基于随机森林算法的颈动脉支架植入术后残留预测模型构建分析

目的 探讨基于随机森林算法的颈动脉支架植入术(Carotid artery stent, CAS)后残留预测模型构建。方法 回顾性选取2018年10月-2021年10月于本院接受CAS治疗的颈动脉狭窄患者181例作为研究对象，根据术后残留狭窄情况分为残留狭窄组(狭窄率≥30%)和非残留狭窄组(狭窄率<30%);比较2组临床资料，采用多因素Logistic回归分析和随机森林算法分别构建影响CAS后残留狭窄形成的2个预测模型，比较2个预测模型的预测效能。结果 术后残留狭窄发生51例(28.18%)归为残留狭窄组，其余130例归为非残留狭窄组。2组术前体质量指数(Body mass index, BMI)、年龄、吸烟史、高血压病占比、术前狭窄处血管内径、支架类型、斑块形态、斑块钙化情况比较均有明显差异(P<0.05)。多因素Logistic回归分析显示，术前狭窄处血管内径(OR=0.012,95%CI=0.001～0.114)为CAS后残留狭窄的保护因素，高血压病(OR=1.057,95%CI=1.035～1.079)、闭环支架(OR=2.773,95%CI=1.067～7.20...

Construction and analysis of prediction model for residual stenosis after carotid artery stenting based on random forest algorithm

ObjectiveTo construct the prediction model for residual stenosis after carotid artery stenting(CAS)based on random forest algorithm.Methods A total of 181 patients with carotid artery stenosis who received CAS in our hospital from October 2018 to October 2021 were retrospectively selected as the research subjects, and they were divided into residual stenosis group(stenosis rate ≥30%)and non-residual stenosis group(stenosis rate <30%)according to the degree of residual stenosis after CAS. The clinical data of these two groups were collected, and multivariate logistic regression analysis and random forest algorithm were respectively applied to construct two predictive models for predicting occurance of residual stenosis after CAS. The predictive performances of these two models were assessed.Results 51 cases(28.18%)of postoperative residual stenosis were identified as the residual stenosis group, while the remaining 130 cases were identified as the non-residual stenosis group. There were significant differences between the two groups in preoperative body mass index(BMI), age, smoking, proportion of hypertension, diameter of preoperative stenosis artery, type of stent, plaque morphology, and plaque calcification(P<0.05).Multivariate logistic regression analysis showed that the diameter of preoperative stenosis artery(OR=0.012, 95%CI=0.001～0.114)was a protective factor for residual stenosis after CAS, but hypertension(OR=1.057, 95%CI=1.035～1.079), closed-loop stent(OR=2.773, 95%CI=1.067～7.202), irregular plaque surface(OR=2.698, 95%CI=1.079～6.750)and plaque calcification(OR=5.488, 95%CI=2.073～14.525)were risk factors for residual stenosis after CAS(P<0.05). All variables ranked by importance in the random forest algorithm were the diameter of preoperative stenosis artery, plaque morphology, plaque calcification, hypertension, type of stent, BMI, age, and smoking. The results showed that the diagnostic performance of the prediction model based on random forest algorithm(area under the curve(AUC)was 0.884)was better than that based on multivariate logistic regression analysis(AUC was 0.821).Conclusion The prediction model based on random forest algorithm could more effectively predict the risk of residual stenosis after CAS. The diameter of preoperative stenosis artery, plaque morphology, plaque calcification, hypertension, and the type of stent are the risk factors for predicting residual stenosis after CAS.