Artificial intelligence could help doctors detect serious heart valve disease years earlier, potentially saving thousands of lives, a new study suggests.

Researchers led by the University of Cambridge analysed heart sounds from nearly 1,800 patients using an AI algorithm trained to recognise valve disease, a condition that often goes undiagnosed until it becomes life-threatening.

The AI correctly identified 98% of patients with severe aortic stenosis, the most common form of valve disease requiring surgery, and 94% of those with severe mitral regurgitation, where the heart valve doesn’t fully close and blood leaks backwards across the valve.

The technology, which works with digital stethoscopes, outperformed GPs at detecting valve disease and could be used as a rapid screening tool in primary care. The results are reported in the journal npj Cardiovascular Health.

“Valve disease is a silent epidemic,” said Professor Anurag Agarwal from Cambridge’s Department of Engineering, who led the research. “An estimated 300,000 people in the UK have severe aortic stenosis alone, and around a third don’t know it. By the time symptoms appear, outcomes can be worse than for many cancers.”

Valvular heart disease affects more than half of people over the age of 65, with around one in ten having significant disease. In its early stages, it is often symptom-free. “By the time advanced symptoms develop, the risk of death can be as high as 80% within two years if untreated,” said co-author Professor Rick Steeds, from University Hospitals Birmingham. “The only current treatment is surgery to repair or replace the valve.”

Currently, diagnosis of valve disease relies on echocardiography, which is the gold standard, but is expensive and time-consuming. Wait times on the NHS can stretch to many months, meaning it cannot be used as a screening tool for the general population.

Doctors may listen to the heart using a stethoscope, but this is not routinely done in short GP appointments, and is known to miss many cases.

“Cardiac auscultation is a difficult skill, and it’s used less and less in busy GP surgeries,” said Agarwal. “That’s a big part of why so many cases of valve disease are being missed.”

The new study – a collaboration between engineers and cardiologists, research nurses and other clinicians from five NHS Trusts – used digital stethoscopes to record heart sounds from 1,767 patients. Each study participant also had an echocardiogram, which was used as a reference.

Rather than training the algorithm to recognise heart murmurs — the traditional diagnostic marker — the researchers trained it directly on echocardiogram results. This allowed the system to learn subtle acoustic patterns that humans might miss, including cases with no obvious murmur.

When tested against 14 GPs who listened to the same recordings, the algorithm outperformed every single one, and did so consistently. Individual GPs varied widely in their judgments, with some prioritising sensitivity and others specificity. The AI delivered reliable results every time and was particularly accurate for severe disease.

The system was designed to minimise false alarms, reducing the risk of overwhelming already-stretched echocardiography services. The researchers say that the technology is not intended to replace doctors, but could be a useful screening tool, helping doctors decide which patients should be referred for further investigation and treatment.

Only a few seconds of heart sound recording is needed, and the test could be carried out by staff with minimal training. “If you can rule out people who definitely don’t have significant disease, you can focus resources on those who need them most,” said Agarwal.

The researchers say that further trials, carried out in real-world GP settings with a diverse group of patients, will be needed before the device can be rolled out to the general population. In addition, they say that more moderate forms of valve disease are more difficult to detect.

However, they say that AI could help address growing pressures on the health service caused by an ageing population.

“Valve disease is treatable. We can repair or replace damaged valves and give people many more years of healthy life,” said Steeds. “But timing is everything. Simple, scalable screening tools like this could make a real difference by finding patients before irreversible damage occurs.”

The research was supported in part by the National Institute for Health Research, the British Heart Foundation, and the Medical Research Council (MRC), part of UK Research and Innovation (UKRI). Anurag Agarwal is a Fellow of Emmanuel College, Cambridge. 

Reference:
Andrew McDonald et al. ‘Development and Validation of AI-Enhanced Auscultation for Valvular Heart Disease Screening through a Multi-Centre Study.’ npj Cardiovascular Health (2026). DOI: 10.1038/s44325-026-00103-y