基于非负矩阵分解的肌肉协同分析方法在脊髓损伤患者步行功能评估中的应用

关键词：步行功能评估  脊髓损伤  肌肉协同  约束非负矩阵分解

Keywords:assessment of walking function  spinal cord injury  muscle synergy  constraint nonnegative matrix factorization

摘要：

      摘要 目的：引入一种基于约束非负矩阵分解（CNMF）的肌肉协同分析方法，应用于脊髓损伤后患者的步行肌肉激活模式分析。 方法：选择6例脊髓损伤患者和8例年龄相仿的正常受试者，采集6个完整的15m独立行走的下肢肌肉表面肌电信号，比较脊髓损伤患者和正常受试者的肌肉协同数和协同稳定性，并与非负矩阵分解（NMF）进行比较。 结果：CNMF相对于NMF在非稀疏协同计算方面有更高的准确性（P<0.05），CNMF在患者肌肉协同计算方面具有更好的稳定性（CNMF和NMF受试者内部一致性分别为0.79±0.10和0.74±0.11），与健康受试者相比，脊髓损伤患者的步行运动肌肉协同具有较低的一致性（患者和正常受试者的受试者内部一致性分别为0.79±0.10和0.98±0.01），6位脊髓损伤患者中，4位的步行肌肉协同数量比正常受试者多。 结论：通过添加平滑约束，CNMF能够显著改善肌肉协同的计算效果；脊髓损伤患者的步行肌肉协同具有显著的特异性。本研究为肌肉协同用于脊髓损伤步行运动功能评估奠定了理论基础。

Abstract:

      Abstract Objective： A muscle synergy analysis method based on constrained non-negative matrix factorization （CNMF） and to apply it to the analysis of muscle activation patterns during walking in patients with spinal cord injury (SCI). Method：Six patients with SCI and eight age-matched healthy subjects were selected. Surface electromyography data of the lower limb muscles were collected during six complete 15-meter independent walking trials. The number of muscles between the patients with SCI and the healthy subjects were compared， and it was also compared with Non-negative Matrix Factorization （NMF）. Result：CNMF demonstrated higher accuracy than NMF in the computation of non-sparse synergies（P<0.05），and exhibited better stability in the computation of muscle synergies in patients（intra-subject consistency：0.79±0.10 for CNMF vs. 0.74±0.11 for NMF）. Compared with the healthy subjects，SCI patients showed lower consistency in the muscle synergies of walking（intra-subject consistency：0.79±0.10 for patients vs. 0.98±0.01 for healthy subjects）. Among the six SCI patients，four exhibited a higher number of muscle synergies than healthy subjects. Conclusion：By adding smoothness constraints， CNMF can significantly improve the computational effect of muscle synergy. Compared with normal subjects， the muscle synergies of patients has significant specificity and is a potential method for walking function evaluation.