US researchers have created an AI-driven screening tool named EchoNext, which utilizes data from affordable electrocardiograms (ECG) to uncover concealed heart diseases.
EchoNext pinpoints patients who require an ultrasound, a non-invasive procedure employed to diagnose structural heart issues. The tool has been shown to surpass cardiologists in accuracy when it comes to detecting diseases that are not identifiable through ECG alone.
The deep learning model was developed using over 1.2 million ECG-echocardiogram pairs sourced from 230,000 patients. In a validation study conducted across four hospital systems, EchoNext exhibited a high level of accuracy in recognizing structural heart conditions, such as heart failure resulting from cardiomyopathy, valve disease, pulmonary hypertension, and significant thickening of the heart.
When compared to 13 cardiologists analyzing 3,200 ECGs, EchoNext successfully identified 77% of structural heart issues. In contrast, cardiologists relying on ECG data achieved an accuracy rate of 64%.
This technology has the potential to transform the estimated 400 million ECGs conducted globally this year into 400 million opportunities to screen for structural heart disease, potentially providing life-saving treatment at the most critical moments.
