An AI model applied to standard 12-lead ECG images demonstrated exceptional performance in identifying hypertrophic cardiomyopathy (HCM) across a multi-center cohort of 1,095 genotype-positive sarcomere variant carriers. At baseline, the AI-ECG achieved an AUROC of 0.91 for phenotypic detection and 0.92 for manifest HCM. Critically, among 231 phenotype-negative carriers, higher AI-ECG scores predicted future HCM development with an adjusted hazard ratio of 1.38 per standard deviation. In 57,007 UK Biobank participants, combining high AI-ECG scores with a polygenic risk score produced 60-fold elevated HCM odds versus 15-fold and 4-fold for either tool alone.
HCM affects roughly 1-in-500 adults and is a leading cause of sudden cardiac death in young people, yet surveillance currently depends on expensive serial cardiac MRI or echocardiography. This AI-ECG approach — applied to a ubiquitous, low-cost test — could dramatically reshape cascade screening protocols, prioritizing intensive imaging for the highest-risk carriers while safely deferring it for others. The 59% negative predictive value at the chosen threshold warrants caution, however; a substantial minority of future converters could be missed. The cohort skews toward established referral centers, potentially inflating performance versus community settings. Integration with polygenic scores is genuinely novel and suggests a layered risk architecture that neither tool achieves alone. This is a preprint posted to medRxiv and has not yet undergone peer review, so findings require independent validation before clinical adoption. Still, the mechanistic synergy between monogenic and polygenic pathways makes this work potentially paradigm-shifting for HCM surveillance strategy.
