Computed tomography angiography-based radiomics model for predicting carotid atherosclerotic plaque vulnerability

• DOI: 10.3389/fneur.2023.1151326
• Publication Date: 2023-06-16T05:15:34Z
• Authors: Dezhi Shan; Siyu Wang; Junjie Wang; Jun Lu; Junhong Ren; Juan Chen; Daming Wang; Peng Qi
• Keywords: Vulnerable plaque; Computed Tomography Angiography

Vulnerable carotid atherosclerotic plaque (CAP) significantly contributes to ischemic stroke. Neovascularization within plaques is an emerging biomarker linked vulnerability that can be detected using contrast-enhanced ultrasound (CEUS). Computed tomography angiography (CTA) a common method used in clinical cerebrovascular assessments employed evaluate the of CAPs. Radiomics technique automatically extracts radiomic features from images. This study aimed identify associated with neovascularization CAP and construct prediction model for based on features. CTA data patients CAPs who underwent CEUS between January 2018 December 2021 Beijing Hospital were retrospectively collected. The divided into training cohort testing 7:3 split. According examination CEUS, dichotomized vulnerable stable groups. 3D Slicer software was delineate region interest images, Pyradiomics package extract Python. Machine learning algorithms containing logistic regression (LR), support vector machine (SVM), random forest (RF), light gradient boosting (LGBM), adaptive (AdaBoost), extreme (XGBoost), multi-layer perception (MLP) models. confusion matrix, receiver operating characteristic (ROC) curve, accuracy, precision, recall, f-1 score performance A total 74 110 included. In all, 1,316 extracted, 10 selected machine-learning construction. After evaluating several models cohorts, it discovered model_RF outperformed others, achieving AUC value 0.93 (95% CI: 0.88-0.99). 0.85, 0.87, respectively. Radiomic obtained. Our highlights potential radiomics-based improving accuracy efficiency diagnosing CAP. particular, model_RF, utilizing extracted CTA, provides noninvasive efficient accurately predicting status shows great offering guidance early detection patient outcomes.