姿态与肌电融合的脑卒中患者上肢运动功能实时评估方法

关键词：脑卒中  上肢运动功能评估  Brunnstrom分期  姿态测量  表面肌电  模糊支持向量机

Keywords:stroke  upper-limb motor function assessment  Brunnstrom stage  posture measurement  surface electromyography  fuzzy support vector machine

摘要：

      摘要 目的：提出了一种基于运动姿态与肌电融合的脑卒中患者上肢运动功能实时评估方法。 方法：选取一定数量处于不同Brunnstrom分期的脑卒中患者。首先运用姿态与肌电检测传感器，分别采集康复动作评估过程中不同分期患者上肢运动姿态与表面肌电数据；其次分别提取患肢运动姿态与表面肌电数据特征，在融合姿态与肌电特征信息基础上获取患者上肢运动功能评估数据集；最后，选取具有较好实时分类性能的模糊支持向量机作为分类器，运用评估数据集测试上肢运动功能评估效果。 结果：基于模糊支持向量机构建的运动功能分类器经过离线训练后，姿态与肌电融合的分类器对脑卒中患者上肢运动功能实时评估准确率平均达到83%，且较单纯基于运动姿态(75%)或肌电(74%)，具有更好的评估效果。 结论：姿态与肌电融合的上肢运动功能实时评估方法，能较准确地对脑卒中患者上肢运动功能进行实时Brunstrom分期。

Abstract:

      Abstract Objective： To study an upper-limb motor function real-time assessment method for stroke patients using posture and electromyography. Method： A certain number of stroke patients with different Brunnstrom stages are recruited. Firstly, the impaired limb’s movement postures and electromyography data during executing the assessment movements are measured using posture and electromyography sensors, respectively. Secondly, the features of the impaired limb’s movement postures and electromyography are extracted, and assessment data sets are constructed after integrating the posture and electromyography features. Finally, the fuzzy support vector machine based motor function classifiers with superior real-time classification performance are developed, and the assessment efficacy is verified using the assessment data sets. Result： After offline training the fuzzy support vector machine based motor function classifiers, classifiers combined postures with electromyography have an average 83% accuracy to the stroke patient’s upper-limb motor function assessment, and have more accurate performance compared to single movement postures (73%) or electromyography (74%). Conclusion： The upper-limb motor function real-time assessment method using postures with electromyography, can accurately differentiate the Brunnstrom stages of the upper-limb motor function for stroke patients.