基于XGBoost算法构建颈动脉粥样硬化患病风险初筛模型

Construction of the Risk Screening Model of Carotid Atherosclerosis Based on XGBoost Algorithm

目的:分析颈动脉粥样硬化(Carotid Atherosclerosis,CAS)人群的临床指标特征,筛选CAS的最佳预测指标,建立简便有效的CAS患病风险预测模型,为普通人群CAS的早期识别及拟定筛查指标提供依据。方法:回顾性分析2017年~2018年福建省人民医院健康体检中心的9405份体检数据,选取其中行颈部血管彩超等检查的870位体检者为研究对象,分为CAS组和非CAS组,分析两组人群的一般临床资料和实验室检查资料。采用特异性、灵敏度、F1值、曲线下面积、查准率-查全率曲线下面积作为评价指标,对比逻辑回归、随机森林和XGBoost算法预测模型评估CAS患病风险的优劣。结果:(1) XGBoost模型在测试集上的评价指标均最高。(2)对于识别CAS,临床特征中贡献度最大的是年龄,其次是收缩压、肌酐。结论:基于XGBoost模型识别CAS的患病风险效果最佳,该模型能较好的CAS患病风险并对普通人群进行CAS患病风险的初筛。

Objective:By analysing clinical characteristics of carotid atherosclerosis,to screen the best predictors of carotid atherosclerosis and establish of a simple and effective risk prediction model for carotid atherosclerosis.So as to provide basis for early identification of carotid atherosclerosis in the general population and development proposal of screening indicators.Methods:A retrospective analysis of 9405 physical examination data from the Health Examination Center of Fujian Provincial People's Hospital from 2017 to 2018 was conducted.870 physical examiners who underwent color Doppler ultrasound and other examinations of neck vessels were selected as the study subjects,and were divided into carotid atherosclerosis group and non-carotid atherosclerosis group.The general clinical data and laboratory examination data of the two groups were analyzed.The specificity,sensitivity,F1 value,the area under the receiver operating characteristic curve(AUC) and the area under the precision-recall curve(AUPRC) were used as evaluation indicators to compare the advantages and disadvantages of logistic regression,random forest and XGBoost algorithm prediction models to evaluate the risk of carotid atherosclerosis.Results:(1)XGBoost model has the highest evaluation index in the test set.(2)For the identification of carotid atherosclerosis,age is the largest contributor to clinical characteristics,followed by systolic blood pressure and creatinine.Conclusion:The best effect is to identify the risk of carotid atherosclerosis based on XGBoost model.This model can better identify the risk of CAS and carry out the preliminary screening of CAS risk for the general population.