Cardiac arrest & resuscitation
Artificial intelligence for predicting shockable rhythm during cardiopulmonary resuscitation: In-hospital setting
Authors' conclusion
This study was the first to accurately predict shockable rhythms during compression using an AI model trained with actual patient ECGs recorded during resuscitation.
Study context
- Population
- Adults with cardiac arrest presenting to or occurring in the emergency department (ED), September 2021–February 2024
- Participants
- 172 adults · 1,889 electrocardiograms (ECGs)
- Outcome
- AUROC for shockable rhythm prediction from ECGs with compression artifact (5 s before each rhythm check)
Methodological detailsArtificial intelligence (AI) model development, prospectively collected database, convolutional neural network, per-patient split into training, validation and testing sets, Grad-CAM explainability
- AI model development
- Prospectively collected database
- Convolutional neural network
- Per-patient split into training, validation and testing sets
- Grad-CAM explainability
Main finding
As reported in the abstract
Shockable rhythm prediction from ECGs with compression artifacts (testing set)
AUROC for shockable rhythm prediction
0.8672
95% CI 0.8161–0.9122
Observational data: associations, not effects.
Further findings
| Outcome | Manual compression | Mechanical compression |
|---|---|---|
| Shockable rhythm prediction, by compression type | ||
| AUROC | 0.8771 (95% CI: 0.8054-0.9408) | 0.8466 (95% CI: 0.7630-0.9138) |
Original publication
Artificial intelligence for predicting shockable rhythm during cardiopulmonary resuscitation: In-hospital setting. Resuscitation. Published 17 July 2024. doi:10.
Composed by AI from the published abstract and checked against it. How Briefs are made →