Guemann M.1, Lapeyre E.2, Ricard D.2, Paclet F.1, Cattaert D.1, de Rugy A.1
1University of Bordeaux, Gironde, Bordeaux, France, 2Hôpital du Val de Grace, Paris, France
Background: Today's myoelectric prostheses offer a large range of mechanical movements. However, the interface between the human and the machine is still linear, and does not integrate the important regulatory functions of spinal sensorimotor loops. The SLR is a spinal sensorimotor network whose dynamics is able to control a movement.
Purpose: In this study, we've tested the capacity of this spinal network model (SLR) to produce a simple gesture that consists in a flexion of the forearm on the arm with a pre-set range of motion and speed. In this context, the complex time course of sensorimotor commands needed to realize the movement had to be totally produced by the spinal network dynamics.
Methods: A 3D elbow model was developed with two antagonist muscles driven by the network. The network stimulation was done in two parts. The first part (SET) regulates the network excitability and the synapse properties, in order to put the network in a functional state ready to produce a movement, but without triggering it. The second phase (GO) triggers the movement by applying a stimulation to the network, with the exception of α and γ motoneurons, whose activations are managed by the dynamics of the network.
The model has been tested in presence of gravity (1) and perturbation (2). The mean square error (MSE) between the target kinematics and the obtained kinematics were used to quantify movement quality. This MSE was optimized by simple trial and error search on the level of current stimulation to each neuron at the SET and GO states, as well as on the values of pre synaptic threshold and synaptic conductance.
Results: For a target flexion of the elbow of 40° executed in 300ms, the MSE obtain after optimization was 5deg2 for condition 1 and 42,62deg2 for the condition 2. Those results produce a kinematic that is reasonably close to a desired one.
Conclusion(s): These results indicate that an arbitrary elbow flexion can be generated and stabilized without direct commands to α and γ motoneurons in the presence of gravity, and with perturbation. This work illustrates the important regulatory function of spinal networks, which could prevent the brain to keep a memory of all possible patterns and regulation that would be necessary to manage any unpredictable perturbation. Instead, this spinal regulatory function offers the opportunity to treat perturbations locally, in a close and fast loop manner.
Implications: Biomimetic network regulation of this kind should be able to substantially improve control systems for prosthetic limbs.
Funding acknowledgements: This project has been selected and awarded by a PhD grant by the French General Army Direction (DGA).
Topic: Disability & rehabilitation
Ethics approval: No ethics approval was needed for this research.
All authors, affiliations and abstracts have been published as submitted.
