An AI-powered stethoscope has shown remarkable potential in detecting severe heart valve disease earlier than traditional methods, according to a study led by the University of Cambridge. The technology, which analyzes heart sounds using an algorithm trained on echocardiogram data, correctly identified 98% of patients with severe aortic stenosis and 94% with severe mitral regurgitation—two of the most common and serious forms of valve disease. Valvular heart disease, often called a "silent epidemic," affects over half of people aged 65 and older, with about one in ten having significant disease. In its early stages, it typically causes no symptoms, but if left untreated, the risk of death within two years can reach 80%. Currently, diagnosis relies on echocardiography, the gold standard, but it is costly, time-consuming, and not feasible for widespread screening. NHS wait times for the test can stretch to months, limiting its use in primary care. Doctors traditionally listen to heart sounds with a stethoscope, but this practice is declining in busy GP appointments due to time constraints and the difficulty of mastering the skill. As a result, many cases go undetected until symptoms appear, often too late for optimal treatment. The new study involved 1,767 patients across five NHS Trusts, with heart sounds recorded using digital stethoscopes. Each participant also underwent an echocardiogram, which served as the reference standard. Instead of training the AI to detect heart murmurs—commonly used indicators—the researchers trained it directly on echocardiogram results, enabling it to identify subtle acoustic patterns that even experienced clinicians might miss, including cases without obvious murmurs. When tested against 14 general practitioners who listened to the same recordings, the AI outperformed every single doctor and delivered consistent, reliable results. While individual GPs varied widely in their accuracy, the AI maintained high performance across all cases, especially in identifying severe disease. The system was designed to minimize false positives, reducing the risk of overburdening already stretched echocardiography services. It requires only a few seconds of recording and can be operated by minimally trained staff, making it ideal for use in primary care settings. The researchers stress that the AI is not meant to replace doctors but to act as a screening tool, helping identify patients who need urgent referral for further investigation. Early detection could allow for timely valve repair or replacement, significantly improving long-term outcomes. Further trials in real-world GP environments with diverse patient populations are needed before widespread rollout. Detecting milder forms of valve disease remains a challenge, but the technology offers a promising path forward. “Valve disease is treatable. We can fix damaged valves and give people many more years of healthy life,” said Professor Rick Steeds from University Hospitals Birmingham. “But timing is everything. This kind of tool could help catch cases before irreversible damage occurs.” The research was supported by the National Institute for Health Research, the British Heart Foundation, and the Medical Research Council, part of UK Research and Innovation. Anurag Agarwal, who led the study, is a Fellow of Emmanuel College, Cambridge. The findings were published in npj Cardiovascular Health.