急性缺血性卒中患者静脉溶栓后3个月不良预后的机器学习预测模型构建OA
Construction of machine learning model for predicting 3-month poor prognosis after intravenous thrombolysis in patients with acute ischemic stroke
目的 构建急性缺血性卒中(AIS)患者重组组织型纤溶酶原激活剂(rt-PA)静脉溶栓后3个月不良预后的预测模型,并分析相关影响因素.方法 选取396例发病4.5 h内接受rt-PA静脉溶栓治疗的AIS患者为研究对象.以3个月改良Rankin量表(mRS)评分3~6分作为不良预后的界定标准.将患者按7∶3比例分为训练集和测试集,分别建立最小绝对收缩与选择算子(LASSO)回归、逻辑回归(LR)和随机森林(RF)模型,并采用10折交叉验证进行参数优化.采用受试者工作特征曲线的曲线下面积(AUC)、校准分析、Brier分数、决策曲线分析(DCA)以及SHAP值评估、解释模型性能与变量贡献.结果 rt-PA静脉溶栓治疗3个月后,不良预后患者235例,占59.3%.测试集中,RF模型AUC为0.886,且在准确率、F1值、Brier分数和临床净获益方面表现较优.24 h美国国立卫生研究院卒中量表(NIHSS)评分、入院NIHSS评分、D-二聚体和年龄是主要预测因素.结论 机器学习模型可用于AIS患者静脉溶栓后3个月不良预后的早期评估,其中RF模型综合表现较优.
Objective To construct predictive models for 3-month poor prognosis after intrave-nous thrombolysis with recombinant tissue plasminogen activator(rt-PA)in patients with acute ische-mic stroke(AIS)and to analyze the related influencing factors.Methods A total of 396 patients with AIS who received intravenous thrombolysis with rt-PA within 4.5 hours of symptom onset were enrolled.A 3-month modified Rankin Scale(mRS)score of 3 to 6 was defined as poor prognosis.The patients were divided into training set and test set at a ratio of 7:3.Least absolute shrinkage and se-lection operator(LASSO)regression,logistic regression(LR)and random forest(RF)models were constructed,and 10-fold cross-validation was used for parameter optimization.The area under the re-ceiver operating characteristic curve(AUC),calibration analysis,Brier score,decision curve analy-sis(DCA)and SHAP values were used to evaluate and interpret model performance and variable con-tributions.Results At 3 months after rt-PA intravenous thrombolysis,235 patients(59.3%)had poor prognosis.In the test set,AUC of the RF model achieved 0.886 and demonstrated superior performance in terms of accuracy,F1 score,Brier score and clinical net benefit.The 24-h National Institutes of Health Stroke Scale(NIHSS)score,NIHSS score at admission,D-dimer level and age were the main predictive factors.Conclusion Machine learning models can be used for early assessment of 3-month poor prognosis after intravenous thrombolysis in patients with AIS,among which the RF model exhibited the best overall performance.
王淼淼;郑强;王晓蕊;李梦瑶;郑怡情;薄静如;郭惠恩;王子康;骆嵩
蚌埠医科大学第一附属医院神经内科,安徽 蚌埠,233004蚌埠医科大学第一附属医院神经内科,安徽 蚌埠,233004郑州市第八人民医院急诊医学科,河南郑州,450000蚌埠医科大学第一附属医院神经内科,安徽 蚌埠,233004蚌埠医科大学第一附属医院神经内科,安徽 蚌埠,233004蚌埠医科大学第一附属医院神经内科,安徽 蚌埠,233004蚌埠医科大学第一附属医院神经内科,安徽 蚌埠,233004蚌埠医科大学第一附属医院神经内科,安徽 蚌埠,233004蚌埠医科大学第一附属医院神经内科,安徽 蚌埠,233004
医药卫生
缺血性卒中静脉溶栓机器学习模型重组组织型纤溶酶原激活剂随机森林模型不良预后改良Rankin量表美国国立卫生研究院卒中量表
ischemic strokeintravenous thrombolysismachine learning modelrecombinant tissue plasminogen activatorrandom forest modelpoor prognosismodified Rankin ScaleNational Institutes of Health Stroke Scale
《实用临床医药杂志》 2026 (15)
73-78,85,7
安徽省教育厅重点项目(2022AH051480)安徽省"江淮名医"培育工程项目(2022)蚌埠市科技创新指导类项目(20230131)
评论