基于eStroke国家溶取栓影像平台的随机森林模型预测醒后卒中预后的价值

Prognostic value of random forest model based on eStroke National Thrombolysis Imaging Platform for post-awakening stroke

目的探讨基于eStroke国家溶取栓影像平台的随机森林模型预测醒后卒中(WUS)预后的价值。方法选取2019-01—2023-03广西医科大学第七附属医院285例WUS患者为研究对象,根据取栓治疗后90 d改良Rankin量表(mRS)评分分为预后良好组和预后不良组。统计2组临床资料及eStroke国家溶取栓影像平台自动反馈定量资料(缺血半暗带体积、梗死核心区体积),构建随机森林模型和Lasso-Logistic回归模型,以Lasso-Logistic回归模型为参照,评价随机森林模型预测WUS患者不良预后风险的价值。结果2组入院时Hcy、WBC、DSA-CS评分、NIHSS评分及取栓次数、穿刺至再通时间、颅内血管狭窄程度、房颤、吸烟史、静脉溶栓、缺血半暗带体积、梗死核心区体积比较差异均有统计学意义(P<0.05);缺血半暗带体积、梗死核心区体积、入院时Hcy水平、入院时NIHSS评分、入院时DSA-CS评分、颅内血管狭窄程度、静脉溶栓、房颤是WUS患者不良预后的影响因素(P<0.05);Logistic回归模型与随机森林模型预测WUS患者不良预后风险的AUC比较差异无统计学意义(0.891,95%CI:0.875~0.902比0.900,95%CI:0.894~0.923)。结论基于eStroke国家溶取栓影像平台的随机森林模型可用于WUS患者早期预后的预测评估,为临床针对性展开后续治疗提供参考,以改善患者预后。

Objective To investigate the prognostic value of random forest model based on clinical data and eStroke National Thrombolysis Imaging Platform in predicting post-awakening stroke(WUS),in order to provide reference for early prognosis prediction and intervention planning.Methods A total of 285 patients with WUS in the Seventh Affiliated Hospital of Guangxi Medical University from January 2020 to January 2023 were selected as subjects.According to the modified Rankin scale(mRS)score 90 days after thromposectomy,the patients were classified into good prognosis group and poor prognosis group.The clinical data of the two groups and the quantitative data(volume of ischemic penumbra and volume of infarction core area)automatically fed back by eStroke National Thrombolysis Imaging Platform were collected.The random forest model and Lasso-Logistic regression model were constructed.The risk of poor prognosis in WUS patients predicted by random forest model was evaluated.Results There were significant differences in Hcy,WBC,DSA-CS score,rLMC score,NIHSS score,thrombolysis times,puncture to recirculation time,intracranial stenosis degree,atrial fibrillation,smoking history,intravenous thrombolysis,ischemic penumbral zone volume and infarct core area volume between the two groups at admission(P<0.05).Ischemic semidark zone volume,infarct core volume,Hcy level on admission,NIHSS score on admission,DSA-CS score on admission,degree of intracranial stenosis,intravenous thrombolysis,and atrial fibrillation as factors influencing poor prognosis in patients with WUS(P<0.05).There was no significant difference in AUC between Logistic regression model and random forest model in predicting the risk of adverse prognosis in WUS patients(0.891,95%CI:0.875-0.902 vs 0.900,95%CI:0.894-0.923).Conclusion The random forest model based on clinical data and eStroke National Thrombolysis Imaging Platform can be used to predict the early prognosis of WUS patients,and provide reference for clinical follow-up treatment,so as to improve the prognosis.