基于随机森林和Logistic回归的大脑半球凸面脑膜瘤显微镜术后继发性癫痫影响因素分析

Factors Influencing Secondary Epilepsy after Microscopic Surgery for Cerebral Convexity Meningiomas Based on Random Forest and Logistic Regression

目的探讨大脑半球凸面脑膜瘤显微镜术后继发性癫痫影响因素,建立多因素Logistic回归模型、随机森林模型,对比两种模型对显微镜术后继发性癫痫的预测效能。方法选取2020年1月—2023年1月收治的90例大脑半球凸面脑膜瘤显微镜术后继发性癫痫作为研究组,同期90例大脑半球凸面脑膜瘤显微镜术后无继发性癫痫作为对照组。采用单纯随机抽样方法,分别从研究组、对照组抽取80%样本量作为训练集,剩余20%作为测试集。统计2组人口学特征、实验室指标,分别构建多因素Logistic回归模型、随机森林模型,采用受试者工作特征(ROC)曲线及曲线下面积(AUC)评价两种模型对显微镜术后继发性癫痫的预测效能。结果研究组血清同型半胱氨酸(Hcy)、S100钙结合蛋白B(S100B)、神经元特异性烯醇化酶(NSE)水平高于对照组,血清甘丙肽(GAL)水平低于对照组(P<0.05);研究组术后瘤腔出血、术中神经电生理监测异常、术后脑积水占比高于对照组(P<0.01,P<0.05);多因素Logistic回归分析显示,Hcy、NSE、S100B、术后脑积水、术中神经电生理监测异常为显微镜术后继发性癫痫影响因素(P<0.01);随机森林模型分析显示,不同变量中排名前5位依次为NSE、S100B、Hcy、术后脑积水、术中神经电生理监测异常;随机森林模型预测显微镜术后继发性癫痫的AUC大于多因素Logistic回归模型(P<0.05)。结论Hcy、NSE、S100B、术后脑积水、术中神经电生理监测异常为大脑半球凸面脑膜瘤显微镜术后继发性癫痫影响因素,基于上述因素构建的多因素Logistic回归模型可直观解释不同变量对显微镜术后继发性癫痫发生的风险度,而随机森林模型对显微镜术后继发性癫痫的预测效能更优,有助于临床制订防治措施。

Objective To investigate the influencing factors of secondary epilepsy after microscopic surgery for cerebral convexity meningioma,to establish multivariate Logistic regression model and random forest model,and to compare the predictive efficacy of the two models for postoperative secondary epilepsy.Methods A total of 90 patients with secondary epilepsy after microscopic surgery for convex cerebral convexity meningioma treated from January 2020 to January 2023 were selected as the research group,and 90 patients without secondary epilepsy after microscopic surgery for cerebral convexity meningioma during the same period were selected as the control group.Using simple random sampling method,80%sample size was randomly selected from the research group and the control group respectively as the training set,and the remaining 20% was used as the test set.Demographic characteristics and laboratory indicators of the two groups were calculated,and multivariate Logistic regression model and random forest model were constructed,respectively.Receiver operating characteristic(ROC)curve and area under the ROC curve(AUC)were used to evaluate the predictive efficacy of the two models for postoperative secondary epilepsy.Results The levels of homocysteine(Hcy),S100 calcium-binding protein B(S100B)and neuron specific enolase(NSE)in the research group were higher than those in the control group,while the levels of galanin(GAL)were lower than those in the control group(P<0.05).The proportion of postoperative lumen hemorrhage,abnormal neuroelectrophysiological monitoring during surgery and postoperative hydrocephalus in the research group was higher than that in the control group(P<0.01,P<0.05).Multivariate Logistic regression analysis showed that Hcy,NSE,S100B,postoperative hydrocephalus and abnormal neuroelectrophysiological monitoring during surgery were the influential factors for postoperative secondary epilepsy(P<0.01).Random forest analysis showed that the top 5 variables were NSE,S100B,Hcy,postoperative hydrocephalus,and abnormal neuroelectrophysiological monitoring during surgery.The AUC of the random forest model for predicting postoperative secondary epilepsy was greater than that of Logistic regression model(P<0.05).Conclusion Hcy,NSE,S100B,postoperative hydrocephalus,and abnormal neuroelectrophysiological monitoring during surgery are the factors that affect the incidence of secondary epilepsy after microscopic surgery for cerebral convexity meningioma.A multivariate Logistic regression model constructed based on these factors can visually explain the risk of different variables for the occurrence of secondary epilepsy after surgery.The random forest model has better predictive efficacy for postoperative secondary epilepsy,which helps to formulate prevention and treatment measures in clinical practice.