AI-derived carotid plaque vulnerability from B-mode ultrasound for cerebrovascular risk stratification: a multicenter cohort study

Routine carotid B-mode ultrasound is widely used for plaque assessment, but vulnerability evaluation remains subjective, and its prognostic value is uncertain. We developed and validated a cascaded multitask AI framework for plaque detection, characterization of five B-mode features, and vulnerability assessment. The multicenter cohort included 6618 participants at high cardiovascular risk and 38,090 images from four centers.

Plaque presence and B-mode features were evaluated against expert B-mode consensus, whereas vulnerability was evaluated against multimodal expert consensus. Internal and external AUCs were 0.95 and 0.96 for plaque detection, 0.85–0.94 and 0.81–0.92 for feature characterization, and 0.90 and 0.88 for vulnerability classification. The model outperformed six independent B-mode readers and showed greater net benefit in decision-curve analysis.

Prognostic validation included 2174 participants; 277 ischemic cerebrovascular events occurred over a median of 37 months. AI-defined high-risk status remained independently associated with events after adjustment for clinical risk factors (hazard ratio, 2.09; 95% confidence interval, 1.69–2.58), and adding the AI score improved the C-index from 0.769 to 0.795. These findings support standardized plaque vulnerability assessment and clinically meaningful cerebrovascular risk stratification from routine B-mode ultrasound.

This work was supported by the Zhejiang Provincial Natural Science Foundation (grant KLY25H180038), the Wenzhou Science and Technology Project (grant Y20220071), and the Zhejiang Provincial Medical and Health Science and Technology Plan Project (grant 2021KY1079). The funders had no role in the study design; data collection, analysis, or interpretation; preparation of the manuscript; or the decision to submit the manuscript for publication. These authors contributed equally: Qi Xu, Huanhuan Ding.

Department of Ultrasound, Wenzhou Third Clinical Institute Affiliated to Wenzhou Medical University, Wenzhou People’s Hospital, Wenzhou, China Qi Xu, Huanhuan Ding, Xiaoli Ji, Xiang Zhang & Zengqiao Lin Department of Ultrasound, The Second Affiliated Hospital and Yuying Children’s Hospital of Wenzhou Medical University, Wenzhou, China Wenzhou Key Laboratory of Structural and Functional Imaging, Wenzhou Municipal Science and Technology Bureau, Wenzhou, China Department of Radiology, The Second Affiliated Hospital and Yuying Children’s Hospital of Wenzhou Medical University, Wenzhou, China Department of Ultrasound, Dongtou District People’s Hospital, Wenzhou, China Center for Cardiovascular Disease, The Affiliated Suzhou Hospital of Nanjing Medical University, Suzhou Municipal Hospital, Gusu School, Nanjing Medical University, Suzhou, China Department of Cardiology, Wenzhou Third Clinical Institute Affiliated to Wenzhou Medical University, Wenzhou People’s Hospital, Wenzhou, China Correspondence to Wenbing Jiang or Zhihan Yan. The authors declare no competing interests. Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Open Access This article is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License, which permits any non-commercial use, sharing, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if you modified the licensed material. You do not have permission under this licence to share adapted material derived from this article or parts of it. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material.