Comparison of muscle synergies before and after 10 minutes of running.

[Purpose] The purpose of this study was to investigate the modular control of locomotor tasks and compared the modules before and after a running intervention.
[Subjects and Methods] Electromyographic measurements were performed on eight young, healthy males engaged in a 60s run on a treadmill at 2.8 m/s before and immediately after the 600s of running intervention. Electromyographic data for 15 trunk and lower-limb muscles on the right side were recorded. Muscle synergies were extracted from the electromyography signals using non-negative matrix factorization.
[Results] Four modules explained the electromyographic activity of all muscles and had the functions of load acceptance (module 1), push-off (module 2), preparation of landing (module 3), and trunk-stabilization activity during the stance phase (module 4). Modules 1, 2, and 3 matched the basic modules reported in previous studies; whereas, module 4 was different before and after the intervention.
[Conclusion] Before the intervention, module 4 engaged the trunk muscles and it was activated in the stance phase during running. However, after the intervention, module 4 engaged the muscles around the pelvis and it was activated after landing. This result suggests that the posture control changes from the trunk muscles to the lower-limb muscles after 10 minutes running.

INTRODUCTION

Human movements are highly complex in terms of both neural activation and biomechanical output because the brain needs to control many musculoskeletal functions. Bernstein proposed the concept of “muscle synergy,” which posits that human motion is controlled through the cooperation of several muscles, rather than through the individual muscles1). The cooperation of several muscles control can reduce information processing in the Central Nervous System (CNS) because the CNS commands only a few modules responsible for muscle synergy2, 3). The synergy can be identified from electromyographic (EMG) patterns recorded from numerous muscles via an algorithm with two components2, 4,5,6,7), “muscle weighting,” which represents the relative weighting of each muscle within each module, and ‘‘activation coefficient,’’ which represents the relative activation of the muscle weights (Fig. 1). Based on the previous investigation of muscle synergies while running, it was concluded that there are five identified modules. These modules have the function, respectively; load acceptance, push-off to end the stance phase, trunk muscles during the transition from the stance to the swing phase, forward swing of the lower limbs during the swing phase, and landing preparation to stabilize the hip and trunk8). These five modules are believed to be the basic modules activated while running9).

Fig. 1.

The concept of extracting muscle synergy M1, M2, and M3 means muscle1, muscle2, and muscle3, respectively. Muscle weighting represents the relative weighting of each muscle within each module and activation coefficient represents the relative activation of the muscle weights. The vertical axis of muscle weighting is the proportion of muscle activity (from 0 to 1) and that of activation coefficient is the activity value of muscle weighting. NMF: Non-negative Matrix Factorization Created from Lacquaniti et al.6)

The incidence of running related injury is 29–56% and the majority of injury occurs due to overuse10,11,12,13). One of the risk factors of running related injury is running distance per week14, 15). As running distance increases, muscle fatigue accumulates. In fact, it has been reported that muscle fatigue of the lower limbs occurs because of extended running; this subsequently changes muscle activity16). The results of the aforementioned study indicated that thigh and calf muscle activity in the fatigued condition was lower than before running and the authors suggested that the alteration of muscle activity might facilitate injury16). As the modules are related to muscle activity, the modules activated while running may change with prolonging running time. However, there has been no study investigating module changes before and after an exercise intervention. We hypothesized that the modules, especially “muscle weighting”, during running changes before and after a running intervention. Thus, the purpose of this study was to investigate whether muscle synergy is changed because of prolonged running.

SUBJECTS AND METHODS

Eight young, healthy males (age: 22.4 ± 2.9 years, height: 174.1 ± 6.5 cm, weight: 65.3 ± 6.3 kg) participated in this study. They all performed physical activities two or three times a week at the recreational level. Exclusion criteria included a history of lower limb disorder, neurological disorder, or lower limb surgery. This study was approved by the Ethics Committee of our university (Ethics ID: 2013-033). All subjects provided informed consent to participate in the study.

Subjects performed stretching and running on a treadmill (R-16S, Alpen Co., Ltd. Japan) as a warm-up for 600s. The running pace was set by the subjects. After warm-up, the EMG measuring equipment was set up. Muscle activity during a 60s run at 2.8m/s on the treadmill was measured before and immediately after a running intervention (defined as pre and post session, respectively). In the running intervention session, subjects ran for 600s and the intensity of the intervention was set at 70% of maximal oxygen consumption using the Karvonen formula. The target heart rate was calculated with the following formula: target heart rate=0.7 ×(220 −subject age −resting heart rate) + resting heart rate. The heart rate was measured using a cardiometer watch (FT2, Polar Co., Japan).

The activities of the trunk and lower extremity muscles were measured using a wireless EMG telemeter system (BioLog DL-5000, S&ME Co., Japan) at 1,000 Hz. Before the surface electrodes were attached, the skin was rubbed with a skin abrasive and alcohol to reduce skin impedance to a level below 2,000 Ω. Pairs of disposable Ag/AgCl surface electrodes (BlueSensor N-00-S, METS Co., Japan) were placed on the rectus abdominis (RA), external oblique (EO), internal oblique/transversus abdominis (IO/TrA), the erector spinae (ES), rectus femoris (RF), vastus lateralis (VL), vastus medialis (VM), gluteus maximus (GMA), gluteus medius (GME), biceps femoris long head (BFL), biceps femoris short head (BFS), tensor fascia lata (TFL), adductor (ADD), tibialis anterior (TA), and gastrocnemius (GAS) muscles on the right side because the right foot was dominant in all the subjects. Before measurement in the pre session, subjects performed maximum voluntary contraction (MVC) tests using manual resistance to normalize the EMG data. Additionally, the electrodes were remained in place during pre and post sessions and intervention.

To divide the running cycle, a 6-camera 3D motion capture system (OQUS, QUALYSIS, Sweden) at 200 Hz was used and it was synchronized with the EMG system. The laboratory (global) orthogonal coordinate system (frame) followed the right-hand rule and had the positive x-direction orientated to the right, the positive y-direction orientated forward, and the positive z-direction orientated vertically upward. Reflective markers (QPM190, QUALYSIS AB, Sweden) were attached to the edge of the toe and the heel of the shoes, and four additional markers were placed on the front and back of the treadmill.

To analyze the EMG data, analysis software (BIMUTAS-Video, Kissei Comtec Co., Ltd., Japan) was used. The raw data were band-pass filtered between 20 to 500 Hz and full-wave rectified. The EMG data was normalized relative to the MVC data. Afterward, we extracted three running cycles with less noise, starting from one right heel contact to the next right heel contact, using coordinate data for the markers. This data was extracted with less disturbance of the waveform during the pre and post sessions. Each running cycle was interpolated to 200 time points. Then, as previously described, non-negative matrix factorization (NMF) was performed to extract muscle synergies based on the study by Lee and Seung17). Matrix factorization minimizes the residual Frobenius norm between the initial matrix and its decomposition, given as: E=W × C + e, where, E is a p-by-n initial matrix (p=number of muscles and n=number of time points), which represents the normalized EMG matrix, W is a p-by-s matrix (s=number of synergies), which represents the “muscle weighting,” C is an s-by-n matrix, which represents the “activation coefficient,” and e is a p-by-n matrix, which is the residual error matrix. ||E −WC|| establishes the Frobenius norm. To avoid the local minima, the algorithm was repeated 10 times for each running cycle of each subject. The lowest cost solution was retained (i.e., minimizing the squared errors between original and reconstructed EMG patterns). The initial matrix E consisted of a cycle for each of the 15 muscles; therefore, E was a matrix with 15 rows and 200 columns. Each line for E and C was normalized according to its maximum value. For each subject, we iterated the analysis by varying the number of synergies between 1 and 15, and then selected the least number of synergies that accounted for >90% of the Variance Accounted For (VAF)18,19,20). VAF was calculated based on the study by Torres-Oviedo et al.18): where, i goes from 1 to n and j goes from 1 to p. Thus, i was from 1 to 15 and j was from 1 to 200 in this study.

To compare W before and after the intervention, the scalar product (SP) was calculated based on the study by Cheung et al21). We defined the module as the same if SP>0.75.

SPSS 21.0 for Windows (IBM, SPSS Statistics 21.0, Japan) was used for data analysis. Paired t-tests were used to compare the pre and post heart rates. The significance level was set at an alpha level of less than 0.05.

RESULTS

The mean heart rate for the post session (161.1 ± 9.0 beats/minute) was significantly higher than that during the pre-session (62.1 ± 6.5 beats/minute, p<0.001). The mean speed of intervention was 3.4 ± 0.2 m/s and the mean distance covered during the 600s running intervention was 2,100 ± 200 m.

From the results of NMF, four or five modules were extracted. To facilitate the comparison of the set of synergies between subjects, four muscle synergies were extracted for all subjects22). Table 1 shows the VAF of the number of muscle synergies and the VAF of the four muscle synergies was for the pre (VAF=91.3 ± 1.2%) and post sessions (VAF=94.6 ± 0.9%). Figure 2 shows the mean of all subjects of extracted muscle synergies. The SP from module1 to module3 was 0.96, 0.86, and 0.83 respectively, and there were no changes between the pre and post sessions. Module1 primarily engaged the knee extensor muscles. It was active after landing and exhibited a peak activity during the stance phase. Overall, this module was associated with weight acceptance. Module2 engaged the hip extensors and GAS and was active during the double floating phase just after right push-off. Module3 mainly engaged the trunk muscles, GMA and TA. This module was active in 80–85% of the running cycle to prepare for landing. The SP of the fourth module was 0.45 and this module was different during the pre and post sessions. Module4 in the pre session, defined as module4pre, primarily engaged the trunk muscles and was active during the stance phase. However, in the post session, module4, defined as module4post, engaged the muscles around the pelvis such as the ES, GMA, BF, and ADD (Table 1).

Fig. 2.

Comparison of extracted muscle weighting and activation coefficient between pre and post session running. Muscle weighting represents the relative weighting of each muscle within each module and activation coefficient represents the relative activation of the muscle weights. The vertical axis of muscle weighting is the proportion of muscle activity (from 0 to 1) and that of activation coefficient is the activity value of muscle weighting. SP: scalar-product, if SP>0.75, these modules are judged to be identical. In this figure, only module4 is different between pre and post. P-flex: planter flexor, K-ext: knee extensor, K-flex: knee flexor, H-ext: hip extensor

Table 1.

The mean of VAF of each number of modules

Number of modules		1	2	3	4	5	6	7
VAF (%)	pre	69.8 ± 5.3	81.5 ± 2.4	86.5 ± 2.1	90.5 ± 1.6	93.0 ± 1.4	95.1 ± 1.1	96.2 ± 0.9
	post	71.0 ± 7.8	81.6 ± 5.7	86.8 ± 4.1	90.5 ± 2.6	93.0 ± 2.2	95.2 ± 1.6	96.6 ± 1.4

The number of modules is decided when the VAF exceeds 90% for the first time. VAF: variance accounted for

DISCUSSION

This study investigated the activity of 15 muscles functioning while running, before and after a 600s running intervention. The main findings of this study were that four modules were active while running and that module4 was different between the pre and post sessions.

We selected the four modules based on Vaz et al.22) to facilitate the comparing the modules between pre and post sessions although previous studies reported that there are five identified modules during running. On the other hand, it is suggested the possibility that long-term training changes modules23,24,25). The subjects performed the several sports activities as swimming, badminton, and track and field at least five years in this study, therefore it is considered that the difference of the number of modules was occurred.

Module 1 was activated to absorb the weight load during the stance phase, and module 2 was activated for push-off. The function of module 3 was to prepare for landing. These modules were roughly similar to the basic modules reported by Cappellini et al.8); therefore, these three modules are speculated to be essential for running because there were no changes following the intervention.

Module 4pre was similar to the basic module. However, the SP for module4 between pre and post sessions was low; hence, module4 activity was different between the pre and post sessions. Module 4pre engaged the trunk muscles and it was activated in the stance phase. Researchers reported that trunk muscles are activated to enhance the rigidity of the trunk, stabilize the lumbar spine and pelvic girdle, and transmit ground reaction forces efficiently during push-off, which ends the stance phase26,27,28). From the activation coefficient of module 4pre, the timing of the trunk muscle activity was similar to that reported by Saunders et al.26); therefore, module 4pre may be activated to stabilize posture through the trunk muscles during the stance phase. Conversely, module 4post engaged the muscles around the pelvis and the activation coefficient indicated that it was activated after landing. Thus, the role of module 4post may be related to postural control for landing impact using the muscles around the pelvis. From our findings, postural control function changed from the trunk muscles to the muscles around the pelvis because of 10 minutes running. In a study targeting long distance running as marathon, it has been indicated that running distance is a risk factor of running related injuries14, 15). Therefore, the alteration of module due to running may be relevant to running related injuries of lower limb.

Turpin et al.29) reported that there was no change for the modules due to muscle fatigue during rowing exercise although our results showed a couple of different modules. Rowing exercise is performed in a stable condition in the sitting position. Our data indicated that the module with an altered function was the posture control module. Running may be a less stable exercise than rowing because of the inherent double floating phase during which the body is unsupported. Additionally, our intervention was not a fatigue protocol. Thus, our data was different from those obtained by Turpin et al29).

As a limitation of this study, there was a possibility that the intensity of intervention was low. Before this study, we intervened the high intensity exercise than this study as preliminary experiments. However, we were unable to attach and maintain the surface electrodes because the subjects perspired while running. Therefore, we performed the study with 70% running intensity and intervened for 10 minutes. Actually, the exercise intensity during sports competitions is higher and the exercise period is longer than the intervention in this study. Our results indicated that module 4pre, which had the role of posture control through the trunk muscles, changed to module 4post, which engaged muscles around the pelvis with 10 minutes running. Therefore, the increased contribution from module 4post may induce running related injury because most running related injuries are caused by overuse12, 13). Future study should investigate the relationship between the modules and prospective injury survey.

In conclusion, we investigated muscle synergy during running using NMF. The module that engaged the trunk muscles to control initial posture while running changed to engage the muscles around the pelvis for posture control after 10 minutes running. The change of posture control function from trunk muscles to muscles around the pelvis may be a risk for running related injuries.