A Perspective on Muscle Synergies and Different Theories Related to Their Adaptation

Round 1

Reviewer 1 Report

This is a well written narrative review paper on muscle synergies and the different theories related to their adaptation.

General comments

There are very few references in this paper. Even though this is not a systematic review or meta-analysis, it can be enriched with more content and references to other important publications in the field.

There are already several reviews on muscle synergies, some of them mentioned by the reviewers. I understand that the goal is to set apart from previous reviews by “discussing muscle synergies adaptation based on primarily tested hypothesis”. However, I missed the reference to at least two renowned perspective papers: Ting et al, 2015 (https://doi.org/10.1016/j.neuron.2015.02.042) and Giszter et al, 2015 (http://dx.doi.org/10.1016/j.conb.2015.04.004).

In general, I think the Introduction can be improved. Importantly, sensory feedback information is missing in Figure 1 schematics. As the authors mention, motor behavior does not work as an open-loop control.

Lines 94-96: “We hypothesize that significant alteration in activation coefficients results in alteration in the synergy structure, which could be due to challenging task requirements or underlying neurological conditions.”. Based on the schematics from Figure 1, how can activation coefficients change the structure of muscle synergies?

Lines 123 – 128: “Also, knee flexors and extensors showed higher co-activation during slacklining due to co-contraction for stability. Moreover, we suggest that the variability in sensory feedback due to challenging situations is more likely to result in the alteration of synergy structure. This is not negating the fact that even task specific modulation in the synergy structures is governed by the afferent and efferent drives.” In this regards, authors might also consider the recent work done by Tresch and colleagues, where they show that CNS chooses muscle activations not only to achieve behavioral goals but also to minimize stresses and strains within joints (Alessandro et al, 2018 - https://doi.org/10.7554/eLife.38215.001; Barroso et al, 2019 - https://doi.org/10.1038/s41598-019-56888-9; Alessandro et al, 2020 - https://doi.org/10.1073/pnas.1916578117).

Lines 128-129: “Therefore, muscle synergy analyses during the performance of a certain tasks resulting in a poor performance can be a good marker.” In this context, it was not clear to me how authors reached this conclusion.

Figure 2. The normalized scalar product is also used frequently to calculate the similarity in synergies.

Fiure 3. Instead of “samples”, maybe it makes sense to use “% cycle” or “% task%.

Lines 2016-216: “Therefore, the researchers use computational analyses such as muscle synergies analysis from EMG signals, identifying patterns that may reflect various levels of neural function.”. In this context, it seems that the authors message is that computational analyses such as the analysis of muscle synergies somehow solve the previously mentioned drawbacks of EMG recordings. Please clarify it.

Lines 218-220: “For example, stroke patients exhibit differences in the number of muscle synergies, which may reflect disturbances in neural pathways and are related to deficits in motor function”. This was repeated, in other words, in the previous paragraphs.

Figure 5. Other metrics than VAF (e.g.,r^2) are also used. It would be important to add a title to the upper (e.g., “Task 2”) and lower (e.g., “Task 1”) panels, to help the reader.

Figure 5. Synergies 2 and 3 in the lower panel could perfectly be considered part of the same synergy.

Minor comments

Line 53. “It was seen previously” instead of “it is seen previously”.

Lines 145-146: “activities of daily living (ADL).” If ADL is not mentioned again then it is not needed.

Caption of Figure 4. “Task 1 and Task 2 show differences” instead of “Task 1 and Task 2 shows difference”. It seems that authors meant to say 90% instead of 90.

Line 212: “have” instead of “has”.

Line 232: “We also discussed importance”. Please add “the” between “discussed” and “importance”.

Author Response

• There are very few references in this paper. Even though this is not a systematic review or meta-analysis, it can be enriched with more content and references to other important publications in the field.

We have added more references as suggested by the reviewer. Now there are 40 references.

• There are already several reviews on muscle synergies, some of them mentioned by the reviewers. I understand that the goal is to set apart from previous reviews by “discussing muscle synergies adaptation based on primarily tested hypothesis”. However, I missed the reference to at least two renowned perspective papers: Ting et al, 2015 (https://doi.org/10.1016/j.neuron.2015.02.042) and Giszter et al, 2015 (http://dx.doi.org/10.1016/j.conb.2015.04.004).

We have added some of these manuscripts as suggested.

• In general, I think the Introduction can be improved. Importantly, sensory feedback information is missing in Figure 1 schematics. As the authors mention, motor behavior does not work as an open-loop control.

We have introduced a new figure that is talking about closed loop and modulation in synergies in introduction section. This figure can clearly explain the hypothesis of muscle synergies using feedback loop.

• Lines 94-96: “We hypothesize that significant alteration in activation coefficients results in alteration in the synergy structure, which could be due to challenging task requirements or underlying neurological conditions.”. Based on the schematics from Figure 1, how can activation coefficients change the structure of muscle synergies?

We have changed the figure task specific sensory input can alter synergy structure whereas for neurological conditions feedforward mechanism is subject to this change.

“A schematic of muscle synergy theory. Muscle synergies (S1, S2, S3), Activation Coefficient (C1, C2, C3), Muscles (M1, M2----M6). The figure shows feedforward and feedback mechanism of neural drives (efferent and afferent). The alteration in activation coefficients relates with modulation in supraspinal control directly from the cortex. This kind of alteration is generally observed in neurological conditions. The changes in supraspinal control can also be caused by the modulation in afferent drives carrying feedback information from muscle spindles and surface of skin. The alteration in activation coefficient due to modulation in afferent drives due to changes in task or neurological issues can also cause changes in the structure of muscle synergies.”

• Lines 123 – 128: “Also, knee flexors and extensors showed higher co-activation during slacklining due to co-contraction for stability. Moreover, we suggest that the variability in sensory feedback due to challenging situations is more likely to result in the alteration of synergy structure. This is not negating the fact that even task specific modulation in the synergy structures is governed by the afferent and efferent drives.” In this regards, authors might also consider the recent work done by Tresch and colleagues, where they show that CNS chooses muscle activations not only to achieve behavioral goals but also to minimize stresses and strains within joints (Alessandro et al, 2018 - https://doi.org/10.7554/eLife.38215.001; Barroso et al, 2019 - https://doi.org/10.1038/s41598-019-56888-9; Alessandro et al, 2020 - https://doi.org/10.1073/pnas.1916578117).

We have added some of these manuscripts as suggested.

• Lines 128-129: “Therefore, muscle synergy analyses during the performance of a certain tasks resulting in a poor performance can be a good marker.” In this context, it was not clear to me how authors reached this conclusion.

We have rephrased the statement which makes more sense.

The muscle synergy analyses during the performance of a certain task can provide a muscle coordination strategy for that task. Moreover, additional measures for efficient performance such as gait symmetry, step length, margin of stability, boundary of stability etc. can be used to correlate with the altered synergies and validate their role with efficient performance. Hence, altered synergies with other measures can be studied to understand appropriate muscle coordination for efficient task performance.

• Figure 2. The normalized scalar product is also used frequently to calculate the similarity in synergies.

We do not understand this question. Can reviewer be clearer so that we can make appropriate changes?

• Fiure 3. Instead of “samples”, maybe it makes sense to use “% cycle” or “% task%.

Changes were made as suggested

• Lines 2016-216: “Therefore, the researchers use computational analyses such as muscle synergies analysis from EMG signals, identifying patterns that may reflect various levels of neural function.”. In this context, it seems that the authors message is that computational analyses such as the analysis of muscle synergies somehow solve the previously mentioned drawbacks of EMG recordings. Please clarify it.

Therefore, the researchers use computational analyses such as muscle synergies analysis from EMG signals. The muscle synergy analysis allows researcher to identify patterns that may reflect various levels of neural function in a low dimensional space. In short, the low dimensional space of muscle synergies extracted from EMG makes the interpretation for movement control easier as we can break down a complex movement into its fundamental components.

• Lines 218-220: “For example, stroke patients exhibit differences in the number of muscle synergies, which may reflect disturbances in neural pathways and are related to deficits in motor function”. This was repeated, in other words, in the previous paragraphs.

 Dear reviewer we have removed this sentence as it was creating repetition in the manuscript.

• Figure 5. Other metrics than VAF (e.g.,r^2) are also used. It would be important to add a title to the upper (e.g., “Task 2”) and lower (e.g., “Task 1”) panels, to help the reader.

We have made the changes as suggested

• Figure 5. Synergies 2 and 3 in the lower panel could perfectly be considered part of the same synergy.

Dear reviewer this figure is just to show the changes in dimensionality and how it looks in a synergy space. It does not show any neurophysiological relevance. The data is a sample data from a task. We agree with you but the loadings values within these synergies shows some significant alteration for some muscles. That is why we kept them separate.  

Minor comments

• Line 53. “It was seen previously” instead of “it is seen previously”.

We corrected it to “It was seen previously”.

• Lines 145-146: “activities of daily living (ADL).” If ADL is not mentioned again then it is not needed.

We deleted (ADL).

• Caption of Figure 4. “Task 1 and Task 2 show differences” instead of “Task 1 and Task 2 shows difference”. It seems that authors meant to say 90% instead of 90.

We corrected it.

• Line 212: “have” instead of “has”.

We corrected to have instead of has

• Line 232: “We also discussed importance”. Please add “the” between “discussed” and “importance”.

We corrected it.

Reviewer 2 Report

This manuscript first gave a narrative account of muscle synergy, with a focus on its intuition rather than mathematical details. Then it expanded on the robustness and variation of muscle synergy across different conditions,  including neurological diseases such as stroke and so on. The later parts would be most useful and relevant for readers who want to take advantage of muscle synergy in engineering/medical applications. The article was short. To someone who already knew of muscle synergy it could be unsatisfying. I get a bit confused whether this paper should be called a “narrative review” as stated by the authors, or it should stay as “perspective” as shown in the title. Above all, my comments/suggestions below are anchored to a perspective article. 

First and foremost, the value proposition was not clear but it must be. Whether it’s a good idea to use muscle synergy in biomechanics/medical applications? Are the tools ready enough? What kind of theoretical preparations do we have? Overall, the perspective must include whether muscle synergy is a worthwhile concept to follow. Without pinpointing these questions, it reads bland, with all due respect.

Changes in muscle synergy across tasks reflect the task-specific nature of the task, rather than a choice of the modularity in control. I am glad that the authors talked about this. Could the authors review more on what happens when the same task was given to subjects, but their muscle synergies (especially the vector patterns) are different? Since synergy can be robust among healthy controls, it is unavoidable to add literature on disease and patients. 

Minor comments:

- The Abstract started with “locomotor system” but the article also covered upper-limb. Consider reword the abstract to reconcile the discrepancy. 
- Figs 2 and 3, what were the sources of EMG data? Were they collected, simulated, or adapted from somewhere?
- Line 148-149, there have been several studies on synergy similarities already. I would cite the relevant ones.
- Line 167-168, using muscle synergies in rehabilitation is a strong case to make, definitely put references in.
- The distribution of literature seemed unbalanced. For example, the part with badminton had four references. Is this truly necessary?

Author Response

• First and foremost, the value proposition was not clear but it must be. Whether it’s a good idea to use muscle synergy in biomechanics/medical applications? Are the tools ready enough? What kind of theoretical preparations do we have? Overall, the perspective must include whether muscle synergy is a worthwhile concept to follow. Without pinpointing these questions, it reads bland, with all due respect.

The muscle synergies concept plays a crucial role in clinical field and rehabilitation. The theory of muscle synergies can be used as diagnostic technique, and for developing assistive technologies.

For diagnostic technique, we have presented some literature where differences in the number of muscle synergies for hemiparetic stroke patients, cerebral palsy, and spinal cord injury has been observed. The number and structure of muscle synergies have reflected disruptions in descending neural pathways and are associated with the deficits in motor function ¹. Hence, muscle synergy analysis may provide a better view to the clinician, physiotherapist about changes occurring at the CNS and muscular level. This kind of information may help in diagnostics of neurological disorders and assist clinicians and physiotherapist in developing rehabilitation interventions for such patients ².

For developing the assistive technology, some researchers used the concept of muscle synergies to control the myoelectrical assistive technology (AT) because, from a computational point of view, a modular organization based on muscle synergies reduces the computational burden for the controller. This burden is due to the higher dimensionality of neural input in the machine interface. This causes non-intuitive control of myoelectrical AT. The people who use the myoelectric controlled assistive technology (AT) for upper extremities suffer from difficulties in controlling these technologies in daily life because of non-intuitive control. The dimensionality reduction of EMG signals simplifies motor control and learning for myoelectric prosthesis, and it may contribute to the adaptability observed in biological systems ³. The roboticists and control engineers are implementing the concept of muscle synergies to develop artificially intelligent controllers for myoelectric devices. In addition to the dimensionality reduction, the modularity of such a scheme has the advantage to improve the device performance by introducing additional synergies to the controller. Thus, researchers propose that intuitive control may be accomplished via a limited number of these fixed muscle synergies and can be updated to perform more complex task by adding more synergy or reshaping existing synergies.

Alternate theories

There are two alternate theories based on different computational and theoretical model that we can present and talk about.

a). Uncontrolled manifold hypothesis (UCM): As observed by Bernstein, the degrees of freedom involved in a postural or movement task exceeds the number of control variables necessary to describe the task ⁴. The UCM hypothesis states that the CNS focuses on controlling the task relevant DOF and leaves others uncontrolled.

The UCM hypothesis shows that in a high-dimensional space of elemental variables, the controller represents and coordinates in that space a subspace depending on a desired value of a specific execution variable. Moreover, the controller arranges co-variation amongst the elemental variables in such a method that their variance in mainly constrained to the UCM. This will be explained as stabilizing the execution variable ⁵. Several researchers utilized the approach of the UCM hypothesis to quantify synergies stabilizing a difference of execution variables by various groups of elemental variables, kinetic, kinematic, and electromyographic ⁶ in a different movements.

The concept of this theory trying to be characterizing various degrees of freedom depending on their control-theoretical stability. The researchers called these motion variables “more stable” or “controlled” that structure fluctuations in joint space, this kind of that variations of joint conformation maintaining the values of these variables are less constrained than changes in joint conformations that alter the values of the task variables ⁷.

b). The equilibrium point hypothesis (EPH): The EPH explains how motor neuron threshold have been controlled out of the value of lambda (λ), which corelated to an organization for a muscle, joint, or group of joints ⁸. The researchers consider the combination of posture and movement control into a single mechanism as a very important characteristic of the equilibrium point hypothesis ⁸. Some researchers defined the equilibrium point hypothesis (EPH) as method that the central nervous system utilized to control the movement of extremities via a simple shift in equilibrium place and the central nervous system doesn’t need to recompense for task dynamics ⁹.

• Changes in muscle synergy across tasks reflect the task-specific nature of the task, rather than a choice of the modularity in control. I am glad that the authors talked about this. Could the authors review more on what happens when the same task was given to subjects, but their muscle synergies (especially the vector patterns) are different? Since synergy can be robust among healthy controls, it is unavoidable to add literature on disease and patients. 

There are different designs of inter subject variability arises from muscular skeletal differences, so we can see the muscle synergies are different for the subjects when they perform same task.

Minor comments:

• The Abstract started with “locomotor system” but the article also covered upper-limb. Consider reword the abstract to reconcile the discrepancy. 

We changed it.

• Figs 2 and 3, what were the sources of EMG data? Were they collected, simulated, or adapted from somewhere?

We collected the data in the lab and we used this data in our research  

• Line 148-149, there have been several studies on synergy similarities already. I would cite the relevant ones.

We added two relevant references ¹⁰,¹¹.

• Line 167-168, using muscle synergies in rehabilitation is a strong case to make, definitely put references in.

We added two references ²,¹².

• The distribution of literature seemed unbalanced. For example, the part with badminton had four references. Is this truly necessary?

We removed three references ¹³.

Round 2

Reviewer 2 Report

Most of my concerns were addressed.

I appreciate the effort of adding competing theories for synergy-based control. However, UCM and EPH are by no means an exhaustive list. It's fine to dissect these two if the authors deem they are the most relevant / comparable theories. But a few others should at least be mentioned and referenced, such as the optimal control theory, internal model, hierachical control... This is a good opportunity to pico-review how muscles should be organized during movement. Not all theories in motor control have the specificity down to the muscle level. Highlighting this should shed light to the uniqueness and value of muscle synergy.

Author Response

• I appreciate the effort of adding competing theories for synergy-based control. However, UCM and EPH are by no means an exhaustive list. It's fine to dissect these two if the authors deem they are the most relevant / comparable theories. But a few others should at least be mentioned and referenced, such as the optimal control theory, internal model, hierachical control... This is a good opportunity to pico-review how muscles should be organized during movement. Not all theories in motor control have the specificity down to the muscle level. Highlighting this should shed light to the uniqueness and value of muscle synergy.

We added internal model theory only because our manuscript is narrative perspective. It is discussing the concept of muscle synergy for this reason we don’t like to go further.