Coronary disease & ACSDiagnostic or prognostic study
Deep learning electrocardiogram model for risk stratification of coronary revascularization need in the emergency department
- Diagnostic accuracy study
- external validation
- Trained on 144 691 US ED visits
Authors' conclusion
The model was able to detect revascularization and Type 1 MI with competitive performance, suggesting a potential role to complement current clinical assessment.
Study context
- Population
- ED visits; US cohort (training/test) and European external validation cohorttest n = 35,995
- Comparison
- Deep learning electrocardiogram (ECG) model vs clinician ECG interpretation and cardiac troponin T
- Outcome
- AUROC for coronary revascularization
Performance
As reported in the abstract
Revascularization, test cohort
| Test or model | AUC |
|---|---|
| ECG model | 0.91 (95% CI 0.91-0.91) |
| Clinician ECG interpretation | 0.65 (95% CI 0.54-0.76) |
| Troponin T (TnT) | 0.71 |
Further findings
| Outcome | ECG model | Clinician ECG interpretation | TnT |
|---|---|---|---|
| External validation (Europe, n = 18 673) | |||
| AUC revascularization | 0.81 (95% CI 0.81-0.82) | 0.70 (95% CI 0.57-0.83) | Not reported in the abstract |
| AUC Type 1 MI | 0.85 (95% CI 0.84-0.85) | 0.74 (95% CI 0.56-0.92) | Not reported in the abstract |
| AUC revascularization (hs-TnT) | Not reported in the abstract | Not reported in the abstract | 0.85 |
| AUC Type 1 MI (hs-TnT) | Not reported in the abstract | Not reported in the abstract | 0.87 |
External validation (Europe, n = 18 673)
- ECG model had higher specificity but lower sensitivity vs hs-TnT
Original publication
Deep learning electrocardiogram model for risk stratification of coronary revascularization need in the emergency department. Eur Heart J. Published 29 March 2025. doi:10.1093/eurheartj/ehaf254
Summarised from the published abstract; every number is checked against it