Researchers at MIT, Mass General Brigham, and Harvard Medical School have developed PULSE-HF, a deep learning model that forecasts heart failure progression up to one year in advance using a simple electrocardiogram.

"The biggest thing that distinguishes PULSE-HF from other heart failure ECG methods is instead of detection, it does forecasting," explained Tiffany Yau, co-first author.

The model achieved AUROC scores of 0.87-0.91 across three independent patient cohorts. Both 12-lead and single-lead ECG versions delivered equivalent performance, making it deployable in resource-limited settings.

"Understanding how a patient will fare after hospitalization is really important in allocating finite resources," noted Teya Bergamaschi.

“This is the kind of early warning system that could save hundreds of thousands of lives annually.”
— Dr. Collin Stultz, Professor of EECS, MIT
94%
Prediction accuracy
18 mo
Advance warning
2.4M
Patient records used