Many ischemic strokes do not start in the brain. They begin with an unstable plaque in the carotid artery, the vessel carrying blood from the heart to the head. When that deposit ruptures, debris travels upward and chokes off circulation.
The catch is that most plaques are harmless. Distinguishing the vulnerable ones, marked by intraplaque hemorrhage, a lipid-rich core and thin inflamed caps, has long depended on the eye and experience of a radiologist. A study in BMC Medicine shows how much of that judgment AI can now shoulder.
The team, led by Xuanwu Hospital of Capital Medical University in Beijing, built a fully automated pipeline called CFAPS. It reads high-resolution vessel wall MRI scans with multiple contrast weightings and, without human input, traces the plaque boundary and labels it vulnerable or stable.
A four-center stress test
The evaluation drew on 1,610 carotid arteries from 1,315 patients scanned between January 2019 and July 2025 across four centers, with collaborators at Shandong Provincial Hospital, Peking University Shenzhen Hospital, Weifang Traditional Chinese Hospital and the University of Science and Technology Beijing.
The clinical stakes are concrete. Patients flagged as high risk may warrant surgery or aggressive medical therapy; those with stable plaques can be spared it. An automated, consistent read could reduce the variability that today sends similar patients down different paths.
The system is a research pipeline, not a cleared product, and prospective trials remain ahead. But it shows the diagnostic bottleneck in stroke prevention may be narrowing to a matter of computation.
