AI-ECG for Time-Resolved Prediction of HFrEF

Overview

This study aims to develop and validate a deep learning-based electrocardiogram (ECG) model for predicting the future risk of heart failure with reduced ejection fraction (HFrEF). The model is trained using raw 12-lead ECG data and generates individualized, time-resolved risk estimates over a 5-year period. Data are obtained from multiple cohorts, including Zhongshan Hospital, Shanghai Tenth People's Hospital, and Beth Israel Deaconess Medical Center, representing diverse populations across China and the United States. The model is designed to identify individuals at elevated risk of developing HFrEF before the onset of overt clinical disease. The performance of the model is evaluated using multiple complementary metrics, including discrimination, calibration, and clinical utility. In addition, interpretability analyses are conducted to explore the physiological relevance of ECG features associated with predicted risk. This study seeks to provide an accessible and scalable tool for early risk stratification of heart failure, with the potential to support timely clinical decision-making and improve patient outcomes.

Heart failure with reduced ejection fraction (HFrEF) is associated with substantial morbidity and mortality worldwide, and early identification of individuals at risk remains a major clinical challenge. Although existing risk models and biomarkers can provide prognostic information, their application is often limited by the need for laboratory testing or imaging, as well as variability in performance across populations. In this study, we develop a deep learning-based survival model using raw 12-lead electrocardiogram (ECG) data to predict the future onset of HFrEF. The model is designed to generate individualized, time-to-event risk estimates over a 5-year follow-up period, allowing for dynamic assessment of risk trajectories rather than static classification. The model is trained on data from Zhongshan Hospital and externally validated in independent cohorts from Shanghai Tenth People's Hospital and Beth Israel Deaconess Medical Center. These cohorts include a broad spectrum of patients, ranging from individuals without known cardiovascular disease to those with diverse clinical conditions, thereby enabling evaluation of model generalizability across different healthcare systems and demographic subgroups. Model performance is comprehensively assessed using multiple metrics, including the concordance index, time-dependent area under the receiver operating characteristic curve, area under the precision-recall curve, Brier score, calibration analysis, and decision curve analysis. Risk stratification capability is evaluated using Kaplan-Meier survival analysis. To enhance interpretability, complementary representation-based and attention-based methods are applied. These include variational autoencoder-derived latent feature analysis, correlation with conventional ECG parameters, and gradient-based visualization techniques to identify waveform regions contributing to model predictions. These approaches aim to ensure that the model captures physiologically meaningful signals associated with myocardial remodeling and cardiac dysfunction. This study is observational and retrospective in nature and does not involve any intervention. The findings aim to support the development of a non-invasive, cost-effective, and widely accessible tool for early detection of individuals at risk of HFrEF, with potential implications for preventive strategies and personalized clinical management.