The effect of head position, electrode site, movement and smoothing window in the determination of a reliable maximum voluntary activation of the upper trapezius muscle

Abstract

Quantitative measures derived from raw myoelectric signal (MES) data must be normalized to allow for comparisons both within and between subjects. The most common method of normalization involves dividing the root mean square (RMS) amplitude of the MES recorded during a given activity by the RMS of the MES elicited during a maximal voluntary isometric contraction (MVIC) of that particular muscle. The objective of this study was to use surface-recorded MES amplitude to determine the combination of electrode site, test position, head posture, and smoothing window that elicits the highest and most reliable MES amplitudes during an MVIC of the upper trapezius (UT) muscle.

Ten volunteers had surface electrodes positioned at five sites on the UT of their dominant side. Three trials of each of three MVIC test positions were performed both with the head in neutral and rotated 45° to the contralateral side. A repeated-measures ANOVA was used for statistical hypothesis testing. Coefficients of variation were used to quantify the between-factor variability introduced in each case. In addition, the data were re-analyzed using moving windows of 100 to 500 ms in length, and an ANOVA was used to determine the effect of window length on both the amplitude and variability of the estimates of maximum voluntary activation (MVE).

Head position had no significant effect on RMS amplitude of the MVIC in any of the test positions. There was a significant electrode site by test position interaction effect. Bonferroni post-hoc analyses were performed on this interaction by fixing test position and testing electrode site, revealing that Sites 1 (2 cm lateral to the midpoint between C7 spinous process and the posterolateral border of the acromion) and 4 (2 cm posterior to Site 1) recorded significantly higher RMS values for all test positions, and were not significantly different from each other. Fixing electrode site, the test position analysis revealed that abduction of the humerus, and abduction with external rotation of the humerus produced significantly higher RMS values than shoulder elevation at both Sites 1 and 4, and that abduction produced a significantly higher RMS amplitude than abduction in external rotation at Site 1. The results confirmed that Sites 1 and 4 consistently produced the highest MES amplitudes for all movements. Pure abduction consistently elicited maximal RMS values; however there is concern regarding supraspinatus cross talk during this movement. Site 1 was found to produce the most reliable data. A moving window of 100 ms was found to generate MVE estimates that were significantly higher than windows ranging from 200 ms to 500 ms in length. There was no effect of window length on the reliability of the MVEs.

Based on this study, it was concluded that abduction or abduction with the arms in lateral rotation should be used as normalization contraction positions for the upper trapezius muscle. During this movement, Site 1 data smoothed with a moving window of 100 ms produces the highest amplitude MVE data but window lengths greater than 200 ms produce more stable estimates in terms of being able to compare studies in which moving windows are used to compute RMS.

Introduction

Surface electromyography is commonly used in biomechanics and ergonomics research to measure muscle activity of the upper trapezius (UT) muscle [9], [19]. Unfortunately, since the acquired myoelectric signal (MES) is dependent on several recording parameters, it is inappropriate to compare the raw amplitude of the recorded muscle activity among subjects, among muscles on the same subject, or from same muscle on the same subject on different days [11]. This difficulty arises because the amplitude of the raw EMG signal is highly variable, and depends on factors such as the exact location and separation of the electrodes, the tissue depth between the electrodes and the active motor units, the location of the active motor units, the muscle temperature, subject posture, and the state of fatigue within the active motor unit pool. In order to make valid comparisons, it is necessary to normalize MES amplitude relative to some standard level of activity [11], [15]. The most common method of normalization is to divide the representative amplitude of the MES by the amplitude of the MES elicited during a maximal voluntary isometric contraction (MVIC) of the same muscle [11]. Normalization schemes often use the root mean squared (RMS) level to estimate MES amplitude over a specified time period [4]. The relative value is represented as a percentage of the maximum voluntary electrical activation (MVE), and is described as the percentage of maximal electrical activation (%MVE) computed by:

$$\text{\%MVE=}\text{RMS}{}_{\mathrm{<mtext>ACT</mtext>}}\text{RMS}{}_{\mathrm{<mtext>MAX</mtext>}}\text{×100,}$$

where RMS_ACT is the RMS amplitude of the MES recorded during the activity of interest, and RMS_MAX is the RMS of the MES recorded during an MVIC. In order for this normalization equation to be accurate, the window lengths over which the RMS values are computed should be the same size for both the activity of interest and for the MVIC [15].

The use of MVEs for normalization of the UT MES amplitude is inappropriate if it is not a true and stable representation of the greatest muscle activity that can be elicited. This is extremely relevant when comparisons are made among research studies that have used different methods to elicit and record an MVE of the UT. Modifiable factors that may influence the amplitude of the MES produced by an MVIC of the UT include: the position and spacing of the electrodes, the movement chosen to perform the contraction, the head position of the individual while the contraction is performed, and the window length used in the analysis of the RMS amplitude of the contraction. Several quality studies have been performed to investigate the optimum combination of electrode positioning, electrode spacing, and test movement (for an excellent review, see Mathiassen et al., 1995 [15]). To our knowledge, there have been no studies in which the effect of head position and its potential interaction with electrode site or movement is considered. The first objective of this work was to investigate the effects of electrode site, movement and head position on the amplitude and relative stability of MVE estimates recorded from the UT. The second objective of this work was to investigate the effect of the moving window length used to compute MVEs from data recorded at the UT.

The surface-recorded MES is the electric representation of the global neuromuscular activity of a contracting muscle [10] and represents a summation of action potentials propagating from many motor units that are activated during a particular movement. It can offer valuable information concerning the timing and relative intensity of muscular activity [4]. The effectiveness of MES in providing information regarding the contractile state of muscle tissue has been demonstrated using the RMS amplitude [12]. The RMS value is dependent on the number of motor units (MUs) firing within the pick up area of the electrodes, the firing rates of the MUs, the size of the MUs, the MU action potential duration, the propagation velocity of the electrical signal, the electrode configuration, and the instrumentation characteristics [12]. For this reason, when studying assessment or treatment protocols using the MES, quantitative measurements from the raw signal are generally of little value.

A standard electrode position is essential for comparison among studies involving MES recordings from the upper trapezius (UT) muscle. In exploring the literature one finds that, despite publication of several recommendations for standard electrode position, one unique site does not exist. In many studies [2], [19], [23], the electrode sites used are only vaguely described.

Early publications by Zipp emphasized the need for a standard electrode position that were located on the muscle bulk in a bipolar configuration [24]. He recommended a standard electrode position for the UT as the midpoint along a “lead line” between the spine of the 7th cervical vertebra and the acromion, but presented no empirical evidence to support the selection of this site. Several researchers have since demonstrated that there is a dip in EMG amplitude at the midpoint of this lead line [7], [21], [22], which has been attributed to the bipolar electrodes straddling the innervation zone, thus resulting in common mode rejection of the motor unit action potential waveforms spreading from the end plate in both directions [7].

Jensen et al. used a 16-channel bipolar electrode array placed along the lead line, and found a plateau region of the highest MES activity with the least amount of variation at 2 cm to 2.5 cm lateral to the midpoint of this lead line [7]. The plateau region was similar when MVICs were performed with the humerus in 0° abduction and in 90° abduction, suggesting that this position would make a better standard electrode position for UT than the Zipp position [7]. Though Jensen et al. showed greater activity 2 cm lateral to the midpoint, [7], [9] this may have been due to the combined activity of UT and supraspinatus, since the UT muscle is thinner laterally. A position more medial, anterior, or posterior to this position may actually give a more valid estimation of UT activity due to a reduction in cross talk from the supraspinatus muscle [9]. Similarly, Hermens and Spaepen suggested that cross talk was less of an issue when shoulder flexion movements are used to generate UT contraction [5]. Work by Jensen et al. [7], [8], [9] suggests that a single optimal electrode position is unlikely to exist for the UT muscle, however, an optimal electrode position for each normalization movement may exist.

Shoulder elevation against manual resistance has been the preferred clinical test contraction in manual muscle testing [1], however the UT has multiple functions. UT functions as a prime mover during elevation and rotation of the scapula [18], and as a stabilizer of the scapula during glenohumeral movements [18] and during glenohumeral torque production [14]. The UT also functions bilaterally to extend the neck and unilaterally to side flex the head [13]. It is not known which of these activities (or which combination) will generate a true MVIC of this muscle. The influence of head position on UT activity recorded during shoulder movements is also unknown.

Schuldt and Harms-Ringdahl [19] and Whalen [23] showed that an MVIC of abduction in the scapular plane produced the greatest MES activity in the UT. No significant differences were found between shoulder elevation and abduction, which may have been due to the variability of activation levels found among individuals [17]. Inadequate descriptions of electrode positioning, muscle loading and methodology limit the usefulness of their conclusion that resisted abduction in the scapular plane is the best position to elicit an MVIC of UT.

Using the Zipp electrode position [24], Nieminen used standardized unilateral MVICs for shoulder elevation, flexion and abduction [17], and his results revealed higher UT MES activity for all flexion positions, followed by shoulder elevation and abduction respectively, however no statistical tests were presented. The lower MES values during abduction were most likely due to their positioning of the arm in 0° of abduction, as the scapula generally does not begin to rotate laterally until about 30° of humeral abduction has been achieved [20], therefore UT would have been acting primarily as a stabilizer, not as a prime mover during this test. The authors did report a large degree of variability in activity between subjects in the elevation and abduction contractions. They stated that the unilateral resistance created difficulty controlling the effect of lateral bending and that bilateral resistance might compensate for this variability.

The combined results from Jensen and Westgaard’s well-controlled studies [7], [8], [9] suggest that normalization procedures using MVICs should include both abduction and elevation contractions, but not flexion. In general, differences in protocols, standardization of contractions, type of resistance and electrode positions limit these generalizations. Mathiassen et al. [15] concur with Jensen et al. in claiming that different parts of the UT muscle may be differentially activated during different MVICs (functional subdivision), and that finding one true maximum may be difficult, depending on electrode location and test position.

None of the studies discussed above directly stated the position of the glenohumeral joint in terms of its rotation during the abduction movements. Illustrations in the studies by Schuldt and Harms-Ringdahl [20], Nieminen et. al. [17] and Jensen and Westgaard [9] represent the position as one of neutral rotation. In this position the supraspinatus muscle may contribute to abduction of the glenohumeral joint, which has led to concerns about EMG cross talk when measuring UT activity as discussed earlier. It may be possible to reduce supraspinatus cross talk by placing the humerus in lateral rotation while performing the abduction contraction. This would change the orientation of the supraspinatus muscle and possibly limit its ability to participate in abduction.

Another possible source of variation in one’s ability to generate an MVIC is the influence of head position during testing. None of the publications discussed above described a standardized head position. While it may be assumed that all subjects maintained a neutral head position, there may have been differences in head positions between subjects and between studies. Different positions or movements of the head would change the length tension characteristics of UT, resulting in differences in activation levels. UT functions unilaterally to side flex the head [13] therefore, use of bilateral resistance while keeping the head in the midline should prevent side flexion, as well as prevent the variability in activation levels for abduction and elevation, as noted by Nieminen et al. [17]. Rotation of the head may also influence activity of UT based on primitive reflex activity, notably the asymmetric tonic neck reflex (ATNR), which results in ipsilateral shoulder abduction and elbow extension when the head is turned to one side and contralateral shoulder abduction and elbow flexion [6]. As has been demonstrated, the UT is active during shoulder abduction [9], [19] thus the influence of ATNR may increase the excitability of the UT, resulting in increased EMG activity with the head turned away from the side of UT recording.

In the review by Mathiassen et al. [15] the effect of smoothing on signal amplitude is discussed, and it is concluded that “the signal smoothing technique may have profound effects on the validity of EMG amplitude normalization” [15]. Jensen et al. showed that the signal amplitude decreased by approximately 15% as the moving window length increased from 200 ms to 1 s. Approximately half of the studies presented in the Mathiassen et al. article [15] used RMS methods to determine signal amplitude, and in many instances (11 of 24 studies), the window length used for smoothing was not reported. Among those window lengths reported, 50 ms was the smallest window length, and 200 ms was the largest. Mathiassen et al. recommended that a moving window of 100 ms be used for smoothing, however there was no empirical evidence presented which supported this recommendation.

Despite the great number of studies into the optimal combination of electrode site and test movement to produce an MVE of the UT that is representative of a true maximal activation [15], there has been little discussion of the differences in the reliability of the data recorded using combinations of these factors. One objective of this study was to use surface MES RMS amplitude to determine the combination of electrode position, MVIC movement, and head position that elicits maximal and stable activity of the upper trapezius muscle. A second objective of this study was to investigate the effect of moving window length on the amplitude and variability of the estimates of MVE computed using the RMS amplitude.