Arrhythmias & electrophysiology
AI-ECG Risk Stratification for Atrial Fibrillation: Real-World Performance and Explainability
Authors' conclusion
AI-ECG provides rapid, low-cost AF risk estimation from a single sinus rhythm ECG; however, its predictive performance is modest compared with clinical scores. At present, AI-ECG may complement, but not replace, traditional risk stratification.
Study context
- Population
- Patients aged ≥40 years who underwent 12-lead ECGs using an AI-enabled electrocardiograph (FCP-9900)patients enrolled
- Participants
- 665 patients
- Index test or model
- Machine-learning models using clinical and ECG-derived variables vs AI-ECG alone for atrial fibrillation (AF)
- Target condition
- atrial fibrillation (AF)
Methodological detailsinternal validation by stratified 10-fold cross-validation, external validation in an independent cohort, SHapley Additive exPlanations
- Internal validation by stratified 10-fold cross-validation
- External validation in an independent cohort
- SHapley Additive exPlanations
Performance
As reported in the abstract
| Test or model | AUC | Area under the receiver operating characteristic curve |
|---|---|---|
| Model 2 support vector machine | 1.00 | — not stated in the abstract |
| Model 2 adaptive boosting | 0.97-0.98 | — not stated in the abstract |
| Model 2 artificial neural network | 0.89-0.95 | — not stated in the abstract |
| AI-ECG alone | — not stated in the abstract | 0.64-0.69 |
AUC
Model 2 support vector machine1.00
Model 2 adaptive boosting0.97-0.98
Model 2 artificial neural network0.89-0.95
AI-ECG alone— not stated in the abstract
Area under the receiver operating characteristic curve
Model 2 support vector machine— not stated in the abstract
Model 2 adaptive boosting— not stated in the abstract
Model 2 artificial neural network— not stated in the abstract
AI-ECG alone0.64-0.69
— not stated in the abstract
Further findings
AF prevalence across AI-ECG risk categories
- AF prevalence increased across AI-ECG risk categories, with significantly higher odds in the mid-high and high groups vs low
Feature contributions (SHapley Additive exPlanations)
- CHA2DS2-VASc the most influential predictor; AI-ECG provided modest incremental value
Original publication
AI-ECG Risk Stratification for Atrial Fibrillation: Real-World Performance and Explainability. JACC Adv. Published 22 July 2026. doi:10.
Composed by AI from the published abstract and checked against it. How Briefs are made →