The U.S. Food and Drug Administration (FDA) has approved the first artificial intelligence algorithm to detect heart-valve defects using a digital stethoscope. The Register reports.
EMAS software, developed by the American startup Eko, is designed to analyse patients’ pulses and assist clinicians in detecting heart murmurs.
Data are sent to the cloud, where the system in a matter of seconds determines the likely type of heart defect. EMAS then outputs the result in JSON format, which is transmitted to the app and displayed in a readable form.
“The software uses signal processing, such as filtering, as well as machine-learning algorithms to analyse the collected data and generate outputs to support clinical decisions for physicians,” the company said.
Developers evaluated the algorithm’s performance against two standard cardiology metrics: sensitivity and specificity. On the first metric, diagnostic accuracy was 85.6%, and on the second, 84.4%. By comparison, general practitioners achieved 44% and 69%, respectively, on the same metrics.
“Cardiologists typically detect murmurs more accurately than general practitioners, but most patients are seen in primary care,” the company said.
Representatives of Eko described the FDA approval as an important step “on the road to commercialisation” of the product.
Earlier in April, the maker of an autonomous AI X-ray analyzer Oxipit announced plans to file with the FDA for device certification.
In March 2021, the regulator approved the first AI device for diagnosing COVID-19.
Subscribe to ForkLog’s news on Telegram: ForkLog AI — all the news from the world of AI!