AI Breakthrough: Hidden Heart Condition Now Detectable via Simple 10-Second EKG

A notorious "ghost" condition that causes chest pain but frequently evades detection on standard tests may finally have a cost-effective solution. Researchers at the University of Michigan Health have developed an artificial intelligence model capable of detecting Coronary Microvascular Dysfunction (CMVD) using nothing more than a standard electrocardiogram (EKG).

The breakthrough, published this week in NEJM AI, promises to close a dangerous diagnostic gap for millions of patients—particularly women—who are often told their heart arteries are "clear" despite suffering from debilitating symptoms.

The "Invisible" Heart Disease
CMVD is a condition that affects the tiny arterioles of the heart rather than the large arteries. Because standard angiograms only visualize major blockages, patients with CMVD often pass these tests with flying colors, leaving doctors baffled and patients untreated. Until now, the only reliable way to diagnose the condition was through PET myocardial perfusion imaging—an expensive, invasive, and scarce test available only at specialized medical centers.

"Our model creates a way for clinicians to accurately identify a condition that is notoriously hard to diagnose—and often missed in emergency department visits—using a 10-second EKG strip," said Dr. Venkatesh L. Murthy, the study’s senior author and a cardiologist at the Frankel Cardiovascular Center.

Self-Taught Intelligence
The University of Michigan team overcame a major hurdle to build this tool: the scarcity of labeled data. Since very few patients undergo the "gold standard" PET scans needed to train an AI, the researchers used a technique called Self-Supervised Learning (SSL).

Pre-training: The AI first analyzed over 800,000 unlabeled EKG waveforms, effectively teaching itself to "read" the electrical language of the heart without human instruction.

Fine-tuning: Once the model understood the basics, researchers trained it on a smaller, high-quality dataset of patients who did have PET scan data.

The result is a tool that outperforms previous state-of-the-art models. Remarkably, the study found that the AI could detect the disease just as well using a resting EKG as it could with data from a stress test, meaning patients may not even need to run on a treadmill to get a diagnosis.

Why This Changes Emergency Care
The implications for emergency rooms are profound. Currently, many patients presenting with chest pain are sent for an angiogram. If the large arteries look clear, they are often discharged with a diagnosis of "non-cardiac chest pain," despite actually suffering from microvascular dysfunction.

"In hospitals with limited resources or non-specialty centers, using our EKG-AI model... will be an easy, cost-effective, and non-invasive way to identify when a patient would benefit from advanced testing," Dr. Murthy noted.

By flagging these "invisible" cases early, doctors can initiate treatment for microvascular disease—such as statins or ACE inhibitors—rather than sending patients home with unanswered questions.