European Journal of Applied Physiology

, Volume 110, Issue 2, pp 415–423

Active biofeedback changes the spatial distribution of upper trapezius muscle activity during computer work

Authors

  • Afshin Samani
    • Laboratory for Ergonomics and Work-related Disorders, Center for Sensory-Motor Interaction (SMI), Department of Health Science and TechnologyAalborg University
    • National Research Centre for the Working Environment
  • Andreas Holtermann
    • National Research Centre for the Working Environment
  • Karen Søgaard
    • Institute of Sports Science and Clinical BiomechanicsUniversity of Southern Denmark
    • Laboratory for Ergonomics and Work-related Disorders, Center for Sensory-Motor Interaction (SMI), Department of Health Science and TechnologyAalborg University
Original Article

DOI: 10.1007/s00421-010-1515-6

Cite this article as:
Samani, A., Holtermann, A., Søgaard, K. et al. Eur J Appl Physiol (2010) 110: 415. doi:10.1007/s00421-010-1515-6

Abstract

The aim of this study was to investigate the spatio-temporal effects of advanced biofeedback by inducing active and passive pauses on the trapezius activity pattern using high-density surface electromyography (HD-EMG). Thirteen healthy male subjects performed computer work with superimposed feedback either eliciting passive (rest) or active (approximately 30% MVC) pauses based on fuzzy logic design and a control session with no feedback. HD-EMG signals of upper trapezius were recorded using a 5 × 13 multichannel electrode grid. From the HD-EMG recordings, two-dimensional maps of root mean square (RMS), relative rest time (RRT) and permuted sample entropy (PeSaEn) were obtained. The centre of gravity (CoG) and entropy of maps were used to quantify changes in the spatial distribution of muscle activity. PeSaEn as a measure of temporal heterogeneity for each channel, decreased over the whole map in response to active pause (P < 0.05) underlining a more homogenous activation pattern. Concomitantly, the CoG of RRT maps moved in caudal direction and the entropy of RMS maps as a measure of spatial heterogeneity over the whole recording grid, increased in response to active pause session compared with control session (no feedback) (P < 0.05). Active pause compared with control resulted in more heterogeneous coordination of trapezius compared with no feedback implying a more uneven spatial distribution of the biomechanical load. The study introduced new aspects in relation to the potential benefit of superimposed muscle contraction in relation to the spatial organization of muscle activity during computer work.

Keywords

BiofeedbackMuscle spatial organisationMuscle subdivisionsPermuted sample entropyNeck–shoulder disorders

Introduction

The assessments of the spatial and temporal changes in the activity of trapezius during computer work are important since discomfort and pain among computer workers are most often located in the shoulder region muscle (Jensen et al. 1999). However, to our knowledge, topographical changes in muscle activity have not been assessed during daily life activities like, e.g. computer work. High-density surface electromyography (HD-EMG) is a non-invasive technique measuring the electrical activity of muscle and allowing spatial as well as temporal assessments of EMG changes. Using HD-EMG recordings, the activation of muscles has been shown to be heterogeneous (Holtermann et al. 2005). The heterogeneity of muscle activation highlights spatial dependency of motor unit recruitment within individual muscles (Farina et al. 2008; Madeleine et al. 2006b). The spatial heterogeneity of muscle activation may have occurred due to peripheral or central mechanisms, i.e. differences in the distribution of the type of motor units within the muscle or in the recruitment order (Holtermann et al. 2005; Kleine et al. 2000). As such, heterogeneity in motor unit activation may imply a beneficial phenomenon as sustained activity of type I fibres during sustained tasks is supposed to lead to muscle damage (Sjøgaard and Søgaard 1998; Visser and van Dieën 2006).

Spatial heterogeneity of muscle activity is reported to play a functional role in terms of higher endurance time in trapezius muscles (Farina et al. 2008; Madeleine and Farina 2008). Additionally, an imposed contraction at higher level during a sustained task is shown to result in longer endurance time and may thus contribute to redistribute muscle load in upper trapezius (Falla and Farina 2007). In line with these findings, superimposed contraction consisting of a brief increase in the exerted force (i.e. active pause) as an alternative to rest (passive pause) is considered to be potentially beneficial (Crenshaw et al. 2006; Samani et al. 2009a). Biofeedback aiming at making computer users aware of a potential muscle overloading (Hermens and Hutten 2002) is a way to address this issue. In contrast to the mentioned potential positive effect of imposed contraction, the introduction of passive pauses during computer work have been found to be insufficient for attaining complete relaxation of the trapezius muscle (Blangsted et al. 2004).

In this study, we investigated the effects of superimposed feedback types (passive and active pauses) generated by advanced biofeedback system on the spatio-temporal organization of the upper trapezius muscle. We hypothesized that active pause will lead to a more heterogeneous pattern of topographical maps extracted from HD-EMG.

Methods and subjects

Subjects

Thirteen subjects participated in this study [mean age 28 (SD 5) years, height 177 (SD 8) cm and body mass 72 (SD 9) kg]. All participants were healthy, right-handed male volunteers experienced with computer use and no history of chronic pain or diseases in the shoulder and neck region. The study was conducted in conformity with the declaration of Helsinki, and experimental procedures were approved by the local ethics committee (N-20070004MCH).

Experimental procedures

Experimental protocol

Once the subjects had received instructions and EMG electrodes were placed (see below), the computer workplace was individually adjusted according to ergonomic guidelines (Kroemer et al. 2001). The recordings were sequentially performed as explained below:

The subjects were asked to (1) sit with both forearms resting on the table for 30 s to compute the resting level of SEMG used for defining the gap threshold in estimating relative rest time (RRT) (Samani et al. 2009a). (2) perform a reference contraction consisting of a 5-kg bilateral shoulder girdle elevation (shrug) for 5 s. A 5-kg load is approximately equivalent to a 30% maximum voluntary contraction level of activation during shoulder elevation (Samani et al. 2009a, b, c). The subjects sat upright on an office chair. Five kilograms weights mounted with handles were placed on each side of the chair. The handles were gripped by the subject holding thus 5 kg on each side during shoulder elevation (reference contraction and active pauses) (3) perform three sessions of 10-min computer mouse-work. Two types of feedback instruction, beside a control trial were used in a randomized order. The two types of feedback instruction during computer work consisted of (1) relaxation for approximately 8 s with palms of the hands on the table (passive), or (2) an isometric bilateral shoulder girdle elevation identical to the reference contraction with a 5-kg load for approximately 8 s (active). Figure 1 illustrates the body posture during computer work and performance of active pause/reference contraction. The feedback instructions were given based on a newly developed advanced biofeedback system. The biofeedback system utilizes fuzzy logic with a set of defined linguistic rules and readjusted membership functions (“LOW”/“MEDIUM”/“HIGH”, see “Appendix”) based on the history of the collected SEMG data (see “Appendix”). The feedback was visually applied by turning a colour bar from green to red displayed on the right side of the screen. During the control session (no feedback), no feedback was applied. The standardized computer work was a timed click and drag computer mouse task previously described in details (Birch et al. 2001).
https://static-content.springer.com/image/art%3A10.1007%2Fs00421-010-1515-6/MediaObjects/421_2010_1515_Fig1_HTML.gif
Fig. 1

Bipolar electrode positions with respect to HD-EMG grid on clavicular and bilateral descending parts of trapezius, body posture during computer work depicting active pauses/reference contraction (dashed lines). Note on the computer screen, mouse work drawing and biofeedback bar

Subjects were also asked to rate continuously their discomfort in the shoulder region during the sessions of computer work using an electronic visual analogue scale (Aalborg University, Aalborg, Denmark) anchored with “0: no discomfort” and “10: extreme discomfort”. Subjects were regularly reminded (approximately every 60 s) to report changes in discomfort level and allowed to change the scale whenever they felt changes in discomfort level. Discomfort may represent both biomechanical and cognitive load, particularly during feedback instances where subjects had to take care of feedback occurrences.

Data acquisition and processing

A force-sensing resistor device (Toptronic, Echternach, Luxembourg) was placed on the mouse click button to determine pause instances in the computer work whilst the feedback instruction took place in response to the biofeedback system.

HD-EMG signals were detected with a semi-disposable adhesive grid of 64 electrodes (LISiN-Spes Medica, Italy, model ELSCH064). The grid consists of 13 rows and 5 columns of electrodes (2-mm diameter, 8-mm inter-electrode distance in both directions) with a missing electrode at the upper right corner serving as the origin of the coordinate system to define electrode location (Fig. 2a).
https://static-content.springer.com/image/art%3A10.1007%2Fs00421-010-1515-6/MediaObjects/421_2010_1515_Fig2_HTML.gif
Fig. 2

Grid for high-density recording of surface electromyographic (HD-EMG) signals from the right upper trapezius muscle. a Schematic description of the HD-EMG recording configuration including the coordinate system and its position on trapezius, b example of root mean square map (interpolation by factor 8) of a representative subject with indication of the innervation zone location (dashed line) and of the position of the centre of gravity (black circle; see text for details on the computation of the centre of gravity) and c illustration of windows location (0–25–50–75–100% of computer work) excluding the pause instances and extracted features from the two-dimensional maps

The silver–silver chloride electrode surfaces in the grid are separated from the skin by a small cavity (~1-mm thick) filled with electrolyte gel. The EMG signals were bipolarly amplified 5,000 times (128-channel surface EMG amplifier, SEA64, LISiN-OT Bioelectronica, Torino, Italy; 3-dB bandwidth, 10–500 Hz), sampled at 2,048 Hz, and A/D converted in 12 bits (resolution 0.2 μV/bit). Before placement of the grid, the main innervation zone of the upper trapezius muscle along the C7-acromion line was identified in a few test contractions with a linear array of 16 electrodes (silver bars, 5-mm long, 1-mm diameter, 5-mm inter-electrode distance) (Farina et al. 2002).

The 64-electrode grid was then placed on the upper trapezius muscle with the fourth row aligned with the C7-acromion line, parallel to the muscle fibre direction. The lateral edge of the grid was 10 mm medial to the identified innervation zone. The lower end of the grid in this position could partly cover the middle part of trapezius. A reference electrode was placed on C7 vertebra. Bipolar signals were computed in the fibre direction; thus 51 bipolar derivations arranged in 13 × 4 were obtained. Additionally, bipolar surface electrodes (Ambu A/S, Neuroline, Ballerup, Denmark) were also aligned (inter-electrodes distance: 2 cm) (1) On the main superior muscle bulk ~20% lateral to the midpoint between the C4 vertebra and posterior lateral third of the clavicle for the clavicular part, (2) ~20% medial to the midpoint between the acromion and the C7 vertebra for the descending (ipsi and contra-lateral sides) part (Fig. 1). Before data analysis, the SEMG signals were amplified (bandwidth, 10–500 Hz), digitally band-pass filtered in the frequency bandwidth 10–400 Hz (4th order Butterworth filter). Furthermore, a notch filter [2nd order Butterworth band stop with rejection width 1 Hz centred at four-first harmonics of the power line frequency (50 Hz)] was used when necessary to remove line interference. Bipolar SEMG from the clavicular and ipsilateral descending parts were used to generate the advanced biofeedback (see “Appendix”) while the contra-lateral descending part was recorded to assess SEMG changes on the contra-lateral side.

Data analysis

Reference voluntary electrical activity (RVE) was calculated as the mean of EMG root mean square (RMS) during the reference contraction over 250-ms epochs moving in steps of 100 ms.

Mouse clicks were used to define periods of computer mouse work and pauses. A detection threshold (0.1 V) was applied to define the onset/offset of mouse clicks. Periods longer than 7 s without a mouse click were considered as pause instances. The pause instances were removed from signal and the following analysis was performed on the signals excluding the pause instances.

RMS of EMG during computer work and mean of discomfort level were calculated over 0.5-s non-overlapping epochs. Normalized RMS (NRMS) obtained by dividing absolute RMS values by RMS values from the RVE were computed. RRT (gap duration set to 250 ms) was computed over 10-s epochs. In addition, permuted sample entropy (PeSaEn), an estimation of sample entropy revealing the complexity of a time-series (Richman and Moorman 2000) was estimated over 0.5-s non-overlapping epochs for all channels. All the processing parts were performed after removal of pause instances.

Topographical maps along the time axis were calculated by applying five equispaced segments (0–25–50–75–100% of computer work time excluding the pause instances) on RMS, NRMS, RRT and PeSaEn for each channel and discomfort level. The segment length was set to 20% of the distance between two consecutives segments and the values within each segment were averaged to represent 51 values of each topographical map. For graphical representation, the 51 values were interpolated by a factor 8 but only the original values were used for data processing and statistics. (Fig. 2b). Figure 2c also illustrates the location of windows (0–25–50–75–100% of computer work) on time axis and the extracted features.

The magnitude as depicted by the mean of squared root of sum of square, the centre of gravity (CoG) and the entropy of RMS, NRMS, RRT and PeSaEn maps were computed. Mean of squared root of sum of square for each map is calculated as \( \sqrt {\frac{1}{51}\sum\limits_{i = 1}^{51} {x_{i}^{2} } } \) where “x” is RMS, NRMS, RRT or PeSaEn.

For the entropy measure, each map was normalized to the map composed of maximum of all topographical maps in each channel and multiplied by 255 (upper limit for an unsigned integer as the image histogram is built on unsigned integer values for each pixel) and then the integer value was taken to depict a grey scale image. Image entropy is a statistical measure of randomness (heterogeneity) and can be used to characterize the texture of the image. Entropy is defined as −sum(p × log2(p)) where p contains the histogram counts in each bin divided by the total number of channels in the map (Gonzalez et al. 2004).

The entropy of each map based on image histogram with 51 bins was then estimated (Farina et al. 2008).

Statistical analysis

Session [feedback consisting of passive, active pauses and control (no feedback)] and time (0–25–50–75–100%) were introduced as factors in a full-factorial repeated measure analysis of variance for the CoG of RMS, RRT and PeSaEn along medial–lateral and caudal–cranial directions as dependent variables. The same analysis was conducted for RMS, NRMS, RRT and PeSaEn image entropy and map magnitude. If the assumption of sphericity was not met, a correction was applied (Greenhouse Geisser with epsilon value below 0.75). In all tests, P < 0.05 was considered significant. Median values [25–75% percentile] are reported.

Results

Magnitude indices

Session played a significant role on the magnitude of RMS, NRMS, RRT and PeSaEn (respectively, F1.06,12.7 = 15.51, P = 0.002; F1.35,16.25 = 26.1, P < 0.001; F1.18,14.2 = 11.37, P = 0.003 and F1.6,13.9 = 9, P = 0.008). Active pause resulted in higher magnitude of RMS and NRMS while the magnitude of RRT and PeSaEn decreased compared with passive pause and control (no feedback) (Fig. 3).
https://static-content.springer.com/image/art%3A10.1007%2Fs00421-010-1515-6/MediaObjects/421_2010_1515_Fig3_HTML.gif
Fig. 3

Median and interquartile range of magnitude of the high-density electromyographic maps for a root mean square (RMS), b normalized RMS (NRMS), c relative rest time (RRT) and d permuted sample entropy (PeSaEn) in relation to superimposed feedback [No (control (No feedback)) / Passive and Active]. *p < 0.05

Spatial features

Time did not play a significant role on any of the assessed measures. Session played a significant role on the CoG of RRT maps along the cranial–caudal direction (F1.28,15.35 = 4.8, P = 0.03). The CoG Y-coordinate of the RRT was located at (48.9 [46.4–64], 48.9 [47–56] and 53.9 [48–75] mm) of the grid for, respectively, no, passive and active pause sessions. Active pause shifted the CoG of RRT maps towards the caudal end of the grid (P < 0.05, Fig. 4).
https://static-content.springer.com/image/art%3A10.1007%2Fs00421-010-1515-6/MediaObjects/421_2010_1515_Fig4_HTML.gif
Fig. 4

Median and interquartile range of centre of gravity (CoG) Y-coordinate in the cranial–caudal direction (CC) of the relative rest time (RRT) high-density electromyographic maps in relation to superimposed feedback [No (control (No feedback)) / Passive and Active]. *p < 0.05

Session also tended to play a significant role on the CoG of RMS, NRMS and PeSaEn along the cranial–caudal direction (respectively, F1.37,16.46 = 3.87, P = 0.055; F1.2,14.45 = 4.16, P = 0.054 and F1.37,16.54 = 3.15, P = 0.08). Active pause tended also to move the CoG Y-coordinate of the RMS and NRMS towards the cranial direction and CoG of PeSaEn towards the caudal direction.

Overall heterogeneity

Session also played a significant role on the RMS and NRMS image Shannon entropy values (respectively, F1.3,15.6 = 8.1, P = 0.01 and F1.2,14.4 = 5.4, P = 0.02). Active pause resulted in higher RMS image entropy compared with control (no feedback) sessions (see Fig. 5).
https://static-content.springer.com/image/art%3A10.1007%2Fs00421-010-1515-6/MediaObjects/421_2010_1515_Fig5_HTML.gif
Fig. 5

Median and interquartile range of magnitude of the Shannon entropy of the a root mean square (RMS), b normalized RMS image in relation to superimposed feedback [No (control (No feedback)) / Passive and Active]. *p < 0.05

Contralateral side

Feedback instruction type played a significant role on RMS and NRMS on the contra-lateral descending part of trapezius (respectively, F2,24 = 12.4, P < 0.001 and F2,24 = 7.9, P = 0.002). RMS and NRMS increased for the contra-lateral descending part during active pause compared with control (no feedback) and for RMS also compared with passive pause. There was significant time × feedback type interaction on the RMS and NRMS from the contra-lateral descending part (respectively, F2.8,33.7 = 3.6, P = 0.02 and F3,36.1 = 4.0, P = 0.01), i.e. RMS and NRMS during sessions with active pause increased over time from (12.6 [5–24] μV and 5.5 [2–9] at 0%) to (18.4 [5.3–35] μV and 6.8 [4.5–21] at 100%), whereas time did not play a significant role during passive and no pause sessions.

Discomfort level

Time played a significant role on discomfort level (F9,108 = 26.9, P < 0.001). It increased monotonically over time from 0.15 [0.004–0.25] at 0% to 2.6 [1.07–3.22] at 100%.

Discussion

Introducing an active pause in response to biofeedback led to higher entropy of RMS maps extracted from HD-EMG. Additionally, active pause lowered the RRT in the cranial end of the recording area. Active pause also led to more regular muscle activation compared with the no feedback and passive pause tasks depicted as lower magnitude of PeSaEn maps from HD-EMG.

Entropy measures

Entropy measures, a way to assess the structure of bio-signal regularity has been used in relation to heterogeneity of muscle activation maps extracted from multi-channel recordings (Farina et al. 2008; Madeleine et al. 2006a; Madeleine and Farina 2008). Higher entropy of the maps is reflecting a more heterogeneous muscular activity. The reported higher entropy values following active pause in response to biofeedback may imply a potential beneficial effect, as increased entropy values are considered healthier (Costa et al. 2002; Vaillancourt and Newell 2002). On the other hand, the magnitude of regularity in HD-EMG time-series increased following active pauses compared with passive pause and control (no feedback). An increase in regularity is synonymous with a decrease in the magnitude of PeSaEn maps in response to active pause. PeSaEn estimates the regularity of temporal variation for each recorded channel which should be distinguished from entropy of maps which represents the spatial heterogeneity over the whole recording grid. This is probably not a favourable trait, but this also emphasizes our limited knowledge regarding the timing and level of active pause aiming at promoting heterogeneity in muscle activation pattern (Samani et al. 2009b). Since the area covered by the electrode grid represented the upper and partly middle trapezius, it may be argued that active pause increases the regularity within a muscle subdivision while it may result in more heterogeneous coordination between different compartments of trapezius. This is of potential interest as we have recently shown an enhanced coupling among trapezius compartments in response to muscle fatigue and delayed onset muscle soreness (Madeleine et al. 2009). Moreover, the present results are in line with the increased regularity of the activation pattern reported in the clavicular and descending parts, and the decreased regularity found in the ascending part of the trapezius muscle (Samani et al. 2009c).

Spatial changes of EMG maps

The current study did not show significant effects of time during the 10-min computer work, but a decrease in the magnitude of relative rest and an increase in the magnitude of RMS during sessions with active pauses were found. These findings might be due to discomfort development during the 10-min computer work. Our previous findings where active pauses were applied at regular time intervals (2 min) did not show any significant effect on RMS or RRT (Samani et al. 2009a). However, in another study with shorter time interval (40 s) between active pauses, a cumulative effect of active pauses was found (Samani et al. 2009b). We may infer that active pauses feedbacks should most likely not be recommended at short interval (e.g. below 2 min). This could be implemented as a constraint in feedback designs promoting heterogeneous muscle activation. Additionally, translation of the present results to a real occupational setting is undeniably challenging as performing frequent active pauses will require devices to facilitate these pauses and may not be accepted by the users.

RRT and RMS are by definition negatively correlated. Then, the shift of the CoG of RRT towards the caudal direction and the trend of the CoG of RMS towards the cranial direction following active pause most likely reflect an increased biomechanical load applied to the upper part of the trapezius. These spatial changes in the pattern of activity were associated with increased heterogeneity in the trapezius activation. Such an association of higher heterogeneity and spatial changes in activity pattern have previously been reported during sustained contraction (Farina et al. 2008). The observed cranial shift in the spatial pattern of activity of the muscle may be due to spatially inhomogeneous changes in motor unit discharge rates or in motor unit recruitment/derecruitment (Westad et al. 2003; Westgaard and De Luca 1999). Moreover, in agreement with previous studies (Farina et al. 2008; Kleine et al. 2000; Madeleine et al. 2006b), no shift of the CoG of the maps was found along the medio-lateral direction due to parallel orientation between the grid and the muscle fibres.

Active pause also increased the EMG amplitude of the contra-lateral descending part of trapezius part. This is in line with studies showing that EMG-based biofeedback from dominant side can cause bilateral modulation of the trapezius activity due to the close neural connections and bilateral activation of the homologous trapezius muscles (Alexander et al. 2007). Holtermann et al. (2008) showed modulation of EMG amplitude in the contra-lateral trapezius due to applied biofeedback to decrease muscle activity. In our study, this modulation only occurred during active pause, and it evolved over time.

In general, the Hennemann principle (Henneman and Olson 1965) is robust and evokes continuous activity of low-threshold motor units at contraction levels as low as those seen during computer work (Søgaard 1995; Søgaard et al. 1996). However, recent findings report that shift in firing rate and changes in recruitment thresholds may point at underlying causes for the change in heterogeneity found in the present study. Westad et al. (2003) showed that the trapezius motor units with sustained firing show a consistent, but variable depression in the firing pattern following a brief superimposed contraction. Moreover, this phenomenon is incorporated with abrupt changes in recruitment threshold of motor units. This can be explained by the activation of motor unit with higher recruitment threshold associated with a decrease in firing threshold (Gorassini et al. 2002). This phenomenon has been attributed to persistent inward currents or plateau potentials. Persistent inward currents contribute to amplify the synaptic inputs and cause the larger motor units to start discharging at relatively lower threshold (Bawa and Murnaghan 2009). This mechanism may eventually lead to a reversal of recruitment order relative to other motor units which are concurrently active. This may represent a motor control adaptation to reduce fatigue in low-threshold motor units during sustained contractions. Superimposed active contractions consisting of bilateral isometric contraction promoted bilateral changes in the trapezius muscle activity pattern most likely highlighting recruitment/derecruitment of active low-threshold motor units.

Conclusion

The study demonstrated that the spatial activation of the trapezius muscle could be reorganized in response to superimposed active contraction, i.e. active pause based on an advanced biofeedback system (fuzzy inference system). Both potentially positive and negative effects of active pause contributed to, respectively, a more heterogeneous coordination of trapezius muscle and a homogeneous localized activation.

Acknowledgements

This work was financially supported by Det Obelske Familiefond, Gigtforeningen and the Danish Agency for Science, Technology and Innovation.

Copyright information

© Springer-Verlag 2010