Four hours of training and an AI: heart scans by non-experts show promise
What happened
Researchers at the Mayo Clinic have tested a way of using artificial intelligence to help people with no ultrasound experience take useful pictures of the heart. The target was a condition called aortic stenosis. The aortic valve is one of the heart's four valves. In aortic stenosis it becomes stiff and narrow, so the heart has to work harder to push blood out to the body. In the study, people with no previous ultrasound training were given about four hours of instruction. They then used a handheld scanner with two AI models working alongside them. One guided them in real time as they moved the probe, helping them get a usable image. The other analysed the pictures for signs of moderate or worse aortic stenosis. The findings were published in JAMA Cardiology on 28 August 2026 and presented at the 2026 congress of the European Society of Cardiology. Coverage from Mayo Clinic and other outlets followed on 3 September 2026. The researchers reported that the combined approach reduced false positives, meaning fewer people were wrongly flagged. They also reported, clearly, that some people who did have aortic stenosis were not picked up.
Why this matters
Aortic stenosis is common and it is serious. It affects roughly 7 per cent of people aged 75 and over, and it is the most common reason worldwide for someone to need a heart valve procedure. It is also easy to miss. In the early stages there may be no symptoms at all. Later, people feel breathless, dizzy, or unusually tired, and it is very tempting to put all of that down to getting older. The test that confirms it is an echocardiogram, an ultrasound scan of the heart. That requires a trained sonographer and a proper machine. In many parts of the world, and in plenty of rural parts of wealthy countries too, those are in short supply. That is the gap this research is aiming at. If someone with four hours of training and a handheld device can reliably flag the people who need a proper scan, then screening could reach places it currently does not. It is a shift in what the technology is for. This AI is not trying to replace a cardiologist. It is trying to work out who should get in front of one.
What the evidence actually says
This is a real result in a serious journal, and it deserves to be taken seriously. It also comes with limits the researchers state themselves. The workflow is designed for screening and triage. It is explicitly not a replacement for a full echocardiogram or for a doctor's assessment. Anyone flagged as possibly having moderate or worse disease still needs a proper scan to confirm the diagnosis and work out how severe it is. The honest weakness is in the misses. The study reports that some patients with aortic stenosis were not identified. In screening terms these are false negatives, and they matter, because a person told they look fine may not come back for a while. The reduction in false positives is genuinely useful, but a screening tool is judged on both sides of that trade. There is also a gap between a research study and everyday practice. Study participants were trained for a specific task, using specific equipment, in a controlled setting. Whether the same results hold in a busy clinic, with mixed equipment and interruptions, is a separate question that needs separate research. And this is one study. Encouraging findings in cardiology have failed to hold up before when tested more widely. JAMA Cardiology published an accompanying editorial alongside the paper, which is usually a sign that editors think a result is important but wants context around it. So: a promising step, with an honest account of its own weaknesses, and some distance still to travel.
Practical advice
Nothing here changes what you should do today, but the underlying condition is worth knowing about. Aortic stenosis becomes more common with age. If you are over 65, the symptoms worth mentioning to your GP are breathlessness when doing things that used to be easy, chest tightness on exertion, dizziness or light-headedness, fainting, or unusual tiredness. The most important thing is not to assume that these are simply age. They might be. But they are also the classic pattern for a narrowed heart valve, and that is treatable. Many cases are first suspected because a doctor or nurse hears a murmur while listening to your chest with a stethoscope. That is a good reason not to skip routine checks such as an NHS Health Check if you are invited. If you already know you have a heart valve problem, keep to your review appointments even when you feel well. Valve disease can progress quietly. And be careful with claims about AI health scans sold directly to the public. This research was about trained users working under clinical supervision, not about consumer gadgets. If you see a private scan advertised, ask what happens to the result, who interprets it, and what follow-up you get.
What to know
Mayo Clinic researchers have shown that people with about four hours of training, helped by two AI models, can capture heart ultrasound images and flag possible moderate or worse aortic stenosis. The study appeared in JAMA Cardiology on 28 August 2026 and was presented at the 2026 European Society of Cardiology congress, with wider coverage on 3 September 2026. The approach cut false positives, but it also missed some people who genuinely had the condition. It is designed for screening and triage, not diagnosis. Aortic stenosis affects around 7 per cent of people aged 75 and over and is the most common reason for heart valve intervention worldwide. The potential value is in reaching places where trained sonographers and full echo machines are scarce. For now, the practical message is unchanged: if you are getting breathless, dizzy or unusually tired, tell your GP rather than putting it down to age. Sources: JAMA Cardiology, "Artificial Intelligence-Enabled Acquisition and Interpretation for Screening Aortic Stenosis", 28 August 2026, https://jamanetwork.com/journals/jamacardiology/fullarticle/2853493 | Mayo Clinic News Network, "AI-assisted approach helps novice users capture heart ultrasound images, identify aortic stenosis", September 2026, https://newsnetwork.mayoclinic.org/discussion/ai-assisted-approach-helps-novice-users-capture-heart-ultrasound-images-identify-aortic-stenosis/ | Medical Xpress, "AI-assisted approach helps novice users capture heart ultrasound images and identify aortic stenosis", 3 September 2026, https://medicalxpress.com/news/2026-09-ai-approach-novice-users-capture.html This article is for general information and does not replace advice from a doctor, pharmacist or other qualified healthcare professional. If you have symptoms such as chest pain, breathlessness or fainting, seek medical advice promptly.
Related articles
Three quiet changes to how the UK regulates medicines and devices
On 1 September 2026, the Medicines and Healthcare products Regulatory Agency announced that the government has tabled three amendments to the Health Bill curren…
Your night-time cough may warn of a flu wave a week early, UKHSA study finds
On 3 September 2026, the UK Health Security Agency published the results of a study it carried out with Sleep Cycle, a Swedish company that makes a popular slee…