An artificial intelligence model for digital stethoscopes can identify patients with weakened hearts that can’t pump blood effectively, according to a new peer-reviewed study published in the March issue of JACC: Advances.
This condition, known as reduced ejection fraction, is an indicator of heart failure. An echocardiogram is typically used to diagnose it, but it’s not widely available because the technology is expensive, it requires specialist training and it’s a time-consuming examination. The new AI model is intended to be used by primary care physicians to detect heart problems earlier before symptoms escalate.
Related: Mayo Clinic, Philips collaborate to speed up cardiac MRI with AI
The study investigated Eko Health’s Low EF AI, which was cleared by the Food and Drug Administration in April for use with the company’s Sensora Cardiac Early Detection Platform. The model, developed in conjunction with Mayo Clinic, was trained using a proprietary dataset of more than 100,000 electrocardiogram and echocardiogram pairs from unique patients.
“We still want to confirm the diagnosis with echocardiography, but instead of letting a lot of people fall through the gaps who never get referred to echocardiography or sending way too many people to echocardiography who don't need it, this is a way to increase your pretest probability of [having] an appropriate patient to send for follow-up testing,” said Rose McDonough, senior manager of medical affairs at Eko Health.
For the study, 2,960 adult patients were initially screened with the Eko’s Low EF AI and within a week they also received echocardiograms. The researchers found that the AI model accurately identified 77.5% of true cases of reduced ejection fraction.
More than 500,000 healthcare professionals currently use Eko’s products, including its digital stethoscopes and AI technology. In addition to the Low EF AI, the company also has FDA-cleared algorithms for detecting atrial fibrillation and structural heart murmurs.
