AHA Scientific Sessions 2026 Abstract · Nov 2, 2026
Artificial Intelligence-Enhanced Electrocardiography for Detection and Prediction of Hypertrophic Cardiomyopathy Across Monogenic and Polygenic Susceptibility
By Croon P, Choi R, Oikonomou EK et al.
Introduction: Genetic testing increasingly identifies individuals at risk for hypertrophic cardiomyopathy (HCM), including carriers of pathogenic sarcomere variants and those with elevated polygenic risk. Yet penetrance is incomplete, resulting in an increasing number of individuals entering surveillance through lifelong serial imaging. Hypothesis: We hypothesized that a validated artificial intelligence-enhanced electrocardiography (AI-ECG) model identifies the HCM phenotype at the first clinical assessment (baseline), predicts incident HCM during follow-up among phenotype-negative carriers, and is complementary to polygenic risk scores (PRS). Methods: We studied 1,095 genotype-positive (G+) carriers of pathogenic or likely pathogenic sarcomere variants from three sites: Yale (n=119), Erasmus MC (n=858), and Prague (n=118). We defined baseline as the first clinical assessment, at which carriers were classified as phenotype-positive (P+) or phenotype-negative (P–) based on the first available echocardiography or CMR report (Yale) or on the treating physician's assessment (Erasmus and Prague). A validated AI-ECG model applied to 12-lead ECG images generated an HCM score. The primary outcome was cross-sectional detection of P+ at baseline, assessed by AUROC. Secondary analyses included manifest HCM (present at baseline or during follow-up) and Cox regression for incident HCM among carriers who were G+P– at baseline. Next, we computed an established PRS in 57,007 general-population UK Biobank participants and tested whether adding AI-ECG to PRS improves HCM identification. Results: Among 1,095 G+ carriers (median age 46 [IQR 34-56]; 52.1% female), 808 (73.8%) were P+ at baseline, 56 (5.1%) were P– at baseline and developed incident HCM during follow-up, and 231 (21.1%) remained P–. AUROC was 0.90 (95% CI 0.88-0.92) for P+ at baseline and 0.92 (0.90-0.94) for manifest HCM (Figure 1) ; at the prespecified threshold, sensitivity was 0.78, specificity 0.87, PPV 0.94, and NPV 0.60. Among P– carriers (median follow-up 6.6 years), higher AI-ECG scores predicted incident HCM (HR 1.55 per 1-SD; 95% CI 1.28-1.88; p<0.001; adjusted HR 1.38; 1.11-1.71; p=0.004). In the UK Biobank, both high AI-ECG and high PRS conferred 60-fold higher odds of an HCM diagnosis (age- and sex-adjusted OR 60.2; 95% CI 26.5-137.2) vs 15.0 for AI-ECG alone and 4.1 for PRS alone (Figure 2) . Conclusions: AI-ECG can detect overt HCM in sarcomere variant carriers and predict incident HCM in those without HCM expression. Moreover, AI-ECG complements PRS in the general population, supporting AI-ECG as a scalable tool to detect HCM and guide surveillance in monogenic and polygenic susceptibility.