Can AI help predict which heart-failure patients will worsen within a year?

2026-03-12 · Source: MIT News - Artificial intelligence · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Data Science & Analytics, Emerging Technologies & Innovation · Depth: Intermediate, medium

Summary

Researchers at MIT, Mass General Brigham, and Harvard Medical School developed PULSE-HF, a deep-learning model designed to forecast heart failure prognosis up to a year in advance. Published in *Lancet eClinical Medicine* on March 12, 2026, the model predicts changes in left ventricular ejection fraction (LVEF), specifically identifying patients whose LVEF is likely to fall below 40 percent. PULSE-HF was retrospectively tested across three patient cohorts from Massachusetts General Hospital, Brigham and Women’s Hospital, and MIMIC-IV, achieving AUROC scores ranging from 0.87 to 0.91. A notable finding is that a single-lead ECG version of PULSE-HF performed as strongly as the 12-lead version, enhancing its applicability in low-resource settings. The project involved significant challenges in collecting and cleaning ECG and echocardiogram datasets due to messy real-world data.

Key takeaway

For AI scientists developing clinical prediction tools, consider the PULSE-HF model's success in forecasting rather than just detecting conditions. Your focus should be on building robust models that can handle "slightly messy" real-world data, as perfect data is often unattainable. Prioritize the development of single-lead versions for broader applicability in diverse clinical environments, especially those with limited resources, to maximize impact and patient benefit.

Key insights

PULSE-HF is a deep-learning model forecasting heart failure worsening via ECGs, achieving high accuracy in LVEF decline prediction.

Principles

• Forecasting future LVEF decline is a novel application.
• Single-lead ECGs can perform comparably to 12-lead ECGs.

Method

The PULSE-HF model takes an electrocardiogram as input and outputs a prediction of whether the patient's ejection fraction will fall below 40 percent within the next year, based on patterns learned from labeled ECG and echocardiogram data.

In practice

• Prioritize high-risk heart failure patients for follow-up.
• Reduce hospital visits for lower-risk patients.
• Deploy in low-resource clinical settings.

Topics

• Deep Learning
• Heart Failure Prediction
• Electrocardiogram Analysis
• Clinical Prognosis
• Medical AI

Best for: AI Scientist, AI Researcher, AI Data Scientist, Research Scientist

Related on AIssential

• Scientists develop AI tool to spot heart failure risk five years before it strikes — An AI tool predicts heart failure risk five years early by analyzing cardiac CT scans for invisible fat inflammation.
• A hidden predictor of sudden cardiac death uncovered by deep learning — Deep learning identifies novel ECG features to predict sudden cardiac death risk more accurately than current methods.
• Shameless plug alert: Win prizes by forecasting real healthcare data to help UK’s health service save lives — Forecasting patient admission risks can directly save lives by enabling proactive hospital interventions.
• Building a Causal AI Prototype for Heart Failure: From Risk Prediction to Counterfactual Treatment… — Causal AI can shift healthcare from predicting risk to prescribing actions that alter patient trajectories.
• A Multimodal and Explainable Machine Learning Approach to Diagnosing Multi-Class Ejection Fraction from Electrocardiograms — Multimodal machine learning can accurately stratify LVEF from ECG and EHR data, enhancing accessibility.
• Clinically Interpretable Sepsis Early Warning via LLM-Guided Simulation of Temporal Physiological Dynamics — LLM-guided temporal simulation improves sepsis early warning by providing interpretable physiological trajectory predictions.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by MIT News - Artificial intelligence.