AI ECG Tool Detects Hidden Heart Disease More Accurately Than Doctors: Study Finds - Video

A recent study published in Nature has demonstrated that an artificial intelligence (AI) model can reliably detect diverse structural heart diseases from electrocardiograms (ECGs), outperforming standard physician review. Developed by researchers across eight NewYork-Presbyterian hospitals, the AI model named EchoNext was designed to act as a multitask classifier capable of identifying subtle signs of structural heart diseases that may otherwise go undiagnosed.

Diagnosis of structural heart diseases often relies on expensive and inaccessible tools like echocardiography. Early detection is critical, but over 6% of older adults with significant valvular heart disease remain undiagnosed. EchoNext aims to fill this gap by providing a low-cost, scalable solution through AI interpretation of routine 10-second ECGs.

To develop and validate the model, researchers compiled over 1.2 million ECG-echocardiogram pairs from 230,318 adults treated between 2008 and 2022. EchoNext processed raw 12-lead ECG waveforms, seven standard ECG parameters, and demographic data. It was evaluated using a held-out internal test set as well as external datasets from Cedars-Sinai, the Montreal Heart Institute, and the University of California, San Francisco.

The model performed strongly, detecting composite structural heart diseases with an AUROC of 85.2% internally and 78–80% externally. In a reader study, EchoNext outperformed 13 cardiologists who reviewed de-identified ECGs, achieving 77% accuracy versus 64% for clinicians. When doctors were shown EchoNext's risk score, their accuracy rose to 69%.

The researchers also conducted a silent deployment on over 84,000 imaging-naive patients in 2023, finding that EchoNext could have flagged nearly 2,000 hidden structural heart diseases cases that were otherwise missed. In a small prospective pilot, 73% of high-risk participants had previously unrecognized structural heart diseases.

These findings suggest that AI-based ECG analysis like EchoNext could serve as a valuable screening tool to triage patients, optimize echocardiography use, and improve early structural heart diseases detection, especially in resource-limited settings.

Reference: Poterucha, T.J., Jing, L., Ricart, R.P. et al. Detecting structural heart disease from electrocardiograms using AI. Nature (2025). https://doi.org/10.1038/s41586-025-09227-0