Model-based design of microstimulation encoders for somatosensory prostheses.

Abstract

There has been much recent interest in somatosensory neural prostheses for restoring the sense of touch and proprioception. While some psychophysical experiments in rats and primates have elucidated the range of perceptual sensitivities to certain stimulus parameters, not much work has been done for developing encoding models for translating mechanical sensor readings to microstimulation through an array of electrodes implanted in the somatosensory pathway. I pro- pose a method of computationally designing these encoders using the tools of model-based con- trol, a methodology that has been applied fruitfully to robotic and chemical process control. By modeling the dynamical effects that microstimulation has on a neural population, optimal stimu- lation patterns can be efficiently obtained in order to mimic naturalistic neuronal activation. Spe- cific modeling challenges are to design nonlinear multi-input multi-output (MIMO) black-box models to express this transformation, and to prevent proliferation of parameters with increasing numbers of input (stimulating electrodes), and outputs (neural readout channels). Once a model is obtained, efficient numerical optimal control techniques optimize multi-channel stimulation control-law, or policy, that respect constraints imposed by safety limits and the operating regime of our models. In this work I apply various low-dimensional linear and nonlinear black-box models to describing the functional input-output dynamics of the thalamocortical system. I then show in silico and in vivo the use optimal control for generating complex spatiotemporal pat- terns of stimulation that drive cortical field potentials through a desired neural trajectory. The results reveal that this model-based control method can faithfully reproduce responses from a wide array of natural touches and accurately convey naturalistic levels of information. This work demonstrates the possibility of sensory prostheses that not only speak the language of the brain, but can drive the brain to produce naturalistic percepts.