over time after cardiac arrest (CA) · peak at 8 h after CA
- Highest predictive performance before 14 h after CA, peaking at 8 h: AUC
- 0.93
- Sensitivity at 0 % FPR at 8 h
- 0.81
Performance difference between 9- and 4-electrode model
- Performance differences over hours negligible between the 9 and 4-electrode model
Original publication
A continuous convolutional neural network: very early EEG most predictive for poor neurological outcome in postanoxic coma. Resuscitation. Published 24 December 2025. doi:10.
Summarised from the published abstract; every number is checked against it