Congenital heart diseaseDiagnostic or prognostic study
A Multitask Deep Learning Model for Pediatric Echocardiography Analysis
- Diagnostic accuracy study
- internal held-out test set
- external validation
Trained on 10 815 studies, validated on 1294 studies; >700 000 videos from >12 000 studies
Authors' conclusion
Our findings demonstrate the remarkable potential for multitask deep learning models to aid the interpretation of pediatric echocardiograms.
Study context
- Population
- Pediatric echocardiography studies at Stanford Medicine (2014–2021) and an external cohort at the Children's Hospital of Philadelphia (2024–2025)
- Comparison
- EchoAI-Peds vs adult-based echocardiography foundation models
- Outcome
- Macroaveraged AUROC for detection of 28 congenital heart defects, abnormalities, repairs and interventions
Performance
As reported in the abstract
Macroaveraged AUROC for 28 congenital heart defects, structural and functional abnormalities, repairs and interventions
| Test or model | AUROC, internal test set | AUROC, external test set |
|---|---|---|
| EchoCLIP | 0.58 (95% CI, 0.56-0.60) | 0.61 (95% CI, 0.59-0.62) |
| EchoPrime | 0.58 (95% CI, 0.57-0.61) | 0.60 (95% CI, 0.59-0.62) |
Further findings
Performance by patient age, sex and number of videos
- Robust performance across patient age, patient sex and studies with varying numbers of videos
Parts of this Brief are withheld: the automatic checks could not confirm them against the abstract.
Original publication
A Multitask Deep Learning Model for Pediatric Echocardiography Analysis. Circulation. Published 28 July 2026. doi:10.1161/circulationaha.126.080619
Summarised from the published abstract; every number is checked against it