Decomposition of Multi-Channel Intramuscular EMG Signals by Cyclostationary-Based Blind Source Separation

Abstract

We propose a novel decomposition method for electromyographic (EMG) signals based on blind source separation. Using the cyclostationary properties of motor unit action potential trains (MUAPt), it is shown that MUAPt can be decomposed by joint diagonalization of the cyclic spatial correlation matrix of the observations. After modeling of the source signals, we provide the proof of orthogonality of the sources and of their delayed versions in a cyclostationary context. We tested the proposed method on simulated signals and showed that it can decompose up to 6 sources with a probability of correct detection and classification >95%, using only 8 recording sites. Moreover, we tested the method on experimental multi-channel signals recorded with thinfilm intramuscular electrodes, with a total of 32 recording sites. The rate of agreement of the decomposed MUAPt with those obtained by an expert using a validated tool for decomposition was >93%.Index Terms -EMG, Decomposition, Motor Unit, Firing Rate, Blind Source Separation, Cyclostationary Process, CycloSOBI, convolutive mixture, pulse train separation I IntroductionThe muscle contraction is generated by functional units --the motor units (MU, [1]) --each composed of a motoneuron (MN) innervating a group of muscle fibers. MNs generate electrical impulses, the action potentials (AP), which induce the activation of muscle fibers. The contraction force is controlled by the number of recruited MUs (recruitment) and by the density of APs within the train of each MU (rate coding) [2].The density of APs can be quantified by the inter-spike interval (ISI) or by its inverse value which is referred to as the firing rate (FR). These quantities can be defined in an instantaneous way, or with average values, or through their distributions. The EMG signal comprises the sum of the MUAP trains (MUAPt) of the active MUs. The identification of the MUAPt from the EMG provides the means to identify the neural drive to muscles. This procedure is often termed decomposition of the EMG and has provided fundamental insights into motor unit physiology and neural control of movement [3][4]. In addition to basic neurophysiologic investigations, decomposition of the EMG has also clinical relevance [5]. Acknowledgments The present work is part of the French national project ECOTECH (www.echotechsan.org) which is supported by the French National Agency for research, contract no ANR-12-TECS-0020. The work was also partly funded by the European Research Council (ERC) via the ERC Advanced Grant DEMOVE (No. 267888; to DF). The authors are grateful to Silvia Muceli for providing the experimental data.In this paper, we address the problem of EMG decomposition. Methods for intramuscular EMG decomposition date back to the 70's. In the earlier approaches, the user usually had to manually select and classify the detected MUAPs in order to identify the [34]. The BSS approach has been recently used also for decomposing multi-channel intramuscular EMG signals [35].Here, we propose a new cyclostationary based B...

On the Prediction of Motor Unit Filter Changes in Blind Source Separation of High-Density Surface Electromyograms During Dynamic Muscle Contractions

“…The number of studies clarifying the potential benefits and applications behind such identification is increasing rapidly with a particular focus on population coding [1] [2] [3]. Namely, the rapid development of systems for acquisition and processing of multichannel intramuscular [4] [5] or surface electromyographic (EMG) recordings [6] enabled the identification of the codes of several tens of simultaneously active motor nerves [7] [8] [9]. In this measuring setup, the skeletal muscles act as natural amplifiers of neural codes.…”

Motor Unit Identification From High-Density Surface Electromyograms in Repeated Dynamic Muscle Contractions

“…Recently, different methodologies for acquisition, assessment and interpretation of large number of neural codes via MU firing pattern identification have been introduced [5], [6], [26], [27], complementing the indwelling electromyography (EMG) with analysis of population coding in relatively large pools of MUs. Both surface and indwelling multichannel recording system have been introduced [1], [22], [24], [27], [30]. They both acquire activities of several tens of MUs, superimposed into interferential EMG patterns.…”

Recursive Decomposition of Electromyographic Signals With a Varying Number of Active Sources: Bayesian Modeling and Filtering

“…The procedures for iEMG signal decomposition have been progressively improved from methods strongly based on the manual intervention of an operator [12], [13], [14] to fully automatic methods [9], [15], [16], [17], [18], [19], [20], [21], [22]. Nevertheless, some limitations remain.…”

“…Other algorithms (e.g., [15], [16], [17]) achieve complete decomposition in an off-line manner by first identifying and clustering the non-overlapped MUAPs and then iteratively searching for their occurrences in the superpositions. More recently, blind source separation approaches have been proposed to decompose multi-channel EMG signals [9], [20], [21] but their performance strongly depends on the number of available channels [22].…”