AI forecasts heart failure progression up to a year in advance

Researchers from MIT, Mass General Brigham, and Harvard Medical School developed a deep-learning model that predicts the risk of decline among heart-failure patients within a year. Known as PULSE-HF, the model forecasts changes in left ventricular ejection fraction (LVEF) using ECG data.

PULSE-HF can process both 12-lead and single-lead ECGs, making it suitable for low-resource settings, including rural clinics without specialised cardiac staff. The model achieved AUROCs of 0.87–0.91 across three patient cohorts, showing high accuracy in identifying patients at risk of severe heart failure.

By predicting future LVEF decline, clinicians can prioritise follow-up care for high-risk patients while reducing hospital visits for those at lower risk. Researchers faced challenges with data cleaning and labelling, but the model remained robust with imperfect real-world inputs.

The team plans to conduct prospective trials in real patients to validate PULSE-HF in clinical practice further. Researchers stressed that AI forecasting of heart failure could greatly improve patient outcomes and healthcare resource allocation.

Would you like to learn more about AI, tech and digital diplomacy? If so, ask our Diplo chatbot!