基于加权案例推理的脑卒中后偏瘫患者的智能康复训练处方决策研究

Intelligent Rehabilitation Prescription Decision Making for Post-Stroke Hemiplegia Patients Using Weighted Case-Based Reasoning

摘要: 及时、精准的康复训练干预可以有效促进脑卒中后的偏瘫(Post-Stroke Hemiplegia, PSH)功能恢复，提升生活独立性。现有康复服务模式具有流程繁琐、专家主导、服务资源分配不均等多重问题。为了提高康复训练决策效率，并提供全面可解释性，本文针对PSH患者的肢体功能障碍，提出了一种基于加权案例推理(Weighted Case-Based Reasoning, WCBR)的康复训练处方决策算法。基于139例临床的脑卒中患者病例构建案例库，WCBR以K近邻(K-Nearest Neighbor, KNN)为基准策略根据患者量表信息提供初始康复训练方案，供康复治疗师参考修正并给出最终康复治疗处方。通过实验验证，该算法的精度指标为87.5%，召回率指标为90.1%。本研究证明，该算法可以基于PSH患者肢体功能评估实现精准的康复训练决策，有望为康复训练提供临床决策支持，改善康复服务流程。

Abstract: Timely and accurate rehabilitation training intervention can effectively promote functional recovery from post-stroke hemiplegia (PSH) and improve independence in daily living. The existing rehabilitation service models have multiple issues, such as cumbersome processes, expert dominance, and uneven distribution of service resources. To improve the efficiency of rehabilitation training decision-making and provide comprehensive explaining ability, this paper proposes a rehabilitation training prescription decision algorithm based on Weighted Case-Based Reasoning (WCBR) targeted at the limb functional impairment of PSH patients. A case library was constructed based on 139 clinical stroke patient cases, and WCBR uses the K-Nearest Neighbor (KNN) as a benchmark strategy to provide an initial rehabilitation training plan based on the patient’s scale information, for rehabilitation therapists to reference, amend, and issue the final rehabilitation treatment prescription. Experimental validation showed that the algorithm has an accuracy metric of 87.5% and a recall metric of 90.1%. This study proves that the algorithm can achieve accurate rehabilitation training decisions based on the limb function assessment of PSH patients, is expected to provide clinical decision support for rehabilitation training, and improve the rehabilitation service process.