2 EMG ELECTRODES INTRODUCTION 3 MATERIALS AND METHODOLOGY. All Rights Reserved 2015 IJARECE. 2.1 Needle Electrodes

Transcription

1 Segmentation of the EMG Signal and Comparison of the Normal and Diseased EMG signals Ashmeet Kaur (Student) Baba Banda Singh Bahadur Engineering College Navneet Kaur Panag (Assistant Professor) Baba Banda Singh Bahadur Engineering College Abstract In this paper, the electromyographic (EMG) signals of a patient and the normal person are considered. In order to decompose both the signals, different segmentation techniques are used. The two segmentation techniques used in this work are: i) Segmentation using daubechies wavelet transform (DWT), ii) by identifying the peaks of the MUAPs. The MUAPs (motor unit action potentials) are the important source of the EMG signal to provide information about any kind of disorder in the muscles or any disturbance in the functioning of the muscle cells. Further for the dissociation of the myopathy signal from the normal signal the time domain features of the EMG signal are studied. The two time domain features discussed in this work are: i) auto-correlation (ACR), ii) Zero-crossing Rate (ZCR). Thus using these features the EMG signals of the patient and the normal person are compared. It is seen that myopathy signals have lower autocorrelation peaks than the normal signals while the myopathy signals have the higher zero-crossing rates than the normal ones. Index Terms Segmentation, MUAPs, ACR, ZCR1 INTRODUCTION Electromyography (EMG) is a technique of recording the activity of muscles. It records the electrical current generated by the muscles. EMG of the muscle is measured by using an instrument called an electromyograph. It produces electrical record of muscles which is known as electromyogram. The EMG signal is the biomedical signal that consists of a train of motor unit action potentials called the MUAPs. The electrical currents generated in the muscles are measured during its contraction. The contraction and relaxation of the muscles is controlled by the nervous system. Thus EMG signal is the complicated signal and it depends upon the properties of the muscles. Muscles get stimulated by the signals from nerve cells called motor neurons. The stimulation of muscles thus causes electrical activity, which in turn results into contraction of muscles. This electrical activity of muscles is measured with the help of needle electrode which is inserted into the muscle and the same electrode is connected to a recorder which records the activity of that particular muscle. The electrode and the recorder together are known as the electromyography machine. EMG signal provide an important source of information for the diagnosis of neuromuscular disorders. It helps to identify the abnormalities of nerves or spinal nerve roots that may be associated with pain or numbness. There are many symptoms for which EMG may be useful which include numbness, stiffness, cramps, deformity, etc. EMG results provide the information that whether the symptoms are due to muscle disease or a neurological disorder [1]. 2 EMG EECTRODES The EMG signal is measured by using the electrodes either on the surface of the muscle or by inserting the electrode in the muscle. Two electrode types are used to record the electromyography (EMG) signal. Needle or wire electrodes are inserted into the muscle for the extraction of the signal. Surface electrodes are used on the surface or skin of the muscle under observation. 2.1 Needle Electrodes These types are very common in EMG recordings and are widely used for neuromuscular evaluations. The tip of the needle is bare. It is used as detection surface. The electrode is connected with the wire which is an insulated wire. The signal obtained with the needle electrode is improved as compared to the signal obtained using other available types. The main advantage of needle electrode is that it is able to detect individual MUAPs during relatively low contraction. The detected MUAPs will be of the required muscle. The other advantage of needle electrode is that it is conveniently repositioned within the muscle after insertion. 2.2 Surface Electrodes Surface electrodes provide non-invasive technique for the detection and evaluation of the EMG signal. These technique used is that these electrodes form a chemical equilibrium between the muscle surface that is needed to be detected and the skin of the body where electrode is placed, through the electrolytic conduction, so that current can flow into the electrodes. The advantage of surface electrode is that they are easy to implement and are simple to use. Unlike needle electrodes, surface electrodes do not require any strict medical supervision. Surface EMG electrodes have found their applications in neuromuscular detections, sports medical evaluation, and for the children where needle insertion is objected. Surface electrodes have some limitations also. As they are applied on the skin, interference from other muscle is a major problem. The position of surface electrodes must be kept stable with the skin otherwise the signal may be distorted. The electrode used in this work is the needle electrode as it does not involve the interference of electrical activity from the nearby muscles [2]. 3 MATERIAS AND METHODOOGY 3.1 Data Acquisition and Pre-Processing The data contains normal person s EMG signal and the EMG of a person suffering from muscle disease called the myopathy disease obtained from the Department of Computer Science, University of Cyprus, Cyprus. All the EMG signals were acquired from the biceps bronchi muscle at the maximum 2767

2 voluntary contraction. The signals were acquired using the standard concentric needle electrode. The recorded EMG signal of a normal person is given below. of wavelet decomposition, a set of approximation signals are obtained. Figure 3.2: Figure 3.1: Raw EMG signal of a Normal Subject Decomposition of DWT Coefficients 3.2 Segmentation The segmentation of the signal is referred to decomposing a given signal into a sum of simpler signals. The EMG signal is the superposition of the electrical activity of the muscles. The interference of the motor unit action potentials forms the EMG signal. Thus in order to understand the mechanism and functioning of the muscles, segmentation of the EMG signal is required. There are two techniques that are discussed and used for the decomposition of EMG signal Segmentation by using DWT There are many Daubechies transforms, they all are very similar. The technique used for the segmentation of the EMG signal in this work is the db4 wavelet transform. The db4 wavelet is the easiest wavelet transform. In this transform, considering a signal f, having an even number of values N, then the first level of db4 waveform is called the mapping and it is given as f (a 1 d 1 ). The mapping goes from the signal f to its first trend sub signal a 1 and to its first fluctuation sub signal d 1 where each value am of a 1 = (a1,a2,,.,an\2) is equal to a scalar product which is given as am=f.v 1 m (1.1) Here V 1 m is the scaling signal. Similarly each value dm of d 1 = (d1,d2,.,dn\2) is equal to the scalar product dm= f.w 1 m (1.2) Here W 1 m is the wavelet. Daubechies wavelet transform is the transformation of the original signal into a wavelet function. It captures the time and frequency domain of the signal. The time and frequency function are obtained by repeatedly filtering the signal by the pair of filters. In this process, the DWT decomposes the raw EMG signal into an approximation signal and a detail signal. The approximation signal is further decomposed into new approximation signal and the detail signal. Thus, at each level Figure 3.3: Segmentation of EMG Signal using DWT for Normal subject The Figure shows the decomposition of the EMG signal with the Daubechies wavelet transform. In this case a portion of MUAP is shown which is extracted using the db4 wavelet transforms [3] Segmentation by identifying the peaks of the MUAPs In order to detect the MUAPs comprising the EMG, the signal is segmented to generate the possible MUAP waveforms. Regions of low activity are removed using a threshold T which depends upon the maximum value maxi{xi} and the mean absolute value (1\) i=1 xi where xi are the discrete values of the EMG signal and is the number of samples. The threshold T is calculated as below: 2768

3 ISSN: X If maxi{xi}> 30 xi i=1, then T = 5 xi i=1 (1.3) else T = maxi{xi}/5 This threshold is used to identify the peaks of MUAPs of the EMG signal. Peaks that are above the mentioned threshold are considered as individual MUAPs. Also a window of 60 samples is placed over the identified peak. If there is any greater peak found then it is also considered in the window otherwise the 60 points are considered as MUAP waveform. The main advantage of using his technique is that it is easy to use. Also, it is a good technique for the identification of the highest peaks of MUAPs. This is the technique that has the capability to adapt different EMG signals [4] This Figure shows the recording of the electrical activity of the muscles of a person who is suffering from the myopathy disease. This is the raw EMG signal taken by measuring the response of a patient s muscles [5]. 3.4 Segmentation of Myopathy Signal The EMG signal of a myopathy patient is also segmented in order to know the mechanism of the muscles of the patient clearly. This segmentation of the myopathy signal will help to diagnose the part of the muscle where the disorder is detected. Basically segmentation is done so that the signal could be decomposed into simpler form. Thus the signal could be easily read when used clinically. Similar to the EMG signal of a normal person, the EMG signal of the patient will be decomposed in the same manner. Two techniques considered for the segmentation of the signal are: i. Daubechies Wavelet Transform ii. Segmentation by identifying the peaks of MUAPs The segmentation of the signal done by Daubechies wavelet transform uses db4 wavelet transform for the decomposition of the signal. Figure 3.4: Segmentation of EMG Signal by Identifying the Peaks of MUAPs for Normal Subject 3.3 Myopathy Disease in EMG signal Myopathy disease is the most common and frequently inherited musculoskeletal disease. The symptoms of the disease can be determined on the type of myopathy. Some generalised symptoms are muscle weakness, cramps, stiffness. In most of the myopathy disease, the weakness occurs primarily in the muscle of shoulders, upper arms, thighs. Other symptoms of myopathy disease are muscle pain, tightness, muscle aching. Figure 3.6: Segmentation of EMG signal using DWT for Myopathy subject Similarly, the peaks of the MUAPs are identified. The peaks are identified so that a small and simpler part of the signal could be obtained and studied properly. The peaks obtained can be studied to know what kind neuromuscular disorder isthere in the muscles [3]. Figure3.5: Raw EMG Signal of a Myopathy Subject 2769

4 Zero-crossing rate is defined as the time domain feature that indicates the number of times a signal crosses the zero-activity line or some other reference line. Mathematically it can be given as: Where sgn[x] = ZCR= 1 2N { 1, x 0 1, x 0 N 1 k=1 sgn x k sgn[x k 1 ] } (1.5) The ZCR is considered as distinguishable feature to detect the diseases in an individual s EMG signal [4]. 4 RESUTS Figure 3.7: Segmentation of EMG signal by Identifying the Peaks of MUAPs for Myopathy Subject 3.5 Dissociation of Myopathy Disease The myopathy disease is the neuromuscular disorder in the muscles of an individual. As the nervous system controls the muscle activity of the human body, thus for identifying the neuromuscular disorder, EMG signal can be observed and analysed in order to recognize the features of the myopathy disease in an individual. According tozeccaat al. [6], there are three types of features that can be extracted in an EMG signal. Time Domain Frequency Domain Time-Frequency Domain In this work, the myopathy disease can be distinguished from a person s EMG by studying and using the time domain features of the EMG signal Time Domain The two time domain parameters used for the EMG feature extraction are: The EMG of the muscles is taken from the biceps bronchi muscle. All the EMG signals from the muscles are taken using the needle electrodes. The recorded EMG signal is shown below. i. Auto-Correlation (ACR) ii. Zero-Crossing Rate (ZCR) Auto-Correlation (ACR) Considering two signals, the cross-correlation between the two signals is defined as the extent of similarity between these two signals. In other words, it can be said that cross-correlation between two signals is the measure of the dependency of these signals on each other. When the signals involved in the crosscorrelation becomes exactly the same then it is known as autocorrelation. Auto-correlation analyses how well the signal is correlated with itself. The auto-correlation indicates the degree of similarity at different portions of a time series data. In this work, the characteristics of the autocorrelation function of the EMG data have been investigated. Considering N-length sequence of EMG data x(n), its autocorrelation function rx(m) can be calculated as: Figure 4.1: Raw EMG Signal of Muscle The Figure shows the EMG signal of the muscle which is the recording of the electrical activity. 4.2 Results of Segmentation of EMG Signal of a Patient and a Normal Person rx(m) = 1 N N 1 m n=0 x n x(n + m ) (1.4) wherem denotes the correlation lag Zero-Crossing Rate (ZCR) Figure 4.2: EMG Signal segmented using Daubechies Wavelet Transform 2770

5 The Figure shows the classification of the EMG signal. The result shows that Daubechies wavelet technique is fast and simple. It can also be used in the study of other motor mechanisms for extracting the single motor units. 4.3 Comparison of EMG Signal of a Patient and a Normal Person Figure 4.3: Peaks of MUAPs of a Normal Person The Figure shows the segmented EMG of the normal person. The peaks of the MUAPs above the assigned threshold value are considered and the peaks below the threshold are left. The MUAPs taken above of the value assigned are enough for studying a person s EMG and knowing the musculoskeletal disorder. Thus this technique is also easier to apply and a good starting point for obtaining the peaks of the MUAPs. This is known as thresholding of EMG signal. Figure 4.5: Comparison of the EMG signal of Normal and Myopathy Patients In this Figure, the first block shows the average of peak value of ACR and the second block indicates the average of ZCR value. The systematic results indicate that the myopathic signals have lower autocorrelation peak than the normal ones. On the other hand, the signals of the muscles having disease have higher zero crossing rates than the normal person s signal. 5 CONCUSIONS Figure 4.4: Peaks of MUAPs of a Patient having Myopathy disease This segmentation technique of recognizing the peaks of the MUAPs showed better results for segmentation. It is an accurate and reliable approach for decomposing the signal into the individual MUAPs. In conclusion, the segmentation done by extracting the peaks of MUAPs using a certain threshold value and the daubechies wavelet transform are seen to be showing the exact results with proper extraction of the peaks. These techniques are considered as the simple, accurate, fast and reliable methods for the decomposition of the signal. It was found that both the methods are good for the processing of the non-stationary signals such as bio-medical signals where both time and frequency information is required. The advantages and disadvantages of both the techniques are discussed. The other advantage of segmentation technique by identifying the peaks is that it has theability to adapt different EMG signals. Also the important matter of concern in this case was choosing a proper window length in order to cover the main duration of MUAP spikes whenever the disease cases are to be considered. If the window length considered will be shorter then it will not be possible to contain the main MUAP spikes. This happens especially in case of motor neuron diseases because in those cases the MUAPs have longer duration. Also, for the dissociation of the myopathy disease and its diagnosis, the time domain features of the EMG signal are studied and observed carefully. In this work the time domain properties used are the auto correlation and the zero crossing rates which are effective for understanding the features of the EMG signals. These time domain features are useful for distinguishing the myopathy signal from the normal EMG signal. Both these time domain features significantly proved the approach to detect the myopathy disease. 2771

6 ACKNOWEDGMENT ISSN: X I thank Dr.Gursewak Singh Brar (H.O.D of Electrical Department) and Prof.Navneet Kaur Panag (Assistant Professor of Electrical Department). REFERENCES [1] Bida O., 2005, Influence of electromyogram (EMG) amplitude processing in EMG-torque estimation. PhD thesis,worcester Polytechnic Institute. [2] Reaz M.B.I., Hussain M.S., Mohd-Yasin F., Techniques of EMG signal analysis: detection, processing, classification and applications, Biological procedures online, Springer, Vol. 8(1), pp [3] Kaur G., Arora A.S., and Jain VK., 2009, Comparison of the techniques used for segmentation of EMG signals. In Proceedings of the 11th WSEAS international conference on Mathematical and computational methods in science and engineering. World Scientific and Engineering Academy and Society (WSEAS), pp [4] Christodoulou C.I. and Pattichis C.S., 1999, Unsupervised pattern recognition for the classification of EMG signals. Biomedical Engineering, IEEE Transactions on, 46(2): [5] Sayeed UdDoulahA.B.M. and Iqbal Md. A., An Approach to Identify Myopathy Disease Using Different Signal Processing Features with Comparison, IEEE Transactions, pp [6] Zecca M., Micera S., Carrozza MC., and Dario P., 2002, Control of multifunctional prosthetic hands by processing the electromyographic signal. Critical Reviews in Biomedical Engineering, Citeseer, 30(4-6):459. [7] Gut R. andmoschytz G.S, 2000, High Precision EMG SignalDecomposition Using CommunicationTechniques, IEEE Transactions onsignal Processing, Vol. 48, No.9,pp [8]Chauvet E.,Fokapu O.,Hogrel J.Y., Gamet D. and Duchene J., 2003, Automatic identification of motor unit action potential trains from electromyographic signals using fuzzy techniques, Medicaland Biology Engineering andcomputing Vol. 41, No.6, pp [9] Chauvet E., Fokapu O., Hongrej J.Y., Garnet D., Duchene J., 2001 A method of EMG decomposition based on fuzzy logic, Proceedings of the 23rd AnnualEMBS International Conference, Vol. 2, pp [10]Katsis C.D., Fotiadis D.I., ikas A. andsarmas I., 2003, Automatic discovery of the number of MUAP clusters and superimposed MUAP decomposition in electromyograms, Proceedings of the4th Annual IEEE Conf. on InformationTechnology Applications inbiomedicine, pp [11] Katsis C.D.,Goletsis Y., ikas A., Fotiadis D.I,Sarmas I., 2006, A novel method for automated EMG decomposition and MUAP classification, ArtificialIntelligence in Medicine, Vol. 37, No.1, pp [12] Katsis C.D, Exarchos T.P.,Popaloukas C.,Goletsis Y., Fotiadis D.I.,Sarmas., 2007, A two stage method for MUAP classification based on EMG decomposition, Computers in Biology and Medicine, Vol. 37, No. 9, pp [13] S ornmo. and aguna P., 2005, Bioelectrical signal processing in cardiac and neurological applications. Elsevier Academic Press. 2772