HumansPreprint

AI-based ECG detects and predicts hypertrophic cardiomyopathy

In 1,095 sarcomere variant carriers, an AI-ECG model detected cardiomyopathy with an AUROC of 0.91 and predicted future disease with a hazard ratio of 1.55.

medRxiv4 Oct 2026, 07:50 BST

In a study of 1,095 people carrying sarcomere gene variants, researchers evaluated an artificial intelligence model applied to standard 12-lead electrocardiogram images to identify hypertrophic cardiomyopathy (HCM). At first clinical assessment, 808 participants presented with the HCM phenotype, while 287 did not. Over subsequent follow-up, 56 of the initially phenotype-negative carriers developed manifest HCM.

The AI-ECG model detected baseline phenotypic HCM with an AUROC of 0.91. In carriers who were initially phenotype-negative, each one-standard-deviation increase in the AI score predicted future HCM onset with an adjusted hazard ratio of 1.38. In an additional analysis of 57,007 UK Biobank participants, having both high AI-ECG and high polygenic risk scores was associated with an adjusted odds ratio of 60.2 for HCM, compared to 15.0 for high AI-ECG alone and 4.1 for high polygenic risk alone.

Why it matters

Hypertrophic cardiomyopathy exhibits incomplete and age-dependent penetrance, meaning structural heart disease often emerges gradually across the lifespan. Low-cost screening tools that detect early electrophysiological shifts could improve long-term surveillance of late-onset cardiovascular disease before severe remodeling occurs.

Caveats

This study is a preprint that has not yet completed peer review. The findings are based on retrospective observational datasets and require prospective validation to show whether AI-guided monitoring improves patient outcomes.