Header
Header
Article

Evaluating Muscle Synergies With EMG Data and Physics Simulation in the Neurorobotics Platform




doi: 10.3389/fnbot.2022.856797.


eCollection 2022.

Affiliations

Item in Clipboard

Benedikt Feldotto et al.


Front Neurorobot.


.

Abstract

Although we can measure muscle activity and analyze their activation patterns, we understand little about how individual muscles affect the joint torque generated. It is known that they are controlled by circuits in the spinal cord, a system much less well-understood than the cortex. Knowing the contribution of the muscles toward a joint torque would improve our understanding of human limb control. We present a novel framework to examine the control of biomechanics using physics simulations informed by electromyography (EMG) data. These signals drive a virtual musculoskeletal model in the Neurorobotics Platform (NRP), which we then use to evaluate resulting joint torques. We use our framework to analyze raw EMG data collected during an isometric knee extension study to identify synergies that drive a musculoskeletal lower limb model. The resulting knee torques are used as a reference for genetic algorithms (GA) to generate new simulated activation patterns. On the platform the GA finds solutions that generate torques matching those observed. Possible solutions include synergies that are similar to those extracted from the human study. In addition, the GA finds activation patterns that are different from the biological ones while still producing the same knee torque. The NRP forms a highly modular integrated simulation platform allowing these in silico experiments. We argue that our framework allows for research of the neurobiomechanical control of muscles during tasks, which would otherwise not be possible.


Keywords:

EMG; Neurorobotics; biomechanics; muscle control; muscle synergies; simulation; spinal cord.

Conflict of interest statement

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Figures



Figure 1

Human knee extension experiment: Participant study with EMG recordings (left) and experiment replication in the Neurorobotics Platform (right).


Figure 2


Figure 2

Muscles of the simulated lower limbs considered in our experiment: (A) Rectus Femoris, (B) Vastus Lateralis, (C) Vastus Medialis, (D) Semitendinosus, (E) Biceps Femoris, (F) Medial Gastrocnemius, and (G) Tibialis Anterior.


Figure 3


Figure 3

Experiment setup in the Neurorobotics Platform with two distinct control approaches: EMG data from the participant study is processed and applied to the simulated model to record resulting knee forces. In a second step optimal synergies are calculated by Genetic Algorithms using the simulation model as test and evaluation bed.


Figure 4


Figure 4

EMG data knee extension trial localization: (top) Activity spectrogram of all muscle recordings: Frequency spectograms for all recorded muscles are added up to identify overall muscle activation during a participant trial session. (Middle) Summed activity spectrogram: Frequencies are added up at any given time point to spot knee extension trial durations. (Bottom) Identified knee extension trials in original recorded EMG data (colors represent individual muscles): a participant executes several trials of knee extension in a row, we apply signal processing to identify trial duration (vertical lines) and centers (crosses). Y-axes values are in arbitrary units unless stated.


Figure 5


Figure 5

Frequency domain plot: The FFT-generated data is averaged over all trials, which shows that the most prominent frequencies lie within the lower end of the spectrum. All frequencies higher than the computed cut-off point will be ignored in further analysis.


Figure 6


Figure 6

Knee torque evaluation on the simulated musculoskeletal model in the Neurorobotics Platform: Muscles are activated (muscle color coding: blue—no activation, red—maximal activation) by either processed EMG signals or a Genetic Algorithm and results in a muscle generated knee torque (red arrow). A counter acting knee torque (yellow arrow) is iteratively increased until the goal knee angle is reached.


Figure 7


Figure 7

Muscle to knee angle Pearson correlations for multiple patients: Different colors refer to data of different study participants, the Pearson correlation is shown clustered by examined muscle groups. We observe a strong negative muscle activation correlation with the given angle for RF, VL, and VM, contrary ST and TA show a weak positive correlation and no definite correlation can be found common for all participants for BF and MG.


Figure 8


Figure 8

Knee torque for a single subject: generated torques show a common pattern across trials and non-optimal trial executions can be identified.


Figure 9


Figure 9

Positive comparison of natural (extracted from the experimental study) vs. evolved (computed by the Genetic Algorithm) synergies: The Genetic Algorithm computes synergies that recreate the muscle activation of processed EMG data of the participant study closely. (A) Subject aac Angel 90. (B) Subject aad Angel 20.


Figure 10


Figure 10

Negative comparison of natural vs. evolved synergies: The Genetic Algorithm recreates the knee torque of study participants but identifies different synergies to reach the same torque. (A) Subject aad Angel 60. (B) Subject aad Angel 90.

References

    1. Albanese U., Sharma K., Feldotto B., Akl M., Weber S., Mahmud H., et al. . (2018). HBP NeuroRobotics Platform. Geneva: CERN.

    1. Amezquita-Garcia J., Bravo-Zanoguera M., Gonzalez-Navarro F. F., Lopez-Avitia R., Reyna M. A. (2022). Applying machine learning to finger movements using electromyography and visualization in opensim. Sensors 22, 3737. 10.3390/s22103737



      DOI



      PMC



      PubMed

    1. Amunts K., Ebell C., Muller J., Telefont M., Knoll A., Lippert T. (2016). The Human Brain Project: c reating a European research infrastructure to decode the human brain. Neuron 92, 574–581. 10.1016/j.neuron.2016.10.046



      DOI



      PubMed

    1. Anderson F. C., Pandy M. G. (1999). A dynamic optimization solution for vertical jumping in three dimensions. Comput. Methods Biomech. Biomed. Eng. 2, 201–231. 10.1080/10255849908907988



      DOI



      PubMed

    1. Au C., Dunne J. (2013). Gait 2392 and 2354 Models. Stanford, CA: OpenSim.



Source link

Back to top button