Development of an artificial intelligence model to quantify plaque vulnerability and predict plaque events using routine computed tomography angiography

A routine CT scan of the heart’s arteries already contains hidden clues about which plaques are dangerous and likely to rupture, and a new artificial intelligence model aims to read those clues. This matters because most people who undergo CT coronary angiography are told they have no significant blockages—yet they still suffer heart attacks at higher rates than those with obvious disease. The problem is that standard scans cannot distinguish stable plaques from inflamed, rupture-prone ones. The researcher proposes to combine two sources of information from the same scan: the pattern of inflammation in the fat surrounding each artery segment, and the detailed shape and texture of the adjacent plaque itself. By feeding these data into a machine learning model trained on real patient outcomes, the model will learn to assign a “plaque vulnerability score” to each vessel segment. If successful, this approach could transform how doctors interpret routine CT scans. Instead of simply ruling out obstructive disease, the scan would flag specific high-risk plaques before they cause a heart attack. That would allow cardiologists to target preventive treatments—such as statins or anti-inflammatory drugs—to the patients who need them most, rather than treating everyone with a normal-looking scan the same way. The work is applied, not fundamental: it directly aims to change clinical practice by extracting new meaning from an existing, widely used test.