| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
Laboratorio di Ingegneria del Sistema Neuromuscolare, Dipartimento di Elettronica, Politecnico di Torino, Torino, 10129 Italy
Submitted 20 January 2004 ; accepted in final form 8 April 2004
 |
ABSTRACT |
|---|
 TOP ABSTRACT METHODSRESULTSDISCUSSIONGRANTSACKNOWLEDGMENTSREFERENCES |
|---|
surface electromyography; muscle fiber conduction velocity; mean power spectral frequency; motor unit activation order
The detection of surface electromyographic (EMG) signals during electrical stimulation of muscle allows assessment of the peripheral properties of the neuromuscular system without direct involvement of the central nervous system. The resultant surface EMG signal is a compound action potential, termed the M wave. The properties of the M wave depend on factors including the number of active motor units (MUs), the dispersion of their innervation zones, the distribution of MU conduction velocity (CV), the location of the MUs within the muscle, the thickness of the subcutaneous tissue layers, the orientation of the detection system with respect to the muscle fibers, and the intracellular action potential shape. The influence of these factors on M-wave properties is in most cases not trivial and often counterintuitive.
The M wave has been utilized in fatigue studies at constant stimulation current and at varying stimulation-intensity current levels (35–37). It is generally accepted that the duration of the M wave increases during sustained stimulation, and consequently characteristic spectral frequencies [mean (MNF) and median frequency (MDF)] decrease. Average CV and MNF (MDF) are correlated during sustained stimulation, although a larger relative decrease of spectral frequencies with respect to CV has been reported (37).
With nonconstant stimulation currents, the number of active MUs changes over time. When axons of different size are in a bundle and are affected by the same externally applied and progressively increasing current density field, those with lower threshold (greater diameter) will be activated first (21, 39). However, in the case of transcutaneous stimulation, the current density field decreases rapidly with depth. Thus the likelihood that an axon is stimulated is affected by both axon diameter and distance from the stimulating electrode. The experiments on humans report various indications on MU activation order, partly because of the difficulty in comparing results due to different methodologies. It has been shown from indirect observations (25, 43) that MUs are activated in a reverse order, i.e., from the largest to the smallest, when elicited by a transcutaneous current. However, there are also opposite indications (19, 27).
The analysis of M-wave properties may allow the investigation of MU activation modalities with increasing stimulation current. However, the interpretation of M-wave properties is complex. The relation between M-wave amplitude and duration, MNF (MDF), and CV during nonconstant-intensity stimulation is affected by many factors (7), including activation order and anatomic factors. Solomonow et al. (41) showed increasing characteristic frequencies of the M wave with activation of faster MUs, implicitly indicating a correlation between CV and MNF (MDF) during orderly activation in electrically elicited contractions. The conclusions drawn in their study are often used to justify the application of characteristic spectral frequencies as indicators of MU recruitment during voluntary contractions (2, 3). However, these results were obtained from intramuscular EMG recordings and may not apply to surface techniques. Interpretation of experimental M waves with surface-EMG generation models may be useful to better understand the way MUs are activated with transcutaneous electrical stimulation.
The main objectives of this study were 1) to investigate M-wave average rectified value (ARV), spectral properties, and CV during electrically elicited contractions with and without a significant change of the number of active MUs during the contraction; and 2) to analyze MU activation order during transcutaneous electrical stimulation by examining changes in M-wave properties. The study is based on both experimental and simulation procedures.
|
METHODS |
|---|
|
TOPABSTRACTMETHODSRESULTSDISCUSSIONGRANTSACKNOWLEDGMENTSREFERENCES |
|---|
Subjects. Ten male volunteers (mean ± SD: age 26.0 ± 3.1 yr; height 176.9 ± 5.1 cm; weight 71.6 ± 7.9 kg) participated in the study. No subject had known symptoms of neuromuscular disorders. The study was conducted in accordance with the Declaration of Helsinki and approved by the Local Ethics Committee, and written, informed consent was obtained from all participants before inclusion.
EMG recordings and stimulation method. Surface EMG signals were detected with a linear adhesive array (ELSCH008, LISiN-SPES Medica, Milano, Italy) of eight electrodes with 5-mm interelectrode distance in bipolar configuration. The EMG signals were amplified (EMG 16, LISiN-Prima Biomedical & Sport, Treviso, Italy), band-pass filtered (3-dB bandwidth, 10–500 Hz), sampled at 2,048 samples/s per channel, displayed in real time, and converted to digital data by a 12-bit acquisition board. The stimulator triggered acquisition of the M waves so that subsequent M waves could be averaged. Before electrode placement, the skin was treated with abrasive paste (Every, Meditec, Parma, Italy). To assure proper electrode-skin contact, 20 µl of conductive gel was inserted into the electrode cavities of the array with a gel dispenser (Multipette Plus, Eppendorf AG, Hamburg, Germany).
Stimulation was provided by a programmable multichannel neuromuscular stimulator (LISiN, Torino, Italy) equipped with a hybrid output stage. The stimulation waveform was a biphasic symmetric square wave of 304-µs duration. Stimulation frequency was 20 Hz. An adhesive stimulation electrode (model 4071003545, 35 x 35 mm, Meditec) was applied on the main motor point of the biceps brachii muscle while a large electrode (50 x 80 mm) was placed over the antagonist muscle to close the stimulation current loop (monopolar stimulation).
General procedures. The muscle investigated was the biceps brachii of the dominant side (the right for all subjects). The subject's arm was placed in an isometric brace, and the forearm was fixed at 120° (180° being full extension of the forearm). The location of the stimulating electrode was identified by using a pen electrode (1-cm2 surface) to deliver an increasing electrical stimulus. Points were marked on the skin where the mechanical response of the muscle to the stimulation was the largest. The stimulation electrode was placed on the most distal of these locations. The choice of the most distal location was due to the high probability of detecting propagating signals by placing an array distal to this point.
The surface array for EMG detection was located between the stimulation electrode and the distal tendon, and was aligned in the direction of the muscle fibers. The most proximal electrode of the array for EMG detection was 
25 mm distant from the center of the stimulation electrode. The muscle was then stimulated at 2 Hz, and the M waves generated were monitored on a personal computer. The stimulation intensity was increased until the M-wave peak amplitude reached a plateau. It was often observed that, after the plateau of M-wave amplitude, for much larger stimulation currents, M-wave amplitude could increase. The maximal current was identified as the current intensity leading to the first rapid increase of M-wave amplitude, followed by an absence of changes for an increase of >10 mA. This current will be defined supramaximal, for simplicity, although the definition may not be strictly rigorous. The operations that led to the identification of the supramaximal stimulation current were repeated three to four times, and the maximal value measured was assumed as the reference supramaximal current.
M waves were then recorded as the muscle was stimulated at 20 Hz with constant and variable stimulus intensities. The experimental session included 1) stimulation with stimuli of constant current intensity over time delivered for 15 s at 40, 60, 80, and 100% of the supramaximal stimulation intensity; and 2) three linearly increasing stimulation intensities from 0 mA to the supramaximal current in 5 s. Seven contractions were elicited in total. Ten minutes of rest were given to the subjects after the 15-s constant-intensity contractions, and 3 min of rest were allowed after the 5-s-long nonconstant-intensity contractions. The seven contractions were performed in random order. For each subject, the experimental session was repeated in 3 nonconsecutive days. In addition to the above procedures, 4 of the 10 subjects performed 5-s decreasing ramp contractions in which current declined from the supramaximal current level to 0 mA.
Experimental data processing. The stimulation artifact was removed by offline blanking of 3 ms (29). ARV (over 30 ms), MNF, and CV (37) were computed from the detected M waves. MDF led to similar results (not reported) as MNF. The three ramp contractions of the same experimental session did not lead to significantly different results (see RESULTS). Thus they were averaged to increase the signal-to-noise ratio. EMG variables during the constant-intensity current contractions were computed from the average of groups of 10 consecutive M waves. The first 0.5 s of the constant-intensity contractions were disregarded because they occasionally showed artifacts. EMG variables were computed from a set of three consecutive bipolar recordings, with the central bipolar signal used for computing ARV and MNF, whereas CV was estimated from the two double differential signals obtained by subtraction of consecutive bipolar recordings (37). A total of five consecutive sets of three bipolar recordings were obtained from the array signals. From the sets of signals detected by the array, those selected for further analysis corresponded to the largest correlation coefficient between the aligned double differential signals (37).
A regression line fit the change in EMG variables during the constant-intensity contractions. The intercept of the regression line at time = 0 was considered the initial value of the variable, and the slope of the line was used as an estimate of the rate of change over time. Normalized slopes were defined as the slopes divided by their respective initial values and expressed as a percent (37). The normalization allowed the comparison of relative changes across EMG variables.
Statistical analysis.
The experimental data were analyzed using one-, two-, and three-way repeated-measures ANOVA. Significant interactions were followed by post hoc Student-Newman-Keuls (SNK) pairwise comparisons. The 
level for statistical significance was set to P 
0.05. Data are presented as means ± SD or means ± SE, as indicated.
Simulation Analysis
A structure-based model of surface EMG signal generation was used for the simulation analysis (15). The model allowed simulation of MU action potentials detected at the skin surface by electrodes having physical dimensions. The MU and volume conductor properties were chosen as in Ref. 14. Briefly, 65 MUs were randomly located within the muscle, and the number of fibers of the MUs was uniformly distributed between 50 and 450. The surface MU action potentials were summed together to form the M wave. Average muscle fiber length was 130 mm, with the innervation zone occurring half the distance along the fibers and the end plates and tendon endings scattered in a region of 10 mm. The M waves were detected between the average innervation zone location and the distal tendon. The CV values were associated to the MUs so that larger MUs had higher CV values (1).
EMG variables were extracted from the simulated signals in each condition in the same manner as the experimental recordings. The simulated signals were noise free and sampled at 2,048 samples/s. The parameters varied in the simulations were 1) MU activation order, 2) location of the MUs within the muscle, 3) average CV of the active MUs (Gaussian distribution of CV with mean values of 3–5 m/s, 0.5 m/s increments), 4) standard deviation of CV distribution (0.1–0.6 m/s, 0.1 m/s increments), and 5) range of the positions of the centers of the innervation zones (0–15 mm, 5-mm increments).
Three activation orders with respect to MU size (and thus to CV) were simulated: 1) orderly activation, i.e., from the smallest and lowest CV MUs to the largest and highest CV units; 2) inverse activation order; and 3) random activation order. Two activation orders based on MU location were also considered: 1) from superficial to deep MUs (termed as geometrical activation) and 2) random with respect to location. The geometrical activation was based on the distance between the center of the simulated detection system and the center of the MU territory. Each of the three activation orders according to size was simulated with the two activation orders based on location. In all cases, 50 simulations for each condition were performed, changing the location of the MUs within the muscle. This was done according to the constraints of the specific activation order simulated.
|
RESULTS |
|---|
|
TOPABSTRACTMETHODSRESULTSDISCUSSIONGRANTSACKNOWLEDGMENTSREFERENCES |
|---|
The stimulation electrode and the most proximal electrode of the array were at a distance of (means ± SD) 9.3 ± 1.5 and 7.0 ± 1.6 cm from the elbow crease, respectively. The location of the motor point was similar to that of the innervation zone [located in the biceps brachii at 8.7 ± 1.1 cm from the elbow crease, according to the data by Farina et al. (18)]. The supramaximal stimulation current during the three experimental sessions was 23.8 ± 5.2, 23.6 ± 4.1, and 24.3 ± 3.45 mA. A one-way ANOVA (factor: day) of the supramaximal current was not significant. A three-way ANOVA was used to compare estimated CV, MNF, and ARV during the ramp contractions (3 days x 3 trials of ramp stimulation x 4 contraction levels: 40, 60, 80, and 100%). There was no significant difference across the three trials; therefore, the M waves detected during the three ramps were averaged, as indicated in METHODS. Figure 1 shows examples of M waves detected in the different conditions.
|
|
Rates of change of EMG variables during constant-intensity contractions. Both MNF and CV significantly decreased over time during the constant-intensity contractions. The slopes of EMG variables were statistically analyzed with two-way ANOVA (3 days x 4 contraction levels).
CV slope did not vary across day or current level, although the mean value tended to decrease with increasing current (Fig. 2). MNF slope significantly depended on the current level (P < 0.01). The average MNF slopes decreased with current level and were significantly different between 40% and the other current levels (SNK test, P < 0.05) (Fig. 2). ARV slope significantly increased with current intensity (P < 0.001) and was significantly different (SNK test, P < 0.05) between all current levels, except between 60 and 80%.
Normalized CV slopes did not vary across day or current level, although normalized MNF and ARV slopes increased in absolute value as the current level increased. The same statistical differences were observed for the normalized and nonnormalized slopes between the different current levels. CV normalized slopes were significantly smaller in absolute value than MNF normalized slopes at all current levels (Student t-test for dependent samples) (see Fig. 2).
Ramp contractions with decreasing stimulus intensity. Ramp contractions with decreasing stimulus intensity were performed in an additional experimental session on four subjects. Although the data were not sufficient to make statistical comparisons, the EMG variables showed similar trends to those obtained during increasing stimulus intensity. The estimated CV decreased in all cases when the current level decreased; thus the same trend of CV with current level, described above, was observed with linearly decreasing currents.
Simulations
Figure 3 shows representative simulated M waves. In the following, results are reported 1) for the M wave generated by the entire set of MUs and 2) for the M waves generated by progressively increasing the number of active MUs.
|
|
|
|
|
|
DISCUSSION |
|---|
|
TOPABSTRACTMETHODSRESULTSDISCUSSIONGRANTSACKNOWLEDGMENTSREFERENCES |
|---|
M-wave Features with Stimuli of Constant Current
During constant-current stimulation, the number of active MUs was almost stable; changes may have occurred as a consequence of variations in axonal excitability with fatigue (9). However, in the relatively short contractions analyzed, these changes have probably been limited. The duration of the M wave increased during sustained stimulation, as observed in previous studies (37), and resulted in a decrease in MNF and an increase in ARV. The normalized rates of change in CV and MNF were significantly different, which is in agreement with findings previously reported (37). Factors other than CV that decrease MNF during constant current stimulations were likely the increased variability in CVs (Fig. 4) and prolongation of the intracellular action potential shape (5, 6).
The increase in stimulation current level from 40 to 100% resulted in a significant increase in the absolute rate of change in MNF. This result does not necessarily reflect the activation of progressively more fatigable MUs since, with increasing current levels, a larger number of MUs was active with an increased metabolic production and blood flow occlusion. Because membrane muscle-fiber properties are affected by the amount of metabolic products in the extracellular medium, the activation of a larger number of MUs may have caused a faster change in the MU membrane properties with respect to the case of less-active MUs.
M-wave Features During Increasing Current Intensity
Spectral frequencies and estimated average CV did not always show similar trends during electrical stimulation. In particular, although their rate of change was in the same direction with constant-current intensity, they showed opposite behaviors when the number of active MUs changed significantly over time (Fig. 2). Similar trends of CV and MNF may apply in some contraction modalities while being absent in others. Thus generalization of relations found for certain contraction modalities should be done carefully.
The lack of a positive correlation between CV and MNF was also found in voluntary linearly increasing force contractions (14). In this case, the main factor masking a relation between CV and spectral frequencies was the location of the MUs within the muscle. The distance of the MU from the detection electrodes has an influence on the spectral frequencies, although this effect may be smaller on the estimated CV. In stimulated contractions, there are additional factors that may affect the relation between estimated CV and spectral frequencies. Among these are the standard deviation in CV values, which has a minor effect on spectral frequencies in voluntary contractions (12), and the spread in the innervation zone of the active MUs. In addition, MNF tends to decrease with increasing stimulation level with progressive activation from superficial to deep MUs due to the volume conductor effect. Thus spectral frequencies would be expected to decline during progressive MU activation with transcutaneous stimulation, which was observed both in experimental and simulation conditions (Figs. 5–7). The decrease in MNF did not necessarily reflect an inverse activation order by size and CV, as shown by the simulation analysis.
In simulation, with any activation order by location, estimated CV had a trend in line with the activation order by size (Figs. 5– 7). On the contrary, the simulations demonstrated that it is unlikely to observe an increase in MNF with progressive MU activation (Figs. 5–7). Accordingly, the experimental results showed a decreasing trend of MNF with recruitment (Fig. 4) and an increase of CV. The conditions leading to a similar result in the simulated data were an orderly activation by size and activation by location from the superficial to the deep MUs (Figs. 5–7).
Solomonow et al. (41) reported a progressive increase in M-wave MDF during orderly MU activation. However, those experiments were performed with intramuscular EMG electrodes where the effect of the volume conductor was negligible due to the small detection volume of intramuscular recordings. Because the main determinant of characteristic spectral frequencies for intramuscular recordings was MU CV (41), the decrease of MNF in the present study cannot be explained only by the increase in CV standard deviation or the spread in innervation zones of the activated MUs. A volume conduction effect should, therefore, play a major role, which was confirmed in the simulations (Figs. 5–7).
The present study underlines that generalization of the results shown in Ref. 41 to surface EMG recordings is not possible. As for the voluntary contraction case (14, 16), characteristic spectral frequencies of the surface-detected M wave cannot be used as indicators of activation from MUs with low CVs to those with high CVs. Similarly, the characteristic spectral-frequency initial values do not provide an indication of the CV distribution of the active MUs. The phenomena reflected by MNF (MDF) during progressive MU activation are indeed not only related to the MU membrane properties; MNF is sensitive to the distribution of innervation zones (anatomic factor), the standard deviation of the distribution of CVs (which may be related to the muscle portion being stimulated, in addition to the properties of the activated MUs), the activation order by location, and other factors.
MU Activation Order with Transcutaneous Electrical Stimulation
From the experimental analysis, average CV increased with progressive MU activation. Moreover, there was a concomitant marked decrease of spectral frequencies in the same conditions. An increase of CV may indicate the activation from small to large MUs, with CV directly related to MU size (1). A large decrease of MNF indicated activation from superficial to deep muscle layers, in agreement with the decreased current density with increasing distance from the stimulation electrode. The simulation results supported this interpretation.
The interpretation of orderly activation from the analysis of average CV should be discussed with the factors that may affect CV estimates. CV values may have been positively biased by the signal components generated by the extinction of the intracellular action potentials at the tendon endings (end-of-fiber components). This bias increases with the depth of the fibers (13); thus progressive activation by location could have resulted in the observed CV trends. This effect cannot be ruled out, although it is unlikely that it fully explains the results observed. The detection system was placed between the innervation zone and tendon region, where end-of-fiber potentials have minimum effect (13); moreover, the biceps brachii muscle is covered by a relatively thin subcutaneous layer and has long fibers. In the simulations performed, we did not observe an estimated CV increase with MUs recruited randomly or with an inverse order by size, both with and without a specific order by location. Finally, the estimated MNF significantly decreased with increasing current level. If end-of-fiber components were dominant, estimated MNF should have been observed to increase or remain constant after the initial decrease, since end-of-fiber signals have higher frequency content than the propagating part of the potentials (17).
Additionally, repeated stimulation may have influenced muscle-fiber membrane properties, resulting in increased CV (20, 26). Also, this effect should be considered and probably partly determined the increased trend of CV with increasing current. However, if these changes fully explained the increase in CV with current level, a high positive correlation between CV and MNF would be expected (because no effect of the volume conductor is present when the CV of the active units increases). A further problem with this interpretation is that the same trend of increasing CV with current intensity was observed during the constant-intensity contractions and during the decreasing ramp contractions. Thus a likely interpretation of the results obtained is that activation tended to progress from low CV MUs to high CV ones.
Studies on MU activation during contractions elicited by transcutaneous stimulation are controversial. During electrical stimulation, larger axons have lower stimulation thresholds than smaller axons (21, 39). Because larger axons are associated with muscle fibers that have larger diameter (23) and higher CV (1), the activation order should be inverse when the nerve is electrically stimulated. This has been found in studies applying direct nerve stimulation with cuff electrodes. Methods for changing this activation order have also been proposed (10, 11, 40, 45). However, in the case of transcutaneous stimulation, other factors, such as the size of the axonal branches, their distance from the stimulation electrode, and their orientation with respect to the current field, may play a role in determining the activation order. Knaflitz et al. (27) and Feiereisen et al. (19) reported that, in the tibialis anterior muscle, an orderly activation is more likely than a reverse one. Thomas et al. (42) discussed similar conclusions in patients after spinal cord injury with severe MU loss and reinnervation. The present results, obtained from the biceps brachii muscle, are in agreement with those studies.
With respect to previous work, we used simulated signals to investigate the relation between EMG frequency variables and CV. We observed that an orderly activation by size may be associated with a progressive activation by location. In this case, if the MUs were recruited from superficial to deep muscle layers, the observation of the CV increase implies that deeper MUs had higher CV than superficial ones. This partly contrasts with the larger percentage of type II fibers in the superficial than in the deep muscle layers of the biceps brachii muscle (32, 33). MUs comprising type II fibers, having larger recruitment thresholds than type I units, should have higher CV values. However, it may be assumed that activation progressed from small-diameter muscle fibers to large-diameter ones and that this was relatively independent of fiber type. For the tibialis anterior muscle, Henriksson-Larsen et al. (24) found that both type I and II muscle fibers have a larger diameter in the deep than in the superficial muscle layers. These authors suggested that muscle adaptation to physical demand does not occur only at the level of muscle fiber type and number but also by variations in fiber size over the muscle cross section. Moreover, there is evidence that CV depends mainly on fiber diameter rather than on fiber type, being that CV values of the two main fiber types largely overlap (44). In some muscles, an inverse association between muscle-fiber diameter and fiber type occurs. In low back muscles, for example, small-diameter muscle fibers belong to type II units, and this was associated with recruitment from higher CV MUs to lower CV ones (30, 31).
In conclusion, the experimental and model analysis allowed the investigation of the main determinants of M-wave characteristics and the discussion of the relation among the variables used to describe the signal. It was concluded that spectral descriptors of the M wave reflect many other factors other than average CV; thus their trend is not indicative of MU activation. Direct estimation of muscle-fiber CV from the M wave, together with a model-based interpretation of the results, indicated that MUs tended to be activated from low CV to high CV and from the superficial to the deep muscle layers with increasing transcutaneous electrical stimulation of the biceps brachii muscle.
|
GRANTS |
|---|
|
TOPABSTRACTMETHODSRESULTSDISCUSSIONGRANTSACKNOWLEDGMENTSREFERENCES |
|---|
|
ACKNOWLEDGMENTS |
|---|
|
TOPABSTRACTMETHODSRESULTSDISCUSSIONGRANTSACKNOWLEDGMENTSREFERENCES |
|---|
|
FOOTNOTES |
|---|
The costs of publication of this article were defrayed in part by the payment of page charges. The article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
|
REFERENCES |
|---|
|
TOPABSTRACTMETHODSRESULTSDISCUSSIONGRANTSACKNOWLEDGMENTSREFERENCES |
|---|
This article has been cited by other articles:
|
|
 |
|
  R. D. Stewart, T. A. Duhamel, K. P. Foley, J. Ouyang, I. C. Smith, and H. J. Green Protection of muscle membrane excitability during prolonged cycle exercise with glucose supplementation J Appl Physiol, July 1, 2007; 103(1): 331 - 339. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
| Visit Other APS Journals Online |
) of 0 and 10 mm. Left: MUs located randomly within the muscle. Middle: small and low CV MUs located superficially. Right: large and high CV MUs located superficially. AU, Arbitrary units.
