Plate Nº 67 · recorded October 10, 2026
Health & Medicine ResearchReported finding
Yale AI Reads a Simple ECG Photo to Spot Deadly Heart Disease
A Yale AI platform can flag a deadly, widely underdiagnosed heart condition from a smartphone photo of an ECG, a study in JAMA reports. Untreated, the disease cuts life expectancy to five years.
By Nathan Brooks4 min read716 words
In brief
- Untreated transthyretin amyloid cardiomyopathy reduces average life expectancy to five years.
- The Yale CarDS Lab platform detects at-risk patients from smartphone photos of standard ECG printouts.
- The study, published in JAMA (2026), tested the model in eight patient cohorts across the US and Europe.
- The tool received FDA breakthrough device designation and is currently under FDA review.
- The TRACE-AI Network Study is evaluating the platform at 13 US health centers.
Untreated transthyretin amyloid cardiomyopathy cuts average life expectancy to five years — and now a Yale-built AI platform can flag people at risk using nothing more than a smartphone photo of a standard ECG printout. The study, published recently in JAMA, shows the tool successfully identified individuals with the disease across eight patient groups in the United States and Europe.
The platform, developed by the Cardiovascular Data Science (CarDS) Lab at Yale School of Medicine, targets one of cardiology's most persistently missed diagnoses. Amyloid cardiomyopathy occurs when misfolded proteins clump together and accumulate on the heart muscle, displacing healthy tissue.
"For a disease that's massively underdiagnosed, identifying those at risk is very critical," says Rohan Khera, M.D., director of the CarDS Lab and the study's principal investigator.
Why is this disease so hard to catch?
The study focused on the subtype caused by misfolded transthyretin, a protein made in the liver that can become misshapen due to genetics or aging. As the protein builds up on the heart, the organ grows stiffer and its electrical system — the wiring that regulates the heartbeat — starts to fail.
Diagnosis often comes late. Symptoms overlap with other cardiovascular conditions, so many patients learn they have the disease only after dangerous complications such as heart failure have already set in.
"It's as aggressive as some of the most aggressive cancers," Khera says of the condition.
How does the AI tool work?
An electrocardiogram (ECG) is a noninvasive test in which electrodes on the skin measure the heart's electrical signals. Doctors worldwide perform hundreds of millions of these tests each year. Yet providers currently do not use ECGs to identify people at risk of amyloid cardiomyopathy — the disease-linked patterns are too subtle for the human eye.
The Yale team set out to change that. Their approach works in two stages:
- First, they trained an AI model to interpret ECG results using data from thousands of de-identified patients.
- Then, using the few hundred patients in that data set already diagnosed with amyloid cardiomyopathy, they refined the model to recognize ECG patterns associated with the disease.
The result is a screening tool doctors can access from their smartphones.
"You can take a photo of an ECG like you would take a photo of a bank check," Khera explains. "Our model can pick up from that photo whether the ECG came from somebody at risk for cardiac amyloid or not."
In the new study, the researchers tested the model in eight distinct patient cohorts across the United States and Europe. It successfully identified individuals with transthyretin amyloid cardiomyopathy in that testing.
What could this change in practice?
Rather than replacing doctors, the tool aims to narrow the search. Because testing every heart patient for amyloid is impractical, the AI acts as a filter that prioritizes who needs further evaluation.
"Our tool can really narrow down the funnel for who should be further evaluated for cardiac amyloid," says Philip Croon, MMed, associate research scientist at Yale School of Medicine and the study's first author.
Earlier detection matters because it opens a window for intervention before the heart sustains extensive damage. Cardiovascular diseases remain the leading cause of death worldwide, and their prevalence is growing as risk factors such as obesity and diabetes rise.
The findings come with clear caveats. The study demonstrates the platform's ability to identify people with the disease in tested cohorts, but it is not yet an FDA-approved diagnostic device. The tool is currently under FDA review, and it received the agency's breakthrough device designation through its Breakthrough Devices Program, which offers an expedited review path for technologies with the potential to save lives.
A separate observational study, the TRACE-AI Network Study, is now examining how this and other multimodal AI tools can detect transthyretin amyloid cardiomyopathy at scale across 13 health centers in the United States.
"We've been able to solve a critical bottleneck in deploying therapies by identifying more at-risk people," Khera says. "This is one of our biggest accomplishments — making care accessible by finding people who need treatment the most."
Publication: Philip M. Croon et al., "Electrocardiogram-Based Deep Learning to Prioritize Testing for Transthyretin Amyloid Cardiomyopathy," JAMA (2026). DOI: 10.1001/jama.2026.16785
via Medical Xpress (Source)
More from Nathan Brooks
Nearby plates
- Heart Failure Deaths Double in Younger and Middle-Aged Americans
- Nearly 1 in 3 Diabetes Patients Face Kidney, Heart Disease or Death
- High IgA Levels Identify a High-Risk Subtype of Fatty Liver Disease
- Aging Muscle's Paradox: One Fat Molecule May Hold the Fix
- Brain Structure Changes May Signal Alzheimer's Years Before Plaques