Abstract
OBJECTIVE: Carotid plaque represents an independent risk factor for cardiovascular disease and a significant threat to human health. The aim of the study is to develop an accurate and interpretable predictive model for early detection the occurrence of carotid plaque. METHODS: A cross-sectional study was conducted by selecting coal miners who participated in medical examinations from October 2021 to January 2022 at a hospital in North China. The features were initially screened using extreme gradient boosting (XGBoost), random forest, and LASSO regression, and the model was subsequently constructed using logistic regression. The three models were then compared, and the optimum model was identified. Finally, a nomogram was plotted to increase its interpretability. RESULTS: The XGBoost algorithm demonstrated superior performance in feature screening, identifying the top five features as follows: age, systolic blood pressure, low-density lipoprotein cholesterol, white blood cell count, and body mass index (BMI). The area under the curve (AUC), sensitivity, and specificity of the model constructed based on the XGBoost algorithm were 0.846, 0.867, and 0.702, respectively. CONCLUSIONS: It is possible to predict the presence of carotid plaque using machine learning. The model has high application value and can better predict the risk of carotid artery plaque in coal miners. Furthermore, it provides a theoretical basis for the health management of coal miners.