Proceedings of The Physiological Society
University of Oxford (2011) Proc Physiol Soc 23, PC91
Optimal control of task-specific muscle synergies II: Feedback strategies
K. Nazarpour1, A. Barnard1, A. Jackson1
1. Newcastle University, Newcastle, United Kingdom.
A companion abstract  reported the development of optimal correlation structure in the trial-to-trial variability of muscle activity according to the abstract demands of a myoelectric-controlled interface task . One mechanism by which this can emerge is through modulation of the common drive to motorneurons via divergent corticospinal projections. However, this simple feed-forward model, while accounting for behaviour in the absence of visual feedback, fails to describe several features of the data when visual feedback is present. In particular, the observed negative correlation between muscles implies the action of a feedback mechanism acting predominantly along task-relevant dimensions. In this experiment we sought direct evidence of such a ‘minimum intervention controller’  driven by either visual or proprioceptive feedback. 20 human subjects made repeated movements of a myoelectric cursor controlled by smoothed, rectified electromyogram (EMG) from APB and ADM muscles. Elliptical target shapes imposed task-relevant and irrelevant dimensions within the task space, oriented such that movement variability should optimally be constrained along dimensions of either positive or negative EMG covariance. In half of the trials, a perturbation was delivered 300ms into the 1s hold period. Perturbations were either visual (cursor position was translated along the axis of one of the controlling muscles) or proprioceptive (electrical stimulation of median nerve at the wrist or cutaneous nerve of the little finger in the absence of visual feedback). The existence of a minimum intervention controller acting along the task-relevant dimension predicts that reflex responses should be observed in the unperturbed muscle. Furthermore, these responses should change direction according to the orientation of the target ellipse. This behaviour was observed in the case of visual perturbations although the response in the unperturbed muscle was smaller and delayed relative to the response in the perturbed muscle. The amplitude of the target-dependent reflex component increased over the course of training (paired t-test, df=19, t=1.96, p=0.06), in parallel with improved performance scores (paired t-test, df=19, t=5.22, p<0.001). Neither short nor long latency EMG responses in APB and ADM to peripheral nerve stimulation showed task-specific modulation. We conclude that optimal control of a myoelectric-controlled interface incorporates both feed-forward (modulation of common drive) and feed-back (minimum intervention controller) strategies. However, when visual feedback is withheld, control is achieved predominantly through feed-forward mechanisms with minimal influence of proprioceptive information.
Where applicable, experiments conform with Society ethical requirements