可解释的机器学习模型用于预测静脉输血栓溶解后急性缺血性中风的结果
Fanhai Bu1,2, Runlu Cai3, Wei Zhang2
1Department of Neurology, The First Clinical College, Xuzhou Medical University, Xuzhou, China.
Frontiers in neurology
|October 13, 2025
概括
机器学习准确地预测了急性缺血性中风 (AIS) 治疗后的糟糕结果. 中性粒细胞与淋巴细胞的比率 (NLR) 是一个关键的预测指标,可以早期识别高风险患者,以提供个性化护理.
科学领域:
- 神经学 神经学
- 医疗信息学 医疗信息学
- 生物统计学 生物统计学
背景情况:
- 急性缺血性中风 (AIS) 患者在静脉输血栓溶解 (IVT) 后往往具有较差的功能结果.
- 机器学习 (ML) 提供了新的方法来支持临床决策,预测患者的结果.
- 开发准确的预后模型对于优化IVT后护理至关重要.
研究的目的:
- 开发和验证一种ML模型,用于预测IVT治疗AIS患者的3个月不良功能结果.
- 使用ML技术识别功能结果的关键预测因素.
- 创建一个工具,用于早期识别高风险患者.
主要方法:
- 一项回顾性研究使用导出 (n=938) 和外部验证 (n=324) 的IVT治疗的AIS患者队列.
- 拉索回归从临床,神经成像和实验室数据中选择了预测因素.
- 八个ML算法被训练并使用交叉验证,AUC,准确性,精度,回忆和F1分数进行评估. 用SHAP分析来解释模型.
主要成果:
- 最终的后勤回归 (LR) 模型包括五个预测因素:NLR,NIHSS,ASPECTS,心房动和血糖.
- 在外部验证队列中,LR模型表现出强大的性能,AUC为0.797.
- 中性粒细胞与淋巴细胞的比率 (NLR) 被确定为不利结果的最强有力的预测因素.
结论:
- 一个节制的5变量LR模型有效地预测IVT后3个月的功能结果.
- 由NLR驱动的炎症在AIS预后中起着至关重要的作用.
- 这种预测工具有助于早期识别高风险患者,以进行个性化干预.
相关概念视频
Ischemic Stroke l: Introduction
53
Ischemic stroke is an acute cerebrovascular condition in which blood flow to a brain region is suddenly interrupted, leading to tissue infarction. Neurons depend on continuous oxygen and glucose supply, so even brief reductions in perfusion cause energy failure, ionic imbalance, and irreversible injury. Ischemic strokes are classified into thrombotic and embolic types based on their underlying mechanisms.Thrombotic MechanismsThrombotic stroke develops when a clot forms within a cerebral artery.
53
Ischemic Stroke ll: Pathophysiology
68
An ischemic stroke occurs when a cerebral blood vessel becomes obstructed, most often by a thrombus or embolus, interrupting the delivery of oxygen and glucose to brain tissue. Because neurons rely on continuous aerobic metabolism, energy failure begins within minutes of reduced perfusion. The region receiving the least blood flow becomes the infarct core, an area of irreversible cellular death. Surrounding this core lies the penumbra, a zone of hypoperfused but still viable tissue that is...
68


