Construction of a prediction model for large-artery atherosclerosis ischemic stroke based on blood biomarkers and ultrasonic imaging

The global burden of ischemic stroke (IS) continues to rise annually. This study aims to develop a machine learning prediction model that integrates blood biomarkers and carotid color Doppler ultrasound features to identify high-risk patients with large-artery atherosclerosis (LAA)-type IS among those with atherosclerosis (AS). A retrospective analysis was conducted involving 166 patients with LAA-type IS and 71 patients with AS. The baseline characteristics, blood biomarkers results and carotid color Doppler ultrasonic imaging features of the patients at admission were collected. Multivariate binary Logistic regression was used to identify the independent influencing factors of LAA-type IS, and logistic regression (LR), random forest (RF) and support vector machine (SVM) prediction models were established. Receiver operating characteristic (ROC) curve, calibration curve, decision curve analysis (DCA) and Delong test were used to evaluate and compare the models. Internal validation of the optimal model was performed using the Bootstrap method; no external validation was carried out. Eight independent predictors of LAA-type IS were identified, which were carotid plaque (CP) presence, CP echogenicity, thrombomodulin (TM), platelet distribution width (PDW), glucose (GLU), apolipoprotein A (APA), homocysteine (HCY), and hydroxybutyric dehydrogenase (HBDH). The area under the ROC curve (AUC) of LR, SVM and RF model were 0.846, 0.901 and 0.955, respectively. The Delong test showed statistically significant differences among the three models (p < 0.05). The calibration curve and DCA results further showed good validity and clinical feasibility of all models. Among them, the RF model exhibited the best performance, with a C-index of 0.955 upon internal validation after 200 bootstrap iterations. The RF model shows promising potential for predicting LAA-type IS and may serves as a reference for clinicians to rapidly identify high-risk patients with this stroke subtype.