Summary
Whenever a voluntary contraction of muscles is made, the motor cortex has an exceptionally large number of degrees of freedom to perform the task, which is greater than the dimension of space. Muscle synergies can overcome this problem by simplifying patterns of movement. A muscle synergy is defined as a coordinated contraction of a group of muscles. This project aimed to analyse muscle synergies and how they change within subjects, across subjects, across fatigue level and across conditions. Electromyograms were taken in both lower and upper limb studies and processed to create a root mean squared (RMS) amplitude and onset/offset times for each muscle, which was normalised to the primary muscle and compared across subjects. Results showed a high variation in synergies between subjects, with one muscle 2.94 X higher than the primary compared with an average of 0.351 in all other subjects. Variations could be due to prior muscle training in some subjects. Qualitative analysis showed that synergies were alike within subjects performing the same task. Fatigue impacted the synergies used, with inter-subject significant (p<0.05) t-test results in three out of nine muscles. However the effect of fatigue was hard to compare across subjects due to individual differences in synergy recruitment. Temporal analysis showed muscles were activated as a group, with the exception of the same one muscle in each subject. Overall this project helped improve understandings of synergies and their variation, which in turn will contribute towards the application of synergy analysis in rehabilitation of patients with muscle abnormalities.
Key Words: Degrees of Freedom, Muscle Synergy, Electromyography, RMS Amplitude
1. Introduction
1.1 The Motor System and Muscle Synergies
Movements from automatic reflexes, to small postural changes, and to pre-planned tasks all require contraction of muscles. Voluntary movements specifically are coordinated in the basal ganglia, thalamus and cerebral cortex. The activities in the cerebral cortex primarily involve the motor cortex, where it chooses which movements are most appropriate for the task. Neurophysiologist Nikolai Bernstein was the first person to identify a problem in the motor system. He noted that the number of muscles, joint angles, and connections available meant a large number of degrees of freedom, and therefore many possible ways for the motor system to achieve the same movement. In his 1967 paper, he stated "It is clear that the basic difficulties for co-ordination consist precisely in the extreme abundance of degrees of freedom, with which the centre is not at first in a position to deal" (Bernstein, 1967). The number of degrees of freedom is greater than the dimension of space, leading to the idea of redundancy – that multiple motor signals can lead to the same movement. He theorised that the motor system overcomes this problem by organising itself into synergies, and thus simplifying patterns of movement. A muscle synergy is defined as the coordinated activation of a group of muscles to achieve a movement (Cordo et al., 1997).
1.3 Development of the Theories of Muscle Synergies
One theory is that muscle synergies simplify the computational challenge the motor system faces. According to Bizzi and Cheung (2013), muscle synergies are organised into fixed spinal modules, units of interneurons that cause a specific movement pattern when activated. They theorised that voluntary movement relies on the central nervous system (CNS) selecting the most appropriate spinal modules to produce the best movement. This was supported by an experiment by Cheung et al. (2009) involving mildly to moderately impaired stroke survivors with motor cortex lesions. Synergies during voluntary movement were the same in both the affected and unaffected arm. However, there was a significant difference in actual motor performance. They also stated that when learning new synergies, sensory feedback causes new modules to be created that are specific to the individual's biomechanical properties. This fixed synergy theory can explain how the motor system is simplified, overcoming the computational strain faced by the CNS. However, it does not account for the adaptability of synergies in response to sensory feedback, and leaves no room for flexibility within synergies.
Neural modularity is another concept of muscle synergies, where modules are flexible rather than fixed, so a few synergies can be combined to produce the desired movement (Ting and McKay, 2008). This allows for a more complex organisation of the motor system, and flexibility of modules mean they can be adjusted by sensory feedback throughout the task (Zelik et al., 2014).
The organisation of muscle synergies has not yet been proven. Although both theories provide some explanation of the pathways causing muscle synergies, the flexible synergy theory is more desirable, as it accounts for the complexity of the motor system and it's ability to incorporate sensory feedback into synergy recruitment.
1.4 Methods of Obtaining Synergies
The process of obtaining synergies starts by measuring electrical activity in the muscles by electromyography, a method involving placing electrodes on the skin on the surface of the muscle and measuring the potential difference. This produces an electromyographic (EMG) signal for each muscle. This complex signal can be simplified to analyse further, and one processing method involves using computational factorisation algorithms. These algorithms linearize the data and separate the original signal into a small number of synergies, which create the vector sum of the raw data (Tresch et al., 2006). Examples of these factorisation algorithms include principal component analysis (PCA), nonnegative matrix factorization (NMF) and factor analysis (FA).
Although these methods are commonly used to obtain synergies (Bizzi and Cheung 2013, Patel et al. 2016, Roh et al. 2013, Zelik et al. 2014), their accuracy is varied and some methods perform better than others (Tresch et al., 2006). There is also a possibility that data can be missed during the linearization process. Due to these uncertainties, and to gain a new perspective in the analysis of synergies, this study analysed EMG data with no application of factorization algorithms.
1.5 Aims of this Study
In this study two separate human experiments were carried out. The first was a dynamic lower limb study of two parts involving a stepping ergometer, which had been completed prior to this study. The objectives were to first analyse which synergies are being used, then determine if these synergies change for each participant during repetition of the same movement. Another objective was to investigate the effect of fatigue on the participant's choice of synergies.
The second experiment was an upper limb study working in collaboration with the Robotic Manipulation group at the University of Leeds. The aim of this study was to investigate and explain any differences in synergies across a variety of conditions.
The main outcome of this study was to gain a further understanding of synergies and how they differ under various conditions in different people. Understanding differences in individuals can contribute to improved rehabilitation of patients with muscle injuries or diseases. Treatment can be tailored to a patient's specific needs by building an individual synergy profile and focussing on abnormal aspects of their muscle recruitment (Roh et al., 2013). Therefore, the wider impact of this study is to provide information that will contribute to an improved quality of life for these patients.
2. Methods
Both experiments were approved by the Ethics Committee at the University of Leeds and participants gave full consent prior to experiments.
2.1 Lower Limb Study
The experiments were performed prior to the start of this project by a Sports and Exercise Sciences group at the University of Leeds.
4. Discussion
4.1 Inter-subject Variation
The global mean gave a picture of the expected synergy profile but showed a large variance across subjects. When compared against subjects, individual differences in synergies were illustrated. There is a difficulty in comparing these differences with current studies, as the literature tends to focus on EMG data that has been processed with factorization algorithms in order to linearize the data. For example Zelik et al (2014) used algorithm NNMF to obtain synergies from a subject EMG in five different walking tasks. A range of 5-8 synergies were measured for one subject in each of the tasks, and these synergies were expressed in all subjects. Although a direct comparison cannot be made, these underlying synergies could explain the differences between subjects. Different combinations of multiple synergies can be combined to produce an overall signal which is the vector sum of the individual synergies. However this is usually seen across a temporal range in dynamic, multi-stage movements such as grasping (Overduin et al., 2008).
Inter-subject variation could also occur due to prior training of some subjects. Subjects 1, 7 and 8 seemed to have a more 'efficient' synergy profile, with the majority of activation coming from a small number of muscles. Other subjects, such as 4 and 6, had a more equal activation with more muscles contributing to the task. This was supported in a study by Sawers et al (2015), who found a significant difference in synergies between experts and novices performing the same skilled task. Fewer muscles were coactivated in experts than in novices, suggesting an improved efficiency of synergies. Long-term training has also been shown to improve efficiency of primary motor cortex signalling (Picard et al., 2013), implicating an improved efficiency in muscle contractions. Future developments could focus on the synergies two groups of subjects separated by training level, as this study had no information surrounding subjects' training level.
4.2 Intra-subject Variation
RMS values for the MVCs within subject 1 were largely similar, shown through qualitative analysis. This supports the expectation that the same synergies are used when a subject repeats a task. However in the example given, subject 4 showed more variation within the MVCs than subject 1. The same training concept can provide a possible explanation for intra-subject variation. A lack of training in subject 4 may have caused an increase in exploration of possible synergies and therefore increased variation, whereas subject 1 may have relied on previously fine-tuned synergies (Sawers et al., 2015). However, the differences in variation are small and a quantitative analysis was not performed to identify these.
4.3 Effect of Conditions
4.3.1 Fatigue
The phase and ramp tests allowed for a comparison across two conditions and may be an indication of the effect of fatigue on muscle synergy recruitment. The phase-ramp comparison between MVCs for each muscle across all subjects revealed only three muscles out of nine followed a general trend following the ramp application of force. It was difficult to measure significance of changes in synergies, as they were subject specific and could not be generalised. For example, the phase-ramp RMS in one muscle may increase in one subject as a response to fatigue, but may decrease in another subject. Performing multiple t-tests comparing each muscle phase-ramp for each subject was possible, but qualitative analysis was sufficient to show the different synergy recruitment patterns for each subject.
Instead, specific examples allowed a focussed comparison of the effect of fatigue in individual subjects. In subject 6, a decrease in semitendinosus RMS was matched with an increased RMS of rectus femoris and vastus medialis. This change in muscle synergy may be due to compensatory muscle activity. Studies have shown an increase in synergy variance amongst non-fatigued muscles following a fatigue task (Singh and Latash, 2011). This could be due to adaptive changes in motor commands (Strang et al., 2009), leading to a compensatory activation of muscles.
4.3.2 External Conditions
The upper limb study showed that the synergies chosen to perform the same task under different external conditions are different. In the flat-sharp tool analysis, the synergies observed are similar and no drastic changes in synergy composition was found. The significant RMS decrease seen in four of the muscles could be due to the hesitation of the subject, resulting in an overall lower application of force. In the 0 °-55 ° and the hard-soft analysis, the RMS changes are more indicative of a synergy change, as there are small changes in composition of muscle activity. These changes could be due to sensory feedback mechanisms which cause slight alterations to be made when applying the force. The role of sensory feedback was investigated by Patel et al. (2016) in the grasping hand of humans. They found that memory-guided synergies could not sufficiently reproduce real synergies, highlighting the importance of visual and tactile feedback while performing a voluntary task. To develop this further, a synergy analysis could be performed with subjects performing a force application with and without a blindfold. This would show the effect of visual feedback on muscle synergies.
4.4 Temporal Effect
Qualitative analysis of onset and offset times showed a coordinated contraction of the group of muscles with the exception of tibialis anterior, which was maintained across subjects. To develop the temporal analysis further, a waveform analysis could be used to obtain a time-varying synergy (d'Avella and Lacquaniti, 2013). This can be used to identify pathways of activation and gain a further insight into the temporal interaction of muscle synergies.
4.5 Limitations
There were several limitations of this project. Firstly, there was a significant lack of information surrounding subjects in the lower limb study. As this study was conducted prior to the start of the project, there was no way to retrieve information such as gender, fitness level and prior training. Knowing these may have explained variations both within and across subjects. Another limitation was that in the ramp test, fatigue was not directly measured in subjects. This meant that it was unknown if the subjects were truly fatigued, which could impact the comparisons made.
One constraint of the upper limb study was that there was only one participant, due to limited time and resources. It is not known if the synergy profile is similar across populations, however the results were treated as such to give a generalised analysis. In addition, the EMG failure in biceps brachii meant that the synergies could not be compared reliably between conditions for the muscle. Further study should focus on minimizing these limitations for a more accurate analysis.
4.6 Ethical Implications and Societal Impact
This project involved the use of human data and subjects so faces ethical implications. All experiments were approved by the University of Leeds Ethical Review Committee. Informed consent was taken prior to experiments, and participants could withdraw consent at any point throughout. Participant data was anonymised to subject number and no personal information was released. One possible ethical implication was muscle soreness. There was no certain way to prevent this, however participants were made aware of this prior to the experiment.
This research can have a wider societal impact. Firstly, science can benefit from the information provided in this project, and use it for improved knowledge in the area of muscle synergies. Secondly, this research can benefit society. Improved knowledge of muscle synergies and their variation within populations can improve diagnostics and create treatments for muscle diseases or injuries that are tailored to a patient's specific needs, by creating an individual synergy profile.
4.7 Conclusions
In conclusion, this project analysed muscle synergies and how they change both within and across subjects, over different conditions. Synergies were shown to be largely similar within subjects performing the same task, and different when compared across subjects. When comparing across conditions, both fatigue and external conditions affected synergy recruitment. Temporal analysis showed a coordinated activation of a group muscles, which could be analysed further with waveform analysis. Despite limitations, the project aims have been met and a further understanding of synergies and their variation has been acquired. Future studies could look towards the use of muscle synergy profiles in analysing abnormalities in patients with muscle injuries or diseases.