Valvular & structural heart diseaseDiagnostic or prognostic study
Ensemble Deep Learning Algorithm for Structural Heart Disease Screening Using Electrocardiographic Images
- Diagnostic accuracy study
- development and external validation
- multicentre
- ensemble deep learning (XGBoost)
Authors' conclusion
We developed and validated PRESENT-SHD, an AI-ECG tool identifying a range of SHD using images of 12-lead ECGs, representing a robust, scalable, and accessible modality for automated SHD screening and risk stratification.
Study context
- Population
- Adults with ECG and TTE within 30 days (Yale New Haven Hospital (YNHH)), 4 US hospitals, ELSA-Brasil cohort; SHD = LVEF <40%, moderate-to-severe left-sided valvular disease or severe LV hypertrophy93,693 YNHH development patients
- Outcome
- Detection of SHD (composite) on TTE
Performance
As reported in the abstract
SHD detection, held-out YNHH test set
| Test or model | AUROC | Sensitivity | Specificity |
|---|---|---|---|
| PRESENT-SHD | 0.886 (95% CI: 0.877-894) | 90% | 66% |
Further findings
| Outcome | Values |
|---|---|
| Diagnostic performance at external hospital sites | PRESENT-SHD AUC (AUROC) 0.854 to 0.900 · Sensitivity 93% to 96% · Specificity 51% to 56% |
- Diagnostic performance at external hospital sites: 4 US hospitals
Diagnostic performance in ELSA-Brasil (population-based cohort)
- AUROC
- 0.853 (95% CI 0.811-0.897)
- Sensitivity
- 88%
- Specificity
- 62%
Performance across demographic subgroups, novel ECG formats and smartphone photographs of electrocardiograms (ECGs)
- Consistent performance
New-onset SHD/heart failure (HF) risk after a positive screen (clinical sites and UK Biobank)
- 2- to 4-fold higher risk, independent of demographics, comorbidities and competing risk of death
Original publication
Ensemble Deep Learning Algorithm for Structural Heart Disease Screening Using Electrocardiographic Images. J Am Coll Cardiol. Published 1 April 2025. doi:10.1016/j.jacc.2025.01.030
Summarised from the published abstract; every number is checked against it