Predicting vulnerability status of carotid plaques using CTA-based quantitative analysis

The study aimed to develop a radiomics model to assess carotid artery plaque vulnerability using CTA images. It retrospectively included 107 patients with carotid artery stenosis who underwent carotid artery stenting (CAS) from 2017 to 2022. Patients were categorized into stable and vulnerable plaque groups based on pathology. A training group and a testing group were formed in a 7:3 ratio. Clinical data, including demographics and lipid profiles, were collected alongside pre-treatment CTA images. Radiomics features were extracted and reduced using the LASSO method to minimize redundancy. A radiomics model was then constructed, utilizing 13 features with a minimum penalty coefficient logλ=0.047. Significant differences were found between stable and vulnerable plaques in terms of stenosis degree. The radiomics model showed high accuracy (AUC of 0.959 in training and 0.942 in testing) for identifying vulnerable plaques. When combined with clinical parameters stenosis degree, the model's diagnostic efficacy improved further, with AUC values of 0.985 and 0.961 in the training and testing groups, respectively. Decision curve analysis (DCA) indicated that the combined model offered superior clinical benefits over the clinical model and radiomics model alone. The study concludes that the combined radiomics model, incorporating stenosis degree, presents a promising tool for differentiating vulnerable from stable plaques.