**1. Introduction**

Computational Intelligence in Electromyography Analysis – 260 A Perspective on Current Applications and Future Challenges

Kohonen, T. (2000). Self-Organizing Maps, *Springer* 

Automation, Vol.19, No.5, pp. 842-853

Vol.47, No.3, pp. 239-246

CVIM-93, pp. 233-240 (in Japanese)

Assisted Surgery, Vol.5, No.4, pp. 423-434

*Computer Aided Surgery*, pp. 43-44

Ikuta, K.; Hasegawa, M. & Goto, H. (2007). Total System of Hyper Finger for Remote Minimally Invasive Surgery (The 9th Report) Proposal and Experimental Verification of Safety Operation Strategies, *Proceedings of the 16th Annual Meeting of The Japan Society of* 

Ishii, C.; Mikami, H.; Nakakuki, T. & Hashimoto, H. (2011). Bilateral Control for Remote Controlled Robotic Forceps System with Time Varying Delay, *Proceedings of IEEE* 

Kita, K.; Kato, R. & Yokoi, H. (2010). EMG-to-Motion Classification for Prosthetic Applications ? A Self-Organizing Approach with Level of Proficiency -, Journal of the

Kizuka, T.; Masuda, T.; Kiryu, T & Sadoyama, T. (2006). Biomechanism Library Practical Usage of Surface Electromyogram, Tokyo Denki University Press (in Japanese)

Krupa, A.; Gangloff, J.; Doignon, C.; de Mathelin, M. F.; Morel, G.; Leroy, J.; Soler, L. & Marescaux, J. (2003). Autonomous 3-D Positioning of Surgical Instruments in Robotized Laparoscopic Surgery Using Visual Servoing, IEEE Transactions on Robotics and

Kumagai, T.; Yamashita, J.; Morikawa, O.; Yokoyama, K.; Fujimaki, S.; Konishi, T.; Ishimasa, H.; Murata, H. & Tomoda. K. (2008). Distance Education System for Teaching Manual Skills in Endoscopic Paranasal Sinus Surgery Using "HyperMirror" Telecommunication

Nakaya, Y.; Ishii, C.; Nakakuki, T. & Hikita, M. (2010). A Practical Approach for Recognition of Hand Gesture and Distinction of Its Singularity, *Proceedings of 2010 IEEE International* 

Rosen, J.; Brown, J. D.; Chang, L.; Sinanan, M. N. & Hannaford, B. (2006). Generalized Approach for Modeling Minimally Invasive Surgery as a Stochastic Process Using a Discrete Markov

Sadoyama, T. & Miyano, H. (1981). Frequency analysis of surface EMG to evaluation of muscle fatigue, European Journal of Applied Physiology and Occupational Physiology,

Tada, M.; Omura, R.; Naya, F.; Noma, H.; Toriyama, T. & Kogure, K. (2006). Analysis of Steering Control Behavior Using 3D-Accelerometers, IPSJ SIG Technical Report, SIG-

Tanoue, K. & Hashizume, M. (2007). Advanced medicine and innovative technology,

Tokuda, J.; Fischer, G. S.; Papademetris, X.; Yaniv, Z.; Ibanez, L.; Cheng, P.; Liu, H.; Blevins, J.; Arata, J.; Golby, A. J.; Kapur, T.; Pieper, S.; Burdette, E. C.; Fichtinger, G.; Tempany, C. M. & Hata, N. (2009). OpenIGTLink: an open network protocol for image-guided therapy environment, The International Journal of Medical Robotics and Computer

Yamashita, J. (2009). A distance- and Self-Education System to Learn Experts' Postures for Technical Skills Training in Endoscopic Sinus Surgery, Journal of Japan Society of

Model, IEEE Transactions on Biomedical Engineering, Vol.53, No.3, pp. 399-413

*International Conference on Human System Interaction*, pp. 330-335

Robotics Society of Japan, Vol.28, No.7, pp. 783-791 (in Japanese)

Interface, Proceedings of IEEE Virtual Reality 08, pp. 233-236

Fukuoka Acta Medica, Vol.98, No.4, pp. 100-105 (in Japanese)

Computer Aided Surgery, Vol.11, No.3, pp. 170-171 (in Japanese)

*Conference on Automation and Logistics*, pp. 474-479

Traditionally, in clinical electromyography (EMG), neurophysiologists assess the state of the muscle by studying basic units of an EMG signal, which are referred to as motor unit action potentials (MUAPs). Information regarding the morphology and rate of occurrence of MUAPs is often used for diagnosis of neuromuscular disorders. In addition, recent studies have shown that the analysis of the energy content of MUAPs is a possible way for discriminating among normal, neurogenic, and myopathic MUAPs [38], illustrating, thus, the clinical value of the interpretation of MUAP information.

A common way of obtaining such information is by observing MUAP activities on an oscilloscope and listening to their audio characteristics over the speakers. When doing this, the researcher is implicitly performing a time and frequency analysis of MUAPs. However, the results of this analysis are dependent on the experience of the investigator and on his ability to extract relevant information from the visual and auditory analysis. Furthermore, this procedure is time-consuming and prone to error.

The drawbacks related to the procedure described above have motivated the use of computer-based techniques for extraction of MUAPs from EMG signals [3, 19, 22, 27, 29, 32, 40]. Such methods, also known as EMG decomposition techniques, aim at classifying MUAPs generated by a common source into the same group. The results of this classification may provide information regarding the orchestration of the neuromuscular system, and therefore of the state of the muscle. A similar problem, often referred to as spike sorting, is found in the study of neuronal activities [25]. In this case, neuronal action potentials from the same source are classified into a common group.

Originally, the investigation of MUAP activities belonged to needle electromyographic (NEMG) studies, mainly because surface electrodes may easily produce an integration of many potentials, which precludes accurate study of their individual form. However, some recent studies have shown that the use of surface electrodes may be successfully applied for

detection of MUAPs from superficial muscles [8]. This advancement has received widespread support among researchers and clinicians because of the ease of use, reduced risk of infection, and the greater number of motor unit action potential trains obtained compared to needle sensor techniques [47].

the use of alternative filtering procedures, for instance, wavelets and a spatial filter, known as a Laplacian filter. The implementation of these filters as well as the introduction of an

EMG Decomposition and Artefact Removal 263

Another relevant pre-processing step is the one that segments the EMG signal into windows containing active and inactive segments. This is a signal detection stage which aims to identify the activities of single MUAPs or their combinations, which is known as MUAP overlaps. Furthermore, this step separates noise in inactive segments from useful information in active segments. The detection of active and inactive segments may be performed visually or manually, i.e. the researcher may classify regions of an electromyographic signal into one of those categories, however such a method is time-consuming and requires concentration,

Different approaches may be employed for automation of this pre-processing stage: (i) The use of the root-mean square of the EMG signal together with a pre-defined threshold; (ii) A threshold proportional to the maximum peak in the signal; (iii) A threshold which is manually adjusted; (iv) Wavelets. The main assumption of this method is that there might be similarity between the mother wavelet and action potentials, and when the correlation between an active

As one may note there exists a variety of strategies that could be considered for automation of the detection of active and inactive segments. Techniques that make an assumption about a pre-defined height and width of MUAPs may be more susceptible to failures, i.e. active regions that do not fit the predefined window will not be detected. Note also that the level of the background activity in the signal will also influence the determination of the beginning and end of active segments. The detection of active regions is further discussed in this chapter. After detection of active regions it is possible to group them into logical units (clusters), however, it is very common to obtain features from those regions prior to data clustering. For example, morphological features of MUAPs, i.e. duration, amplitude, area, number of phases (number of baseline crossings) and number of turns (number of positive and negative peaks) have been employed. Other successful approaches include the use of coefficients of the Fourier transform, the coefficients of the Wavelet transform, the use of time samples of the band-pass filtered signal and low-pass differentiated signal, and the use of autoregressive and cepstral coefficients. Some other approaches, known as feature extraction and selection procedures, try to obtain features that maximize cluster separability. Examples of the use of such techniques include the application of Principal Component Analysis (PCA) and

The grouping of MUAPs is commonly performed by means of at least three distinct strategies: (i) Template matching: raw MUAPs, referred to as MUAP templates, are first classified or identified, and then used for classification of new MUAPs. The initial MUAP templates may be manually selected from the EMG signal or be chosen automatically from a clustering procedure. During data classification, usually MUAP templates are modified by an update rule, which takes into account the variability of MUAP shapes in the data set; (ii) Clustering: a clustering technique is used for grouping patterns represented by features selected from active regions; (iii) Hybrid: in this approach, first a clustering technique is used for grouping

alternative method for EMG signal filtering is further discussed in the chapter.

which may introduce inconsistency in the signal analysis.

Independent Component Analysis (ICA).

window and the mother wavelet is high an active segment is detected.

Currently, computer-based EMG has become an indispensable tool for investigations seeking to explain the state of the muscle. Different methodologies, ranging from simple quantitative measures to automatic systems that enable the assessment of neuromuscular disorders, have been developed [30]. Such tools are important for standardization of results and also they may reveal important features in the signals, which might be barely perceived from a manual analysis [35].

A typical system for extraction of MUAPs from EMG signals may require several stages of signal processing, for instance, signal detection and filtering (i.e., artefact removal) [45], feature extraction or selection [36], data clustering or classification [23]. Specific research may be carried out in each of these steps.
