AI model ECG-CLIP improves cardiovascular disease detection

Researchers have developed a new artificial intelligence model, ECG-CLIP, designed to detect and predict various cardiovascular conditions with high efficiency. Published in The Lancet Digital Health, the study demonstrates that the model can adapt to specific clinical tasks using approximately 91% less manually labeled training data than previous systems.

ECG-CLIP functions as a foundational model, pre-trained on over 1.7 million electrocardiograms from more than 542,000 patients. Unlike conventional machine learning models that rely heavily on large, pre-classified datasets for every new condition, ECG-CLIP associates electrical heart signals with corresponding clinical reports to learn general representations. This allows for rapid adaptation to new diseases even when characterized cases are scarce.

The technology aims to provide a preliminary assessment in approximately two seconds, serving as a support tool for healthcare professionals rather than a replacement for cardiologists. By identifying patterns associated with cardiac alterations, the tool can help prioritize urgent cases and reduce patient waiting times in high-pressure medical environments.