|
| 机器学习预测脑卒中上肢运动功能结局：回顾性研究 |
| Machine learning predicts upper limb motor function outcomes in stroke: a retrospective study |
| 投稿时间：2025-03-25 修订日期：2025-03-25 |
| DOI： |
| 中文关键词: 脑卒中康复,上肢运动功能,机器学习 |
| 英文关键词: Stroke rehabilitation, upper limb motor function, machine learning |
| 基金项目:1. 中国康复研究中心青年“pBFS指导下的不同剂量iTBS 治疗脑卒中后上肢运动障碍的随机对照研究”(项目编号：2023ZX-Q3) |
|
| 摘要点击次数: 154 |
| 全文下载次数: 0 |
| 中文摘要: |
| 目的:确定4种机器学习(Machine Learning, ML)模型预测脑卒中上肢运动功能恢复结局的准确性和适用性，并探讨与结局恢复相关预测变量的重要性。
方法:纳入802例中国康复研究中心住院的脑卒中患者，以临床数据作为特征变量、上肢运动功能恢复情况作为结果变量构建预测模型。验证比较了K-近邻 (K-Nearest Neighbor , KNN)、支持向量机(Support Vector Machine , SVM)、随机森林(Random Forest, RF) 和极限梯度提升(eXtreme Gradient Boosting, XGBoost)四种模型的预测性能。
结果: XGBoost相较于其他模型表现出了最佳性能，其准确率最高，为82.59%，其次为RF、SVM、KNN，准确率分别为81.09%、72.64%、70.15%。XGBoost和RF预测脑卒中运动功能改善的重要特征为病程、年龄、入院时美国国立卫生研究院卒中量表(National Institute of Health Stroke Scale, NHISS)评分、Fugl-Meyer评估量表上肢评分(Fugl-Meyer Assessment Upper Extremity, FMAUE)及Barthel指数(Barthel Index, BI)评分。
结论: ML预测模型可以准确预测脑卒中后上肢运动功能结局，XGBoost模型表现出了最佳性能。XGBoost模型和RF模型共同确定，病程、年龄、入院时NHISS、FMAUE及BI评分是影响脑卒中上肢运动功能恢复的重要因素 |
| 英文摘要: |
| Objective: To determine the accuracy and applicability of four machine learning (ML) models in predicting upper - limb motor function recovery outcomes post - stroke, and to explore the importance of related predictive variables for recovery.
Methods: Enrolled 802 stroke patients from China Rehabilitation Research Center. Used clinical data as feature variables and upper - limb motor function recovery as the result variable to build prediction models. Compared and evaluated the predictive performance of K - Nearest Neighbor (KNN), Support Vector Machine (SVM), Random Forest (RF), and eXtreme Gradient Boosting (XGBoost).
Results: XGBoost outperformed others with the highest accuracy of 82.59%, followed by RF (81.09%), SVM (72.64%), and KNN (70.15%). For XGBoost and RF, key features predicting motor function improvement were disease course, age, admission NIHSS score, FMA - UE score, and BI score.
Conclusion: ML models can accurately predict post - stroke upper - limb motor function outcomes. XGBoost performed best. Both XGBoost and RF identified disease course, age, admission NIHSS, FMA - UE, and BI scores as crucial factors for upper - limb motor function recovery in stroke patients. |
|
View Fulltext
查看/发表评论 下载PDF阅读器 |
| 关闭 |
|
|
|