Can AI Help Predict Which Heart-Failure Patients Will Worsen Within a Year?
Innovative Deep Learning Model Aims to Transform Heart Failure Predictions
Researchers from MIT, Mass General Brigham, and Harvard Medical School have developed an advanced deep-learning model called PULSE-HF to predict the progression of heart failure in patients up to a year ahead of time. Heart failure, marked by weakened heart muscles and chronic fluid buildup, poses significant health risks and burdens healthcare systems globally.
How PULSE-HF Works
PULSE-HF stands for “Predict changes in left ventricULar Systolic function from ECGs of patients who have Heart Failure.” The system analyzes electrocardiogram (ECG) data to forecast whether a patient's left ventricular ejection fraction (LVEF)—the amount of blood pumped out with each heartbeat—will decrease below 40%, a threshold indicating severe heart failure.
The model was trained and validated using datasets from Massachusetts General Hospital, Brigham and Women’s Hospital, and the publicly available MIMIC-IV database. It achieved strong predictive performance, with AUROC scores between 0.87 and 0.91 across different patient groups.
Clinical Benefits and Flexibility
PULSE-HF not only helps prioritize follow-up for high-risk patients but can also reduce unnecessary visits for those at lower risk. Uniquely, the model works with both standard 12-lead ECGs and simpler single-lead ECGs, making it practical for use in low-resource settings where access to advanced imaging like ultrasounds may be limited.
Challenges and Next Steps
Developing the model required extensive data collection and cleaning, as labeling clinical data such as echocardiogram results can be time-consuming and complex. Looking ahead, the research team plans to validate PULSE-HF in prospective studies involving real patients with unknown outcomes, moving closer to practical deployment in healthcare.
This advancement highlights the potential of AI in healthcare, offering more accurate resource allocation and possibly improved survival rates for heart failure patients.
For more details, read the original article at MIT News.
