Neural Network Models for Spinal Implementation of Muscle Synergies.

Yunqing Song,Masaya Hirashima,Tomohiko Takei

doi:10.3389/fnsys.2022.800628

Abstract

Muscle synergies have been proposed as functional modules to simplify the complexity of body motor control; however, their neural implementation is still unclear. Converging evidence suggests that output projections of the spinal premotor interneurons (PreM-INs) underlie the formation of muscle synergies, but they exhibit a substantial variation across neurons and exclude standard models assuming a small number of unitary “modules” in the spinal cord. Here we compared neural network models for muscle synergies to seek a biologically plausible model that reconciles previous clinical and electrophysiological findings. We examined three neural network models: one with random connections (non-synergy model), one with a small number of spinal synergies (simple synergy model), and one with a large number of spinal neurons representing muscle synergies with a certain variation (population synergy model). We found that the simple and population synergy models emulate the robustness of muscle synergies against cortical stroke observed in human stroke patients. Furthermore, the size of the spinal variation of the population synergy matched well with the variation in spinal PreM-INs recorded in monkeys. These results suggest that a spinal population with moderate variation is a biologically plausible model for the neural implementation of muscle synergies.

Highlights

Our body is remarkably complex, yet we display a highly stable motor performance
As Hirashima and Oya (2016) demonstrated previously, the nonsynergy model reproduced a shift in the preferred direction of muscles (Figure 1D) relative to their mechanical directions (Figure 1C), and the muscle activation patterns were successfully reconstructed with a linear combination of four muscle synergies (VAF ≥ 0.9, Figures 1E,F)
In experiment 1, we compared two types of neural network models: one with random connections and the other with a smaller number of spinal synergies. We found that both models achieved comparable task performance and similar muscle synergies, confirming the prediction of a previous study (Hirashima and Oya, 2016)

Summary

Introduction

Our body is remarkably complex, yet we display a highly stable motor performance. For example, to reach for a coffee cup on a desk, there are an infinite number of patterns of muscle activity involved in extending the arm because multiple muscles span the same joint. Muscle synergies have been proposed as a solution to control redundant systems by coordinating a number of muscles with a smaller number of control modules, which are called muscle synergies (Tresch et al, 1999; Bizzi et al, 2002; d’Avella et al, 2003). This hypothesis is phenomenologically supported by experimental observations that a linear combination of basic patterns of muscle activity successfully reconstructs muscle activity during a wide range of behaviors, including reflex movements (Tresch et al, 1999; Cheung et al, 2005), postural tasks

Methods

Results

Conclusion