Proprioceptive Feedback Modulates Motor Cortical Tuning During Brain-machine Interface Control
Danielle Rager, Carnegie Mellon University
Loss of proprioception is known to severely impair motor control, but the neural mechanisms by which proprioception aids in the planning and execution of visually guided movements are not well understood. We investigate the impact of providing proprioceptive feedback to a human subject with tetraplegia and intact sensation who was implanted with two 96-channel microelectrode arrays in primary motor cortex (M1). Passive proprioceptive feedback was provided either by manually moving the subject’s arm in conjunction with the brain-machine interface (BMI)-controlled robotic arm or by moving the subject’s arm in an exoskeleton under BMI control. Performance of a BMI-assisted reaching task degrades when we allow a visually trained decoder to leverage the subject’s own proprioceptive signals, indicating that proprioceptive feedback alters M1 tuning structure. We show that velocity tuning is stable across trials in the visual feedback only (V) condition and in the visual and proprioceptive feedback (VP) condition, but is not stable across the two conditions. Paradoxically, reaching task performance was worse when both training and task control of the BMI were performed with vision and proprioception than with vision alone. We show that proprioceptive feedback introduces "noise" correlations that increase the overlap of M1 movement representations when motor tuning is taken to be a function of velocity coordinates. These "noise" correlations make movements less differentiable to our decoder in the VP condition. Our findings suggest that M1 does not encode movements with invariant tuning in velocity space, but rather with a latent tuning structure that is dependent on the sensory feedback modalities available. Therefore, effective closed-loop BMI control with natural or surrogate somatosensory feedback may require the development of new decoding algorithms that do more than map M1 responses to kinematics.
Abstract Author(s): D.M. Rager, J.E. Downey, J.L. Collinger, D.J. Weber, M.L. Boninger, R.A. Gaunt,
V. Ventura