AI tool could eradicate heart attacks, predicting risk a decade early

A cardiologist in Britain has spent years training an artificial intelligence system to see what human eyes cannot: the invisible markers of a heart attack waiting to happen. Professor Charalambos Antoniades, backed by the British Heart Foundation, has developed CaRi-Heart, a tool that examines CT scans of the heart and extracts patterns too subtle for clinicians to detect. The system can estimate a person's risk of suffering a heart attack within the next decade—years before symptoms ever appear, years before the damage becomes irreversible. He wants to be remembered as the doctor who killed heart attacks. The research suggests he might succeed.

The scale of what he is trying to prevent is enormous. Heart attacks send 100,000 people to NHS hospitals every year. They account for 290 admissions daily across the United Kingdom. Cardiovascular disease costs the health service £7.4 billion annually. Yet for decades, doctors have faced a fundamental problem: they could not reliably identify who was truly at risk. Treatment remained reactive, a response to crisis rather than a shield against it. The AI changes that equation. It looks at the arteries supplying blood to the heart and calculates risk by analyzing something called a Fat Attenuation Index—a measurement of tissue composition that appears meaningless to the human eye but carries profound predictive weight.

The discovery emerged from a striking observation. Antoniades and his team studied 40,000 people in the UK whose cardiac CT scans had all been reported as completely normal by clinicians. Within three years, two-thirds of the heart attacks in that group occurred in people whose scans had shown no visible abnormality. Around 85 percent of all cardiac CT scans performed in the NHS are reported as normal. Yet buried within that vast cohort of apparently healthy patients are thousands of people at significant risk. The AI finds them. It sees what is there but invisible.

Paul Randall, a 45-year-old planning manager from Leighton Buzzard in Buckinghamshire, became one of those people. He had grown up watching heart disease claim his family. His uncle died of heart complications at three years old. His grandfather suffered a fatal heart attack at 42. In early 2024, at 43, Randall began experiencing chest pain and dizziness—sometimes while running, sometimes while sitting at his desk, once while walking his dog Roxy through the woods with no one nearby. He did not make a fuss. He simply continued. A visit to his GP revealed high cholesterol, which led to an ECG and a CT scan. When Antoniades' AI analyzed that scan, it found three things: a 20 percent blockage in the arteries of his heart, a mitral valve leak, and a slight narrowing of his aortic valve. His Fat Attenuation Index placed him in the 76th percentile. The AI calculated he faced a significant risk of a cardiac event within eight years.

The diagnosis became a catalyst. Randall changed his life. He extended his sleep from five hours a night to a full schedule. He began running five miles three to four times a week and cycling with his daughter. He reduced red meat to once a week, cut alcohol, and lost weight until he fell into the healthy BMI range. He felt the weight of responsibility—not just to himself, but to the future he wanted to see. He wanted to watch his daughter grow up. He wanted to help with grandchildren. The scan, he said, made his plan for protecting his own heart so much clearer.

If the technology were deployed today across all 350,000 cardiac CT scans performed annually in the NHS, the modeling suggests extraordinary impact. Within five years, it could prevent 1,700 cardiac deaths, 2,700 heart attacks, 1,600 cases of heart failure, and strokes—roughly 6,000 major adverse cardiac events prevented. The NHS found that in five pilot hospitals, the technology changed clinical management in about half of all patients tested. An 11 percent reduction in cardiac deaths. A 10 percent fall in non-fatal heart attacks. A 4 percent decline in strokes and a 4 percent reduction in heart failure cases. Overall, a 30 percent reduction in major cardiovascular events. The financial savings alone—given that cardiovascular disease costs the system £7.4 billion yearly—could be substantial.

The mechanism is straightforward in concept but revolutionary in practice. Once doctors know who is at risk, they can intervene. They can prescribe preventative medications tailored to the individual. They can counsel lifestyle changes before the crisis arrives. They can shift medicine from reaction to prevention. Antoniades describes what the AI accomplishes as remarkable: it detects something clinicians cannot see. The scans appear completely normal. Yet within that appearance of normalcy lies genuine danger, waiting to be found and addressed.

The next hurdle is approval from NICE, the National Institute for Health and Care Excellence. The technology is currently under review. If NICE includes it in formal clinical guidelines, deployment across the NHS could follow. For Randall and thousands like him—people at risk but unaware, people whose scans appear normal but whose hearts are vulnerable—that approval cannot come soon enough. The tool offers something medicine has long promised but rarely delivered: the chance to prevent catastrophe before it strikes.