Forecasting mortality and cardiovascular risks through AI-estimated biological heart age from 12-Lead ECGs

Abstract Background The integration of artificial intelligence (AI) into clinical diagnostics presents novel opportunities for enhancing predictive accuracy in cardiology. Utilizing AI to derive biological heart age from standard 12-lead electrocardiograms (ECGs) introduces a potential paradigm shift in cardiovascular risk assessment. Purpose This investigation aims to evaluate the prognostic capabilities of a deep learning-based algorithm that calculates biological ECG heart age (AI ECG-heart age) from 12-lead ECGs, comparing its predictive power against traditional chronological age (CA) for mortality and cardiovascular outcomes. Methods A deep neural network was developed and trained on a substantial dataset of 425,051 12-lead ECGs collected over fifteen years, with subsequent validation and testing on an independent cohort of 97,058 ECGs. Comparative analyses were conducted among age and sex-matched patients differentiated by ejection fraction (EF). Results The AI demonstrated a mean absolute error in estimating ECG-heart age of 5.8 ± 3.9 years, achieving an R squared of 0.7 (r=0.84, p<0.0001). In adjusted Cox proportional-hazards models, an AI ECG-heart age exceeding CA by seven years markedly increased risks for all-cause mortality (HR 1.62 [1.43-1.84]) and major adverse cardiovascular events (MACEs) (HR 1.92 [1.65-2.23]). In contrast, an AI ECG-heart age that was seven years younger than CA correlated with significantly reduced risks (HR 0.86 [0.77-0.95] for mortality; HR 0.73 [0.63-0.84] for MACEs). Additionally, subjects with reduced EF consistently exhibited elevated AI ECG-heart ages along with prolonged QRS durations and corrected QT intervals. Conclusion Biological heart age estimated by AI from 12-lead ECGs is strongly associated with increased mortality and cardiovascular events, underscoring its utility in enhancing early detection and preventive strategies in cardiovascular healthcare. This study confirms the transformative potential of AI in refining clinical assessments and improving patient outcomes.