AI spots hidden heart attack risk in routine chest scans, study finds

Artificial intelligence can now spot early warning signs of heart attack risk simply by studying the density of muscles in a patient's chest and back, according to a new study that used AI to analyse scans from more than 1,700 patients.

The research, led by the University of Edinburgh and published in the journal Radiology, used an AI tool to examine standard coronary computed tomography angiogram (CCTA) scans — the type of scan around 350,000 people undergo in the UK each year to check for blockages in the arteries supplying blood to the heart.

Rather than only checking for arterial narrowing, the AI system was trained to look beyond the heart itself, analysing muscle, bone, organs and fat in the upper body. In particular, it measured "skeletal muscle attenuation" — essentially how bright or dark muscle tissue appears on a scan. Denser, higher-quality muscle reflects more X-rays and shows up brighter, typically containing less fat, a factor previously linked to poorer heart health.

The standout advantage of the AI approach is speed. What would take a radiologist several hours to measure by hand, the AI completed in under a minute per scan — raising the possibility that muscle-quality analysis could be added to routine cardiac scans without slowing down clinical workflows.

Studying patients mostly in their 50s who had presented with chest pain, researchers tracked outcomes over the following decade. They found that for every 10-point increase in muscle brightness on the scan, a patient was 31 per cent less likely to suffer a heart attack and 39 per cent less likely to die within ten years. These associations held even after accounting for age, sex and the level of calcium build-up in the arteries.

Notably, the size of a patient's muscles made no difference to their risk. It was muscle composition — not bulk — that correlated with better outcomes, suggesting that any form of exercise capable of improving muscle quality, not just strength training, could play a protective role.

Researchers believe people with denser, higher-quality torso muscle tend to be more physically active generally, and that this activity likely protects the heart through multiple pathways rather than muscle density alone being directly responsible.

Professor Michelle Williams, the study's senior author, said the results had personally prompted her to start exercising more, including planks, pilates and cycling — activities she believes may engage the back, pectoral and intercostal muscles picked up in CCTA scans. She cautioned, however, that considerably more research is needed to understand exactly how exercise influences muscle density and how that, in turn, affects heart health.

Professor Bryan Williams of the British Heart Foundation, which part-funded the study, said the findings offered further evidence of the protective power of physical activity, noting that more active people in the study tended to have denser muscle and better heart health as a result.

Looking ahead, researchers say the AI tool could eventually help doctors flag patients with lower-quality muscle during scans they are already having, allowing those at greater risk to be prioritised for closer monitoring, lifestyle advice or preventive medication such as statins. They stressed, though, that further research is required before the technology could be relied upon for that purpose in clinical practice.