A new artificial intelligence model developed by researchers at Scripps Research is showing promise in detecting and predicting heart disease using far less labeled data than conventional systems, a development that could matter for hospitals and clinics with limited resources.
Published in The Lancet Digital Health on September 1, 2026, the model, called ECG-CLIP, was trained on more than 1.7 million electrocardiograms from more than 540,000 people, paired with clinicians’ notes. Researchers said that this approach may make the system more adaptable for a range of disease-detection and prediction tasks in real-world settings.
How the model was tested
The team assessed ECG-CLIP against other models across three clinical tasks: detecting acute myocardial infarction, cardiac amyloidosis and hypertrophic cardiomyopathy; predicting future atrial fibrillation; and estimating outcomes after emergency department visits or surgery. In each of these areas, the model outperformed standard baselines and also matched the next-best model trained on the full dataset while using about 91% less hand-labeled training data on average.
Researchers also found that ECG-CLIP performed well even when only a small number of positive examples were available, including cases with as few as 10 confirmed examples of a disease. They said that could make it especially useful for rare conditions and for settings where the number of ECG leads is limited.
Potential value for routine care
Beyond disease detection, the model also showed strong performance in predicting atrial fibrillation and in forecasting short-term survival after acute care encounters, as well as the likelihood of chronic kidney disease and type II diabetes developing within three years. The researchers added saliency maps to help show which parts of the ECG signal influenced the model’s predictions, with the aim of improving interpretability for clinicians.
While the findings point to substantial potential, the study’s authors said prospective clinical trials will still be needed before the model can be considered ready for widespread use in routine care. For now, the work adds to a growing effort to apply foundation models to cardiovascular medicine, especially in environments where data, staffing or ECG resources are limited.
That combination of scale, flexibility and lower data requirements could make ECG-CLIP a notable step forward in digital cardiology if it continues to validate in real-world settings.
Photo source: AI-generated or edited image
