Decomposition of Motor Unit Action Potential in EMG Signal
(October 2013 – Jan 2017)
Decomposition of Motor Unit Action Potential in EMG Signal: Statistical Shape Model Analysis of MUAPs
Introduction
Electromyogram (EMG) signal is a record of electrical activity of muscle contraction which is made up of many single contraction units. Features of this signal generally are impressed by phiysiological characteristics of the muscle, that is for instance, if the contraction force increases, the number of activated units and the rate of activation increases, too. Also, each unit represents electrical activity of a single fiber bundle in an specified muscle. Studying the behavior and characteristic of contraction in fibers gives a rich and comprehensive information related to physiological investigations and clinical examinations in motor control and healthiness of the muscle. Hence, shape characteristics can be a good help in diagnosis of some neuromascular disorders like myopathic, neurogenic disorders, and the level of abnormality in some motor neuron diseases. So, in recent decades there have been many advances in recognizing and decomposing the components of EMG signal. The enormous number of spike shapes, the fact that single spikes can merge and overlap if two or more Motor Units (MUs) fire simoltaneously or nearly the same time, in addition to the noisy nature of EMG signal, makes it a challengable work to decompose the signal into separate elements. In fact, the biggest problem is that skeletal muscles don’t contract individually. Almost always there are tiny groups of fiber bundles that contract simultaneously. Each group of fiber bundles are stimulated by a MU which are called -motor neurons. Acumulation of all the potential activities of MUs in time are called Motor Unit Potential (MUP). In clinical point of view, analysis of EMG signal
through decomposition and clustering MUPs in similar groups are used for biofeedback training and help in diagnosis and analysis of neuromuscular system disorders. Also, full decomposition can be useful to study discharge irregularities like those associated with doubly innervated muscle fibers[2]. As mentioned earlier, EMG signal is natually noisy. As shown in (1) a mathematical model for describing the EMG signal is made up of a noise term added to a summation of a large number of blended motor unit action potentials (MUAPs).
Unfortunately, distingushing the valid trains of template shapes of MUAP and firing patterns of MU from invalide ones is not an easy work, since they are similar. Also, another misleading factor is variability of MUAP shapes and MU firing patterns. Such inaccurate information should be excluded to have a proper and reliable analysis. To remove mentioned improper information and also background noise and artifacts, preprocessing is needed in the first step. Here, we follow 4 steps for automated EMG signal decomposition:
(1) preprocessing (2) MUAP segmentation (3) feature extraction (4) MUAP clustering and classification (5) Statistical Shape Mode of MUAPs (6) Segmentation and Clustering based on SSM
Applying One Dimensional Active Shape Model in decomposition of
MUAPs in EMG Signal
One of the interesting characteristics of the MUAPs is that they have specific patterns which are almost fixed with little variations for each fiber. Due to this feature, one of the methods that usually is used in image processing analysis can be applied to this field. Statistical Shape Models (SSM) use statistic parameters such as mean and covariance of an object, to draw a mean shape of it in which changing the parameters results in expansion and contraction in different directions. Using this method, a clustering of firing pattern in each fiber can be done. The main approach to this study is using this method to extract shape of the MUAPs and to model the shape variations so that we can inversely decide for a given MUAP, that which fiber-class it belongs to.
Collaborators:
Adjunct Professor of Electrical and Computer Engineering at The University of British Columbia
Farnaz Mohammadi
B.Sc. Student at University of Tehran
