基于机器学习算法的脑梗死伴颅内动脉狭窄预测模型研究

Machine learning-based models for prediction of cerebral infarction with intracranial artery stenosis

目的评价3种机器学习算法模型对脑梗死患者是否伴有颅内动脉狭窄的预测性能,为缺血性卒中诊治策略提供参考依据。方法回顾性收集自2013年1月至2019年9月喀什地区第一人民医院神经内科收治的缺血性脑卒中患者临床资料。通过特征选择筛选出脑梗死伴颅内动脉狭窄的相关因素作为预测因子,基于随机森林、决策树和神经网络3种机器学习算法建立预测模型。利用受试者工作特征曲线下面积(area under ROC,AUC)、灵敏度、准确率等指标评价3种模型对脑梗死患者是否伴有颅内动脉狭窄的预测性能。结果研究分析了2365例脑梗死患者的74种特征,通过特征选择选出23个特征作为预测因子纳入建模。决策树模型、神经网络模型、随机森林模型的AUC值分别为0.78±0.11、0.85±0.12、0.89±0.10;灵敏度依次为0.92±0.05、0.91±0.06、0.88±0.10;准确率依次为0.79±0.12、0.77±0.13、0.85±0.13。结论随机森林模型对脑梗死患者是否伴有颅内动脉狭窄的预测性能最佳,在无条件进行血管成像检查时可应用于脑梗死是否伴有颅内动脉狭窄的预测,具有一定的临床应用价值。

Objective To evaluate the predictive performance of three machine learning algorithm models for cerebral infarction patients with intracranial arterial stenosis,and to provide reference for ischemic stroke diagnosis and treatment strategies.Methods The clinical data of patients with ischemic stroke admitted to the Department of Neurology of the First People’s Hospital of Kashi Prefecture from January 2013 to September 2019 were retrospectively collected.The features of cerebral infarction with intracranial artery stenosis were identified by characteristics selection method.The prediction model was established by using the three kinds of machine learning algorithms,such as random forest,decision tree and neural network.The prediction performance of the three models on whether patients with cerebral infarction had intracranial artery stenosis was evaluated by using the area under the subject operating characteristic curve,sensitivity and accuracy.Results The study analyzed 74 characteristics of 2365 patients with cerebral infarction.By feature selection method,23 features were selected as predictors and included in the model.The AUC values of decision tree model,neural network model,and random forest model were 0.78±0.11,0.85±0.12,0.89±0.10;the sensitivity was 0.92±0.05,0.91±0.06,0.88±0.10;the accuracy rates are 0.79±0.12,0.77±0.13,0.85±0.13.Conclusions The random forest model has the best performance in predicting whether patients with cerebral infarction are accompanied by intracranial arterial stenosis.It can be used to predict whether cerebral infarction is accompanied by intracranial arterial stenosis when unconditional angiography is performed.It has certain clinical application value.