AI ECG Spots Deadly Disease Months Early

A ten-second heart test is now catching a deadly, hidden disease months before doctors usually see it.

Story Snapshot

A standard 10-second electrocardiogram, boosted by artificial intelligence, flagged cardiac amyloidosis in a 77-year-old man with 98 out of 100 confidence.
Mayo Clinic’s artificial intelligence electrocardiogram has been run over one million times to screen for silent heart problems.
Across large studies, these tools reach area under the curve scores around 0.9 and can predict disease more than six months before normal diagnosis.
Experts praise the breakthrough but warn about bias, small early studies, and the need for real-world trials before routine nationwide use.

How a simple heart test exposed a silent killer

Cardiac amyloidosis is one of those diseases that most people never hear about until it is almost too late. It happens when abnormal proteins build up in the heart muscle and slowly stiffen it, often hiding behind common symptoms like fatigue, swelling, or shortness of breath. By the time many patients get the right diagnosis, they have advanced heart failure, and treatment choices are limited. That slow, quiet damage is what makes early detection so valuable in the first place.

Mayo Clinic physicians now use an artificial intelligence system that reads a routine 12-lead electrocardiogram and looks for patterns linked to amyloidosis. In one widely discussed case, the tool scanned a standard 10-second strip from 77-year-old patient Mike Busch and returned a confidence score of 98 out of 100 for cardiac amyloidosis. That single result triggered further testing, confirmed the diagnosis, and allowed doctors to start targeted therapy while his heart still had meaningful reserve left.

From research code to FDA-cleared screening tool

The artificial intelligence electrocardiogram did not start as a flashy gadget. Researchers first trained it on tens of thousands of routine electrocardiograms from patients with and without cardiac amyloidosis. They taught the model to spot subtle changes in voltage, timing, and wave shape that are invisible to the human eye but repeat across many amyloidosis cases. In its main clinical study, the tool reached an area under the curve of 0.91 and a positive predictive value of 0.86 for detecting either major type of cardiac amyloidosis.

Area under the curve is a measure of how well a test separates sick from healthy patients. A perfect test scores 1.0. The artificial intelligence electrocardiogram’s score around 0.9 put it firmly in the “high accuracy” range. In follow-up work, researchers showed that, in patients who had electrocardiograms before they were diagnosed, the model predicted cardiac amyloidosis more than six months before doctors formally recognized it in 59 percent of cases. That is not science fiction; it is a documented lead time that can translate into earlier drug therapy and fewer hospital stays.

Real-world impact and the scale of deployment

Those numbers did not stay trapped in journals. Mayo Clinic reports that clinicians there have now used artificial intelligence electrocardiogram tools more than one million times to screen for a range of heart problems, including low ejection fraction, hypertrophic cardiomyopathy, aortic stenosis, atrial fibrillation, and cardiac amyloidosis. The amyloidosis-focused model was cleared by the United States Food and Drug Administration as a breakthrough device, making it the first commercially available artificial intelligence echocardiography tool for amyloid cardiomyopathy screening.

One related echocardiography model, which reads a single video clip of the beating heart, achieved an area under the curve of 0.93 with 85 percent sensitivity and 93 percent specificity in an external validation cohort drawn from 18 international sites. Those numbers mean the model correctly identified most patients who truly had cardiac amyloidosis and correctly reassured most patients who did not. That kind of performance, repeated across independent centers, starts to look less like hype and more like a useful extension of standard practice.

A simple 10-second heart test with the aid of artificial intelligence is helping physicians detect a serious, often-overlooked disease.

For Rochester businessman Mike Busch, that technology proved life-changing. After months of unexplained symptoms, an AI-enhanced ECG helped… pic.twitter.com/Bo8XBvNcFt

— Mayo Clinic (@MayoClinic) July 13, 2026

Questions about bias, generalizability, and next steps

Researchers also warn about bias tied to the training data. Many artificial intelligence electrocardiogram models, including those for amyloidosis, rely heavily on large archives from a few academic centers in the United States. If most patients in those archives come from limited regions or share certain traits, the model may perform worse in other states, smaller hospitals, or more diverse communities. That matters for fairness and for trust; people want to know that a tool used to guide life-and-death heart decisions works for their background, not just for a narrow slice of the population.

The good news is that serious prospective trials are now under way. One registered study is testing artificial intelligence electrocardiogram screening for cardiac amyloidosis in atrial fibrillation patients and tracking how it changes diagnostic timing and outcomes compared with usual care. Other work is exploring cost-effectiveness, prognostic value in severe aortic stenosis, and performance in lower-prevalence populations.