If a pair of electrodes is applied on the skin above a muscle and the muscle is voluntarily activated, an electrical signal is detected between the electrodes (Figure 1). This signal has the following features, which have been extensively investigated and are well known (1).
(a) Its instantaneous value is apparently random with a gaussian distribution in the range of 10–500 μVrms. Its amplitude is estimated using the root mean square value (RMS), which coincides with the standard deviation of the distribution. Alternatively, the average rectified value (ARV) is used.
(b) Its harmonics (which form the “spectrum” and are obtained by means of Fourier analysis) are in the frequency range of 10–400 Hz. The mean frequency (centroid or center of gravity, indicated as MNF) of the spectrum is in the range of 70–130 Hz, and the median frequency (frequency value that splits the spectrum into two parts containing equal power, indicated as MDF) is in the range of 50–110 Hz .
(c) If two or more electrode pairs are applied in the direction of the fibers, a delay can be observed between similar signals detected from these pairs and the propagation velocity of the signal can be estimated. Such a value is an estimate of the muscle fiber conduction velocity (CV) and is reliable only if the two signals are sufficiently similar (correlation coefficient > 0.7 or 0.8).
These features are affected by the following two groups of factors.
Geometrical and Anatomical Factors
These factors include (a) electrode size, shape, and interelectrode distance; (b) electrode location with respect to the innervation zones (IZs) and the muscle tendon junctions; (c) thickness of the skin and subcutaneous layers; (d) misalignment between muscle fibers and electrodes in the horizontal plane (angle between the line connecting the two [or more] electrodes and the line in the fiber direction, in the plane of the skin); and (e) misalignment between muscle fibers and electrodes in the vertical plane (angle between the line connecting the two [or more] electrodes and the line in the fiber direction, in the plane perpendicular to the skin and longitudinal to the fibers).
These factors include (a) muscle fiber CV as a global average; (b) statistical distribution of muscle fiber CV (dispersion or scatter of the CV values around the average); (c) number of motor units (MUs), territory, number of fibers, fiber size, and histological type of each MU; (d) blood flow and temperature; (e) rate of metabolite production, intramuscular pH, ion concentrations, and shifts across the muscle cell membrane; (f) type and level of contraction (voluntary, concentric or eccentric, electrically stimulated); (g) mean and standard deviation of the inter pulse intervals of the MUs; and (h) degree of MU synchronization (if synchronization is indeed present).
It is evident that the signal contains information about many physical and physiological factors or variables whose contributions to the signal are not easy to separate. Two examples follow:
A change in signal amplitude from one test to another may be due to a change of electrode position, MUs activated, thickness of subcutaneous tissue, conduction velocity, electrode alignment with the direction of the fibers, or level of muscle activation.
The decrement of MNF or MDF during a sustained contraction may be due to a decrement of mean CV, an increase in CV dispersion (scatter) among different MUs, the dropout of superficial MUs (or those with higher CV) or the recruitment of deeper ones (or those with lower CV), or the widening of depolarization zones, a different degree of MU synchronization.
The challenge involves associating the measurable variables defined in the introduction (a, b, and c) to the mostly unknown physical and physiological factors listed, considering that the measured variables are only estimates, which are subject to the statistical errors associated to the detection and calculation methods (7). When a single electromyographic (EMG) channel is available, it is not possible to separate the contributions of these factors. There are simply too many unknowns and too few conditions to disentangle the information content. A linear array of equally spaced electrodes makes available many EMG channels and provides additional information. It allows the identification of single MU action potentials (MUAPs), the location of innervation and tendon zones, and the estimation of CV of the individual MUAPs and of their firing patterns. In the last few years, the array technique opened up many new possibilities of extracting information from the EMG signal. Nevertheless, most possibilities are still unexplored, and considerable research needs to be done. A linear array is also required to learn where a single electrode pair should be located on the muscle and to determine the effects of placing it in different positions. This issue is discussed, among others, in this review.
In addition to the array, a second important tool is an accurate model of EMG generation that describes and teaches the effects of the parameters, listed in the two groups of factors, over the variables mentioned in the introduction (a, b, and c). This issue is not discussed in detail in this review but is addressed elsewhere (7).
Investigators using surface EMG should be aware of the possibilities and limitations of the presently available techniques and of the options for the future; this review attempts to provide this information.
Effect of Electrode Location on EMG Variable Estimates
EMG detection with a linear electrode array is a powerful technique to extract useful information from the surface signal and also helps in understanding the appropriate use of a single electrode pair. Consider, for example, the two monopolar potentials of an MU that are propagating in opposite directions in the upper left of Figure 1. They are very similar. If a differential detection is obtained by means of two electrodes placed one on each side of the IZ at the same distance from it, the resulting voltage will be the difference of almost equal time-varying voltages and will be the following:
- Small with respect to the monopolar potentials
- Sensitive to differences in shape between the two propagating waves, which depend on the distribution of the neuromuscular junctions (NMJs) within the IZ (as well as other factors)
- Sensitive to electrode location and to the spread of CV values among MUs (CV distribution)
- Sensitive to relative movements between the electrodes and the muscle fibers, which will alter the symmetry of the detectors relative to the IZ
Consider now a differential detection obtained between two electrodes placed on one side of the IZ, as indicated in the upper right of Figure 1. The resulting voltage is the difference between two very similar monopolar time-shifted voltages and will show the following:
- Smaller or comparable amplitude with respect to the monopolar potentials, depending on the interelectrode distance
- Small sensitivity to the spatial distribution of the NMJ within the IZ because the two monopolar potentials will be equally affected by it and the effect will cancel in the difference between the signals
- Small sensitivity to electrode location and electrode-muscle relative movement as long as the electrodes are sufficiently distant from the innervation and termination zones (at least two or three times the interelectrode distance)
- High sensitivity to CV and to CV distribution, which will affect the time duration of the differential signal and the delay between two such differential signals detected from adjacent electrode pairs
Figure 2a depicts a set of differential signals detected from a biceps brachii using a linear array of electrodes that extends from one tendon to the other. Each trace corresponds to an electrode pair. The electrodes are silver bars 1 mm thick, 5 mm long, and 10 mm apart. Figure 2b shows eight single differential signals detected with different interelectrode distances and locations near the IZ. The following observations are straightforward from Figure 2a: (a) a very small signal is detected by the seventh pair (trace 7 from the top), which is over the IZ; (b) each firing generates a well-defined propagating signal that begins at the IZ (trace 7) and terminates at the muscle-tendon junctions (traces 1 and 14 or 15); and (c) the contributions of individual firings can be detected, recognized, and classified and attributed to the corresponding MUs. Figure 2b shows how the features of signals detected with a single pair of electrodes can change depending on electrode location and interelectrode distance. In particular, it is evident that electrode pairs placed symmetrically over the IZ (such as B, F, and G) provide small and noisy signals, whereas pairs that are on one side (such as A and H) provide larger signals than pairs that span the IZ.
Figure 3 shows the changes in ARV of the differential EMG detected along an electrode array (14). Large variations are observed near the IZs indicating that a reliable estimate may be obtained only with electrodes placed between the innervation and the tendon zones as suggested by the recent European Recommendations for Surface Electromyography (7). Similar observations apply to spectral variables and CV estimates (8,12).
Because the location of the IZ is not known a priori and it differs across individuals, the zone should be identified in each muscle and each individual by either of two methods: (a) by using a linear electrode array and choosing an electrode pair midway between the IZ and a tendon or (b) by using a detection system with only two adjacent pairs of electrodes and verifying that the two signals are delayed and sufficiently similar to justify the assumption of monodirectional propagation. “Sufficient similarity” may be tested using the correlation coefficient between the two signals after they have been aligned. This coefficient should be higher than 0.7–0.8.
From the above considerations, it is evident that improper electrode placement may lead to total inconsistencies in the detection of signal amplitude and spectral features. For example, if the interelectrode distance includes the IZ, a pair of widely separated electrodes may detect smaller signals than a pair of electrodes that are closer together but do not span the IZ. This may be the reason for conflicting results reported in the literature when this factor was not taken into account (7). Improper electrode repositioning after training or treatment could lead to large variations of amplitude and/or spectral variables that could be incorrectly attributed to the effect of training or treatment.
It is evident that electrode locations away from the IZs are also required for a reliable CV estimation.
Values of interelectrode distance in the arrays range from 2 mm (small muscles of the hand or face) to 10 mm for longer limb or back muscles. The size of electrodes used in arrays ranges from 1-mm-diameter pins to 10-mm-long bars made of 1- mm-diameter silver wire. Electrode size and shape also have some effect on amplitude and spectral EMG features, because the detected potential is the average under the electrode area of the potential distribution over the skin. EMG amplitude, MNF, and MDF decrease with increasing electrode area, and this effect is particularly evident for electrodes that extend in the direction of propagation of the action potentials.
Effect of Electrode Misalignment on Surface EMG Features
Electrodes for surface EMG detection should be placed in the direction of the muscle fibers. A misalignment of the electrodes with respect to muscle fibers has great influence on the estimates of the EMG variables. Figure 4 reports a single MUAP detected with a linear array and simulated with a mathematical model for three angles of inclination between the array and the muscle fibers. From the figure it is evident that depending on the location of the electrodes along the muscle, amplitude, MNF and MDF may be overestimated or underestimated. It can also be shown that CV may be overestimated or underestimated in case of misalignment of the electrodes. Proper alignment is characterized by a symmetric waveform pattern of the action potential propagating in the two directions between the end plate and the two tendon regions (Figure 4). Note that with a single pair of electrodes, it is not possible to evaluate the alignment of the electrodes with the fibers.
MYOELECTRIC MANIFESTATIONS OF MUSCLE FATIGUE: A TOOL FOR NONINVASIVE FIBER TYPING?
Muscle fatigue could be considered as associated to: (a) changes in muscle fiber membrane excitability and MUAP propagation (1), (b) alteration of muscle metabolic conditions, and (c) failure of excitation-contraction coupling (3). During a sustained voluntary contraction, the EMG signal progressively changes its characteristics because of the first factor, which is referred to as myoelectric manifestations of muscle fatigue. These changes indicate modifications of muscle fiber membrane properties that might eventually activate the other two factors, resulting in the inability to sustain the required contraction level (mechanical fatigue). The changes are believed to be related to shifts in ionic concentrations and are reflected by decrements of CV and spectral variables and increments (followed by decrements) of amplitude variables. These variables are usually plotted versus time after normalization with respect to a reference value (the initial value or the intercept of a regression function) so that percent changes of different variables can be compared. The resulting graph is referred to as the fatigue plot. A system used in many previous investigations to obtain fatigue plots is depicted in Figure 5. This system estimates amplitude and spectral features from a single differential signal (SD) and CV from two double differential signals (DD1 and DD2) as described previously (10,13). This approach has been used to describe changes during isometric constant force contractions and may not be suitable to describe changes taking place during variable force, intermittent, or dynamic contractions. Research concerning the latter conditions is under way to identify the effects of relative muscle-electrode movement.
Muscles are known to be made up of MUs consisting of two main fiber types with different metabolism and electrophysiological properties. On the basis of work in animals, type II fibers appear to have larger diameters with higher CV and faster decrement of CV during sustained contractions. These relationships (as well as other factors) are reflected by MNF and MDF as demonstrated by Gerdle et al. (4). Many researchers have investigated the relationship between the fatigue plot and the fiber constituency in the attempt to identify a noninvasive procedure to estimate fiber distribution (by size and/or type) in a given muscle. If the findings from animal studies are demonstrated to be applicable to humans, this possibility would be very relevant in sport and geriatric medicine. Although the solution of the problem is not yet available, Figures 6 and 7 depict very promising results in this direction. The results in these figures strongly suggest a correspondence between surface EMG variables and the proportion (by percent area) of type I and type II fibers in the muscle.
Another result indirectly indicating the possibility of noninvasive fiber typing derives from the experiments of Merletti et al. (10) and Hara et al. (6) who demonstrated that, when isometric force level increases from low (20–50% MVC) to high (80–100% MVC) values, an increment of CV and MDF or MNF is observed. This increment is higher in younger subjects and lower in elderly subjects in agreement with bioptic findings indicating a progressive decrement of size or number of type II fibers with age. In addition, the rate of decrement of CV during a sustained contraction is higher in young subjects providing evidence which further supports this hypothesis (10).
The available evidence strongly suggests that there is a relationship between muscle fiber constituency and the values and/or rates of change of estimates of EMG variables during sustained isometric contractions. There has been a tendency to extrapolate these findings to dynamic (eccentric, concentric) contractions. In these cases the situation is more complicated by, a) the fact that the signal is non stationary and traditional spectral analysis must be used with caution, b) relative movement between electrodes and muscle may create artifacts and c) the MU pool may change during movement.
THE SURFACE EMG: STOCHASTIC OR DETERMINISTIC?
A single electrode pair provides a signal that appears random (stochastic) and may indeed have large random components if the two electrodes are placed on opposite sides of the IZ. When multiple signals are obtained from an array, deterministic patterns appear that are clearly associated to the firings of individual MUs. These patterns are evident from Figures 2a and 8. Figure 8 depicts a set of differential signals detected with an array of silver bars (1 mm diameter, 5 mm length, 5 mm spacing) placed on a biceps brachii. The signal segment shows three MUs with two IZs, one between pair 8 and 9 (MUs 1 and 2) and one under pair 4 (MU 3). The straight lines are added to help in the interpretation. It is clear that under certain electrode pairs (4, 5, 6, and 7) signals travel sometimes in one direction (when MUs 1 and 2 fire) and sometimes in the opposite one (when MU 3 fires). Note that a “global” estimation of CV based on only two electrode pairs would be completely incorrect if propagation is not always in the same direction under the electrode pairs. This condition is verified only for pairs 1, 2, and 3 or pairs 10 and 11. In cases such as this, an MUAP by MUAP estimation of CV may be necessary. With appropriate signal processing software, it is possible to identify each MUAP, estimate its CV, and attribute the MUAP to a specific MU. The deterministic features of the EMG, as summation of MUAPs, can therefore be extracted (2,11) and decomposition may be attempted. It has been proven (2,11) that it is possible to identify single MUAP patterns, at least at low muscle activation levels, with manual and automated techniques and associate them to the firings of different MUs. It has been shown how to extract anatomical and physiological features of the single MUs, such as the position of the end plates, of the tendon regions, the length of the fibers, and the CV of the single MUAPs. Research on the separation and classification of MUAPs detected with linear or two-dimensional arrays is under way in many laboratories.
Surface EMG techniques are very powerful but must be applied with proper knowledge of the signal generation and propagation mechanisms. It is very easy to obtain a signal, but incorrect methodology will lead to incorrect conclusions derived from such a signal (1). There is a need for standardization of the methodology with specific regard to electrode location. An effort in this direction has been sponsored by the European Community by means of the Concerted Action on Surface Electromyography for Noninvasive Assessment of Muscles (7). This effort should be further pursued by the research community and extended to dynamic conditions. Surface EMG shows considerable potential for noninvasive fiber typing, assessment of myoelectric manifestations of muscle fatigue, and monitoring of changes induced by training, disuse, and aging (5).
At the present time, surface EMG techniques for muscle characterization are affected by major limitations, and their application is restricted to superficial muscles with parallel fibers. They provide information that is very different from that obtained with needles. Muscles with pinnated fiber architecture, curved fibers, or multiple IZs are difficult to investigate. Electrode arrays should be used to detect the most reliable location for placing a single electrode pair. There is no doubt that future EMG equipment will incorporate flexible linear or two-dimensional arrays and appropriate software for the automatic identification of the most meaningful channels and extraction of clinically relevant information from these channels.
The authors gratefully acknowledge the collaboration of Dr. Roger Enoka in the preparation of this work. Financial support was provided by Compagnia di San Paolo and Fondazione CRT di Torino.
1. 1. De Luca, C.J. The use of surface electromyography
in biomechanics. J. Appl. Biomechan. 13: 135–163, 1997.
2. Farina, D., Fortunato, E. and Merletti, R. Non invasive estimation of motor unit conduction velocity distribution using linear electrode arrays
. IEEE Trans. BME. 47: 380–388, 2000.
3. Gandevia, S., Enoka, R. McComas, A. Stuart, D. and Thomas. C. Fatigue: Neural and Muscular Mechanisms. Norwell, MA: Plenum Press, 1995.
4. 2 Gerdle, B., Henriksson-Larsen, K. Lorentzon, R. and Wretling. M. Dependence of the mean power frequency of the electromyogram on muscle
force and fiber type. Acta Physiol. Scand. 142: 457–465, 1991.
5. Hainaut, K., and Duchateau. J. Muscle fatigue
, effects of training and disuse. Muscle
Nerve. 12: 660–669, 1989.
6. Hara, Y, Findley, T. Sugimoto, A. and Hanayama. K. Muscle fiber conduction velocity
after fatigue in elderly subjects. Electromyogr. Clin. Neurophysiol. 38: 427–435, 1998.
7. Hermens, H., Freriks, B. Merletti, R. Stegeman, D. Blok, J. Rau, G. Disselhorst-Klug, C. and Hägg. G. European Recommendations for Surface Electromyography
. Enschede, NL, 1999. Roessingh Research and Development.
8. Hogrel, J.Y., Duchene, J. and Marini. J. Variability of some SEMG parameter estimates with electrode locations. J. Electromyogr. Kinesiol. 8: 305–315, 1998.
9. Kupa, E., Roy, S. Kandarian, S. and De Luca. C. J. Effects of muscle
fiber type and size on EMG
median frequency and conduction velocity. J. Appl. Physiol. 79: 23–32, 1995.
10. Merletti, R., Lo Conte, L.R. Cisari, C. and Actis. M.V. Age related changes in surface myoelectric signals. Scand. J. Rehab. Med. 25: 25–36, 1992.
11. Merletti, R., Farina, D. and Granata. A. Non-invasive assessment of motor unit properties with linear electrode arrays
. In: Clinical Neurophysiology: From Receptors to Perception, edited by Comi, G. Lücking, C.H. Kimura, J. and Rossini. P.M. New York: Elsevier Science, 1999, p. 293–300.
12. Merletti, R., and Roy. S. Myoelectric and mechanical manifestations of muscle fatigue
in voluntary contractions. J. Sports Phys. Ther. 24: 342–353, 1996.
13. Merletti, R., Knaflitz, M. and De Luca. C.J. Myoelectric manifestations of muscle fatigue
during voluntary and electrically elicited contractions. J. Appl. Physiol. 69: 1810–1820, 1990.
14. Rainoldi, A., Nazzaro, M. Merletti, R. Farina, D. Caruso, I. and Gaudenti. S. Geometrical factors in surface EMG
of the vastus medialis and lateralis muscles. J. Electromyogr. Kinesiol. 10: 327–336, 2000.
15. Sadoyama, T., Masuda, T. Miyata, H. and Katsuta. S. Fiber conduction velocity and fiber composition in human vastus lateralis. Eur. J. Appl. Physiol. 57: 767–771, 1988.
Keywords:© 2001 Lippincott Williams & Wilkins, Inc.
EMG; electromyography; electrode arrays; electrophysiology; muscle; muscle fiber conduction velocity; muscle fatigue