An artificial intelligence system that can identify people likely to have a heart attack within a decade in the future could soon be in use across Britain.
The technology, which could save thousands of lives a year, is being assessed by the National Institute for Health and Care Excellence (Nice) and a decision on its use in the NHS is expected by the end of the year.
The project’s scientists also revealed that they are working on similar AI systems to predict whether someone is at risk of having a stroke and to recognize people at risk of conditions such as diabetes.
“This technology has now been tested in several hospitals across the UK and the results are extremely encouraging,” said Prof. Charalambos Antoniades, leader of the Orfan (Oxford Risk Factors And Non Invasive Imaging) study. “If installed nationwide, it could help thousands of people avoid a premature heart attack or death from heart disease.”
More than 300,000 people in Britain each year developed severe chest pain and were given CT scans to discover whether they suffered from heart abnormalities such as arterial blockages, Antoniades said. However, less than 20% of those scanned were found to have obstructions or dangerous narrowing of their coronary arteries. “The other 80% plus show no abnormalities. They are reassured and sent home, often without medication,” says Antoniades, professor of cardiovascular medicine at the University of Oxford. “However, these reassurances are often misplaced.”
In fact, about two-thirds of this ‘safe’ group develop serious – sometimes fatal – heart problems, including heart attacks. “It is clear that we have missed signals from our scans that could tell us about those in real danger,” he added. “It’s a huge problem in healthcare and we believe AI is the perfect technology to tackle it.”
The research, led by a team from the University of Oxford’s Radcliffe Department of Medicine, and published in the University of Oxford Lancet last week, is designed to detect the abnormalities missed by standard CT (computed tomography) scans. This knowledge would allow doctors to provide patients with preventative treatments, such as anti-inflammatory medications.
A big part of the problem was that damage to an artery caused by inflammation wasn’t picked up by a CT scan, Antoniades said. “Our discovery was finding a way to surface hidden information by using AI to enhance our CT scan images to show what damage has been done. We couldn’t imagine this before, but now we can.”
The technique uses data on the characteristics of coronary plaques, as well as changes in the fat around inflamed arteries, to provide important information about the health status of our heart arteries. “Essentially, these measurements tell us what the absolute risk is of a patient developing a fatal heart condition in the next decade,” he said.
These risk factors were originally developed from US case studies, but the data has since been evaluated in 40,000 patients in UK hospitals.
after newsletter promotion
“What we found was clear. Patients who were shown to have high inflammation in their coronary arteries were also found to be at extremely high risk for serious heart disease, such as heart attacks. We have found a way to detect the hidden factors that lead to heart attacks.
The study, funded by the British Heart Foundation, found that in 45% of cases, doctors decided to change a patient’s treatment based on data provided by AI analysis. These treatments include giving high doses of statins or medications such as colchicine, which are known to reduce the risk of cardiovascular disease.
Antoniades added: “We also plan to expand supply of this British-made technology to the US, where it is also being evaluated by the Food and Drug Administration, and in Europe, where it has already been approved for clinical use.”