Early detection of aortic stenosis remains a major challenge, with many patients presenting late in the disease course. Findings presented at the European Society of Cardiology (ESC) Congress 2025 suggest that longitudinal electrocardiogram (ECG) analysis using an artificial intelligence (AI) model could offer a scalable solution.

An analysis of 7,860 ECGs from 2,040 transcatheter aortic valve replacement (TAVR) recipients showed that the AI model (AK-AVS) identified high-risk signals up to 4.5 years prior to the procedure. More than 90% of patients exceeded a risk threshold in the months before TAVR.

Three ECG progression patterns—Persistently High, Accelerated Progression, and Stable Low—were identified through unsupervised clustering, with the first two associated with significantly higher one-year mortality. Integration of trajectory data improved predictive accuracy beyond EuroSCOREII and STS risk scores.

These results highlight the potential role of AI-driven ECG analysis as a cost-effective, widely accessible tool for early detection of aortic stenosis and for enhanced risk stratification.