Deep Learning Model Predicts Heart Failure Progression from ECG Data

Heart failure, characterized by weakening heart muscles and fluid buildup, remains a leading cause of death. MIT collaborators created PULSE-HF, a deep learning model that predicts changes in left ventricular ejection fraction (LVEF) using electrocardiogram data. Validated retrospectively on cohorts from Massachusetts General Hospital, Brigham and Women's Hospital, and a public dataset, the model identifies patients likely to have worsening cardiac function.

The model aims to aid resource allocation and patient management by forecasting disease trajectory more accurately than current clinical assessments alone. However, this prediction model awaits prospective studies for clinical implementation. The research highlights the promise of AI in cardiovascular prognosis but does not replace standard clinical judgment yet.