A Radiomics-Based Approach with Automated Segmentation for Identifying Symptomatic Basilar Artery Plaques in Acute Stroke.
Authors
Affiliations (8)
Affiliations (8)
- Lauterbur Research Center for Biomedical Imaging, Shenzhen Institutes of Advanced Technology, Chinese Academy of Sciences, Shenzhen, 518055, China; Department of Biomedical Engineering, Chongqing University of Technology, Chongqing, China.
- Department of radiology, The First Affiliated Hospital of Nanjing Medical University, NO.300 Guang Zhou Road, Gulou district, Nanjing, Jiangsu Province, 210029, China.
- Lauterbur Research Center for Biomedical Imaging, Shenzhen Institutes of Advanced Technology, Chinese Academy of Sciences, Shenzhen, 518055, China.
- Department of Image Advanced Analysis of HSW BU, Shanghai United Imaging Healthcare Co., Shanghai 201800, China.
- Department of Biomedical Engineering, Chongqing University of Technology, Chongqing, China.
- Lauterbur Research Center for Biomedical Imaging, Shenzhen Institutes of Advanced Technology, Chinese Academy of Sciences, Shenzhen, 518055, China; State Key Laboratory of Biomedical Imaging Science and System, Key Laboratory of Biomedical Imaging Science and System, Chinese Academy of Sciences, Shenzhen, China.
- Department of radiology, The First Affiliated Hospital of Nanjing Medical University, NO.300 Guang Zhou Road, Gulou district, Nanjing, Jiangsu Province, 210029, China. Electronic address: [email protected].
- Lauterbur Research Center for Biomedical Imaging, Shenzhen Institutes of Advanced Technology, Chinese Academy of Sciences, Shenzhen, 518055, China; State Key Laboratory of Biomedical Imaging Science and System, Key Laboratory of Biomedical Imaging Science and System, Chinese Academy of Sciences, Shenzhen, China. Electronic address: [email protected].
Abstract
Intracranial atherosclerotic disease (ICAD) is a leading cause of ischemic stroke worldwide, with basilar artery atherosclerosis frequently involved. Despite therapeutic advances, patients with basilar artery atherosclerosis remain at substantial risk of recurrent stroke, highlighting the need for improved strategies to accurately identify symptomatic basilar artery plaques. In this study, we aimed to develop and validate a framework for identifying symptomatic basilar artery plaques by integrating MRI-based vessel wall segmentation with a tabular foundation model for quantitative plaque characterization. In this retrospective study, we analyzed 256 patients with basilar artery stenosis who underwent three-dimensional high-resolution vessel wall imaging (VWI) between May 2018 and November 2023. An automated convolutional neural network-based model (Vessel-SegNet) was applied to segment the basilar artery vessel wall on both pre-contrast VWI (preVWI) and contrast-enhanced VWI (ceVWI) images. Radiomics, morphological, and signal intensity features were subsequently extracted from the segmented vessel walls and used to train a Tabular Prior-Fitted Network (TabPFN) to identify symptomatic versus asymptomatic plaques. Model performance was evaluated using the area under the receiver operating characteristic curve (AUC), accuracy, sensitivity, and specificity. The model based on traditional morphological and signal intensity features achieved an AUC of 0.784 (95% CI: 0.673-0.877) for distinguishing symptomatic basilar artery plaques from asymptomatic basilar artery plaques. In contrast, the radiomics-based model, incorporating features extracted from both preVWI and ceVWI, showed a significantly improved discriminative performance, with an AUC of 0.887 (95% CI: 0.798-0.955). The proposed framework, integrating automated vessel wall segmentation with a tabular foundation model, enables accurate identification of symptomatic basilar artery plaques. This approach provides a scalable and objective tool that may support risk stratification and inform treatment planning in patients with ICAD.