Particle.news
Download on the App Store

University of Michigan AI Detects Microvascular Heart Disease From a 10-Second EKG

Built with self-supervised learning on over 800,000 EKGs, the model was fine-tuned using PET scans to predict myocardial flow reserve.

Overview

  • Peer-reviewed findings appear in NEJM AI, detailing the model’s performance and methodology.
  • The tool significantly outperformed previous EKG-based AI across diagnostic tasks, including myocardial flow reserve prediction.
  • Accuracy with resting EKGs was strong, with only minimal additional gain from exercise stress tests.
  • Researchers say the approach could help flag patients for advanced imaging in settings without PET or other specialty scanners.
  • Coronary microvascular dysfunction is often missed in routine evaluations, and roughly 14 million people seek care for chest pain in the U.S. each year.