首页 >> 肌电 >>最新推荐 >> 高密度肌电图分解算法:上肢肌肉的实验评价
肌电

高密度肌电图分解算法:上肢肌肉的实验评价

ndependent component analysis based algorithms for high-density electromyogram decomposition: Experimental evaluation of upper extremity muscles

Chenyun Dai ⁠a , Xiaogang Hu ⁠a⁠, ⁠∗ a Joint Department of Biomedical Engineering, University of North Carolina at Chapel Hill, North Carolina State University, United States A

ABSTRACT

Motor unit firing activities can provide critical information regarding neural control of skeletal muscles. Extracting motor unit activities reliably from surface electromyogram (EMG) is still a challenge in signal processing. We quantified the performance of three different independent component analysis (ICA)-based decomposition algorithms (Infomax, FastICA and RobustICA) on high-density EMG signals, obtained from arm muscles (biceps brachii and extensor digitorum communis) at different contraction levels. The source separation outcomes were evaluated based on the degree of agreement in the discharge timings between different algorithms, and based on the number of common motor units identified concurrently by two algorithms. Two metrics, the separation index (silhouette distance or SIL) and the rate of agreement, were used to evaluate the decomposition accuracy. Our results revealed a high rate of agreement (80%–90%) between different algorithms, which was consistent across different contraction levels. The RobustICA tended to show a higher RoA with the other two algorithms (especially with Infomax), whereas FastICA and Infomax tended to yield a greater number of common MUs. Overall, through an experimental evaluation of the three algorithms, the outcomes provide information regarding the utility of these algorithms and the motor unit filter criteria involving EMG signals of upper extremity muscles.

INTRODUCTION

Electromyogram (EMG) signals represent a convoluted process of hundreds of motor unit (MU) discharge activities with the corresponding motor unit action potentials (MUAPs). Decomposition of the EMG signals involves the separation of the composite interference activities into the constituent MU discharge activities, which can provide clinical and scientific insights into the neuromuscular systems [1–4]. In addition, recent studies have shown that MU discharge timings could be used as a neural interface signal during human-machine interactions [5].Earlier EMG decomposition mainly based on template-matching of MUAPs via manual expert editing or automated algorithms using intramuscular recordings [6–8]. With the development of algorithms and electrode hardware, MU activities can be extracted through the decomposition of high-density (HD) surface EMG signals using different blind source separation algorithms [9–12]. Despite these initial successes, the performance evaluation of the decomposition is still a challenging task. Previously, the decomposition performance has been evaluated through several methods. First, synthesized EMG signals can be simulated, and the accuracy of the decomposition can be evaluated by directly comparing with the ground truth. Although the model simulation may not fully reproduce the detailed features as in real EMG signals, this technique can directly measure the performance of decomposition [9,10]. Second, two source validation has been used to evaluate the decomposition accuracy of a small sample of MU firing activities [13–15]. Specifically, surface and intramuscular recordings are performed concurrently on the same muscle, and the two types of recordings at different sources are decomposed separately, with the intramuscular decomposition as the reference outcome. The rate of agreement (RoA) of the common MU activities are used as a measure of decomposition accuracy. However, the number of common MUs obtained from the two recording sources are typically small, and, therefore, only a small fraction of the decomposed MUs from the surface recordings can be assessed. Lastly, different metrics extracted from individual MUs (e.g. spike pulse to noise ratio [16], dissimilarities or amplitude of the MUAP [17]) have been used to predict the decomposition accuracy through a correlation analysis. Recent studies have also used a cluster-ing index, the silhouette distance (SIL), as a measure of the degree of separation of the MUAP train from the background noise, including other potential source signals [18,19]. However, the association between the SIL and the decomposition performance has not been fully investigated.Accordingly, the purpose of our current study was to evaluate the RoA of three ICA based EMG decomposition algorithms (FastICA [20], Infomax [21], and RobustICA [22]), based on EMG signals obtained from two arm muscles (biceps brachii and extensor digitorum communis) at different muscle contraction levels. The RoA between algorithms was evaluated based on the notion that the accuracy should be high if the same MUs can be separated repetitively from different algorithms with a high agreement. The decomposition yield of the algorithm was also compared based on the experimental data. Our results showed that the RoA of the decomposition from different algorithms was largely above 80% with the SIL>0.85, although a number of MUs still showed high RoA (>80%) with the SIL ranging from 0.50 to 0.85. Overall, the combination of RobustICA and Infomax showed a consistently higher RoA than other algorithm combinations. In contrast, the number of common motor units between Infomax and FastICA was the highest among the different algorithm combinations. These findings provided experimental evidence on selecting decomposition algorithms and performance assessment metrics for specific applications that have different accuracy and yield requirements.

Discussion

The purpose of this study was to evaluate the performance of three previously developed ICA-based source separation algorithms (Infomax, FastICA and RobustICA) on MU decomposition of EMG signals obtained from two arm muscles (biceps brachii and EDC). Two evaluation metrics, SIL and RoA between algorithms, were used to assess the decomposition performance. Our results revealed a high RoA between different algorithms across different muscle contraction levels. The RobustICA tended to show a higher RoA with the other two algorithms (especially Infomax), whereas FastICA and Infomax tended to yield a greater number of common MUs by these two algorithms. The experimental outcomes were also largely consistent with earlier simulation results in Part 1. These findings can provide guidance on selecting particular decomposition algorithms and particular performance assessment metrics for different applications that have different requirements on the decomposition accuracy and yield.

Conclusion

In general, we performed a systematic evaluation on the performance of different ICA-based algorithms for MU decomposition of EMG signals obtained from arm muscles. Specifically, RobustICA showed higher RoA with other algorithms, whereas FastICA and Infomax can decompose a greater number of common MUs. The selection of specific algorithms or MU filtering metrics may depend on different applications with particular requirement. The outcomes can help us identify reliable MU activities at the population level that can be used to understand mechanistic and clinical aspects of the neural control of muscle activations.

REFERENCES

[1] X. Hu, A.K. Suresh, W.Z. Rymer, N.L. Suresh, Altered motor unit discharge patterns in paretic muscles of stroke survivors assessed using surface electromyography, J. Neural Eng. 13 (2016) https://doi.org/10.1088/1741-2560/13/4/046025, 046025. [2] C. Dai, N.L. Suresh, A.K. Suresh, W.Z. Rymer, X. Hu, Altered motor unit discharge coherence in paretic muscles of stroke survivors, Front. Neurol. 8 (2017). [3] C.J. De Luca, E.C. Hostage, Relationship between firing rate and recruitment threshold of motoneurons in voluntary isometric contractions, J. Neurophysiol. 104 (2010) 1034–1046, doi:jn.01018.2009 [pii]10.1152/jn.01018.2009. [4] J. Duchateau, R.M. Enoka, Human motor unit recordings: origins and insight into the integrated motor system, Brain Res. 1409 (2011) 42–61. [5] D. Farina, I. Vujaklija, M. Sartori, T. Kapelner, F. Negro, N. Jiang, K. Bergmeister, A. Andalib, J. Principe, O.C. Aszmann, Manmachine interface based on the discharge timings of spinal motor neurons after targeted muscle reinnervation, Nat. Biomed. Eng. 1 (2017) 25. [6] E.D. Adrian, D.W. Bronk, The discharge of impulses in motor nerve fibres, J. Physiol. 67 (1929) 9–151. [7] K.C. McGill, Z.C. Lateva, H.R. Marateb, EMGLAB: an interactive EMG decomposition program, J. Neurosci. Methods 149 (2005) 121–133. [8] R.S. LeFever, A.P. Xenakis, C.J. De Luca, A procedure for decomposing the myoelectric signal into its constituent action potentials-part II: execution and test for accuracy, IEEE Trans. Biomed. Eng. (1982) 158–164. [9] A. Holobar, D. Zazula, Multichannel blind source separation using convolution kernel compensation, IEEE Trans. Signal Process. 55 (2007) 4487–4496. [10] M. Chen, P. Zhou, A novel framework based on FastICA for high density surface EMG decomposition, IEEE Trans. Neural Syst. Rehabil. Eng. 24 (2016) 117–127. [11] Y. Peng, J. He, B. Yao, S. Li, P. Zhou, Y. Zhang, Motor unit number estimation based on high-density surface electromyography decomposition, Clin. Neurophysiol. 127 (2016) 3059–3065. [12] Y. Ning, X. Zhu, S. Zhu, Y. Zhang, Surface EMG decomposition based on K-means clustering and convolution kernel compensation, IEEE J. Biomed. Heal. Informatics (2015) https://doi.org/10.1109/JBHI.2014.2328497. [13] B. Mambrito, C.J. De Luca, A technique for the detection, decomposition and analysis of the EMG signal, Electroencephalogr. Clin. Neurophysiol. 58 (1984) 175–188. [14] X. Hu, W.Z. Rymer, N.L. Suresh, Accuracy assessment of a surface electromyogram decomposition system in human first dorsal interosseus muscle, J. Neural Eng. 11 (2014) 26007, https://doi.org/10.1088/1741-2560/11/2/026007. [15] H.R. Marateb, K.C. McGill, A. Holobar, Z.C. Lateva, M. Mansourian, R. Merletti, Accuracy assessment of CKC high-density surface EMG decomposition in biceps femoris muscle, J. Neural Eng. 8 (2011) 66002, https://doi.org/10.1088/1741- 2560/8/6/066002. [16] A. Holobar, M.A. Minetto, D. Farina, Accurate identification of motor unit discharge patterns from high-density surface EMG and validation with a novel signal-based performance metric, J. Neural Eng. (2014) https://doi.org/10.1088/ 1741-2560/11/1/016008. [17] J.R. Florestal, P.A. Mathieu, K.C. McGill, Automatic decomposition of multichannel intramuscular EMG signals, J. Electromyogr. Kinesiol. (2009) https://doi.org/10. 1016/j.jelekin.2007.04.001. [18] F. Negro, S. Muceli, A.M. Castronovo, A. Holobar, D. Farina, Multi-channel intramuscular and surface EMG decomposition by convolutive blind source separation, J. Neural Eng. 13 (2016) 26027. [19] T. Kapelner, F. Negro, O.C. Aszmann, D. Farina, Decoding motor unit activity from forearm muscles: perspectives for myoelectric control, IEEE Trans. Neural Syst. Rehabil. Eng. 26 (2018) 244–251. [20] A. Hyvärinen, E. Oja, Independent component analysis: algorithms and applications, Neural Network. 13 (2000) 411–430. [21] A.J. Bell, T.J. Sejnowski, An information-maximization approach to blind separation and blind deconvolution, Neural Comput. 7 (1995) 1129–1159. [22] V. Zarzoso, P. Comon, Robust independent component analysis by iterative maximization of the kurtosis contrast with algebraic optimal step size, IEEE Trans. Neural Netw. 21 (2010) 248–261. [23] A.J. Fuglevand, D.A. Winter, A.E. Patla, Models of recruitment and rate coding organization in motor-unit pools, J. Neurophysiol. 70 (1993) 2470–2488. [24] J.N. Leijnse, N.H. Campbell-Kyureghyan, D. Spektor, P.M. Quesada, Assessment of individual finger muscle activity in the extensor digitorum communis by surface EMG, J. Neurophysiol. 100 (2008) 3225–3235, https://doi.org/10.1152/jn.90570. 2008. [25] J.N. Leijnse, S. Carter, A. Gupta, S. McCabe, Anatomic basis for individuated surface EMG and homogeneous electrostimulation with neuroprostheses of the extensor digitorum communis, J. Neurophysiol. 100 (2008) 64–75, https://doi.org/10. 1152/jn.00706.2007. [26] M. Chen, A. Holobar, X. Zhang, P. Zhou, Progressive FastICA peel-off and convolution kernel compensation demonstrate high agreement for high density surface EMG decomposition, Neural Plast. 2016 (2016). [27] A. Holobar, D. Farina, M. Gazzoni, R. Merletti, D. Zazula, Estimating motor unit discharge patterns from high-density surface electromyogram, Clin. Neurophysiol. 120 (2009) 551–562, https://doi.org/10.1016/j.clinph.2008.10.160. [28] X. Hu, A.K. Suresh, W.Z. Rymer, N.L. Suresh, Altered motor unit discharge patterns in paretic muscles of stroke survivors assessed using surface electromyography, J. Neural Eng. 13 (2016) https://doi.org/10.1088/1741-2560/13/4/046025. [29] X. Dai, C. Cao, Y. Hu, Prediction of individual finger forces based on decoded motoneuron activities, Ann. Biomed. Eng. (2019). [30] X. Hu, W.Z. Rymer, N.L. Suresh, Assessment of validity of a high-yield surface electromyogram decomposition, J. NeuroEng. Rehabil. 10 (2013) 99, https://doi.org/ 10.1186/1743-0003-10-99.

声明:本站所提供内容均来源于网友提供或网络搜集,由本站编辑整理,仅供个人研究、交流学习使用,不涉及商业盈利目的。如涉及版权问题,请联系本站管理员予以更改或删除。

 

  微信公众号  Sports_training


技术支持: 建站ABC | 管理登录