首页|期刊导航|机器人外科学杂志(中英文)|机器学习早期识别急性前循环大血管闭塞性脑梗死静脉溶栓获益人群的简化模型构建

机器学习早期识别急性前循环大血管闭塞性脑梗死静脉溶栓获益人群的简化模型构建OA

Construction of a simplified machine learning model for early identification of intravenous thrombolysis benefit population in acute anterior circulation large vessel occlusive ischemic stroke

中文摘要英文摘要

目的:构建简化可解释的机器学习模型,早期识别急性前循环大血管闭塞性脑梗死静脉溶栓后 6 个月获益个体.方法:回顾性纳入 2022 年 6 月—2025 年 6 月徐州医科大学附属医院及睢宁县人民医院收治的急性前循环大血管闭塞性脑梗死患者 134 例,按治疗方式分为静脉溶栓组(n=62)和非溶栓组(n=72).收集患者一般资料,采用倾向评分重叠加权控制治疗选择偏倚.基于加权样本,通过纳入治疗交互项的惩罚Logistic 回归筛选变量,构建简化个体化获益模型,计算每例患者在溶栓与未溶栓情景下达到 6 个月良好功能结局[改良 Rankin 量表(mRS)评分≤4 分]的绝对概率差值(ΔP),以 ΔP≥0.10 界定获益人群.采用 Bootstrap 重抽样 1000 次内部验证,比较不同获益分层中溶栓与非溶栓患者的结局差异.结果:最终模型纳入年龄、美国国立卫生研究院卒中量表(NIHSS)评分、阿尔伯塔卒中项目早期 CT 评分(ASPECTS)、大血管闭塞部位及是否机械取栓5个变量.模型预测6个月获益的AUC为0.857(95%CI:0.792~0.911),Brier分数为0.196,校准斜率为 0.982(95%CI:0.812~1.151).全队列 ΔP 中位数为 0.09(0.04,0.16).高获益组(ΔP≥0.10)中,溶栓患者 6 个月 mRS 评分≤4 分比例显著高于未溶栓患者(72.22%vs.35.00%,P<0.05);中等获益组(0≤ΔP<0.10)与无获益组(ΔP<0)中,两组差异均无统计学意义(P>0.05).结论:本研究构建的简化模型提示,年龄、NIHSS评分、ASPECTS、大血管闭塞部位及机械取栓信息联合,有助于识别急性前循环大血管闭塞性脑梗死患者更可能从静脉溶栓获益的人群.

Objective:To construct a simplified and interpretable machine learning model for the early identification of individuals who can benefit from intravenous thrombolysis within 6 months after acute anterior circulation large vessel occlusion ischemic stroke.Methods:134 patients with acute anterior circulation large vessel occlusion ischemic stroke who were treated at the Affiliated Hospital of Xuzhou Medical University and Suining County People's Hospital from June 2022 to June 2025 were retrospectively enrolled.The patients were divided into the intravenous thrombolysis group(n=62)and the non-thrombolysis group(n=72)according to treatment regimens.Clinical data of the patients were collected,and propensity score overlap weighting was used to control treatment selection bias.Based on the weighted samples,a penalized Logistic regression model incorporating treatment interaction terms was used for variable screening to establish a simplified individualized benefit prediction model.The absolute probability difference(ΔP)of achieving a favorable 6-month functional outcome[modified Rankin scale(mRS)score≤4]was calculated for each patient under thrombolysis and non-thrombolysis scenarios,and patients with ΔP≥0.10 were defined as the benefit population.Bootstrap resampling with 1000 iterations was performed for internal validation,and differences in outcomes between thrombolytic and non-thrombolytic patients in different benefit strata were compared.Results:The final model included five variables:age,National Institutes of Health stroke scale(NIHSS)score,Alberta stroke program early CT Score(ASPECTS),large vessel occlusion site,and mechanical thrombectomy.The area under the curve(AUC)of the model for predicting 6-month treatment benefit was 0.857(95%CI:0.792-0.911),the Brier score was 0.196,and the calibration slope was 0.982(95%CI:0.812-1.151).The median ΔP of the entire cohort was 0.09(interquartile range:0.04,0.16).In the high-benefit group(ΔP≥0.10),the proportion of patients with a 6-month mRS score≤4 was significantly higher in the thrombolysis group than that in the non-thrombolysis group(72.22%vs.35.00%,P<0.05).No statistically significant differences were observed between the moderate-benefit group(0≤ΔP<0.10)and the no-benefit group(ΔP<0),with all P>0.05.Conclusion:The simplified model constructed in this study indicates that the combination of age,NIHSS score,ASPECTS,large vessel occlusion site,and mechanical thrombectomy status may help identify patients with acute anterior circulation large vessel occlusion ischemic stroke who are most likely to benefit from intravenous thrombolysis.

王吉慧;陈习;赵耀;周昌虎

徐州医科大学附属医院急诊科 江苏 徐州 221006||睢宁县人民医院急诊科 江苏 徐州 221200睢宁县人民医院急诊科 江苏 徐州 221200睢宁县人民医院急诊科 江苏 徐州 221200徐州医科大学附属医院急诊科 江苏 徐州 221006

医药卫生

急性前循环大血管闭塞性脑梗死静脉溶栓机器学习个体化治疗预后

Acute Anterior Circulation Large Vessel Ischemic StrokeIntravenous ThrombolysisMachine LearningIndividualized TreatmentPrognosis

《机器人外科学杂志(中英文)》 2026 (9)

1663-1671,9

徐州市科技计划项目(KC23338) Xuzhou Science and Technology Program(KC23338)

10.12180/j.issn.2096-7721.2026.09.024

评论