The Effects of Intracortical Microstimulation Parameters on Neural Responses

Microstimulation of brain tissue plays a key role in a variety of sensory prosthetics, clinical therapies and research applications. At present, stimulus parameters are tailored to each application via iterative testing. Computational optimization methods seek to improve tried and tested waveforms developed for specific purposes, however the fundamental understanding of how stimulation parameters influence the neural circuits they activate remains widely unknown. This deficit hinders both the optimization of existing protocols and the development of new stimulation applications. To date, a number of prosthetic devices validated as early as the 1970's linger in the development stages largely due to the inability to effectively interface these devices with the brain. In order to use electrical stimulation to convey messages to the central nervous system, a better understanding of stimulus signal design is required. In this thesis, I investigate the influence that each parameter of the constant-current, symmetric, biphasic square waveform exerts on the responses it evokes through intracortical microstimulation of the caudal forelimb area of the rat motor cortex. The parameters under investigation include the current amplitude, pulse frequency, pulse duration, interphase interval and train duration of the stimulus and effects were assessed by examining the electromyographic responses evoked in the rat forelimb muscles in response to each stimulus. The major findings describe how each parameter of the stimulus signal influences the magnitude, onset latency, and duration of the response. A previously unexplored component of the response signal (which we called ‘residual activation') is analyzed for the first time. Hypotheses as to the origin and underlying neural mechanism of this phenomenon are proposed and the stimulus parameters affecting its occurrence, prevalence and duration are described. The reliability of stimulation signals for evoking consistent responses is also assessed with respect to parameter variations. A methodology for the informed design of stimulation signals is proposed and aided by the development of a simple computational model representing the input-output relationships between stimulation parameters and the responses they evoke. This model uses an artificial neural network approach and can be used to predict the properties of the response when the parameters of the stimulus are known. Given the prevalence of brain stimulation in clinical, research and therapeutic applications the proposed methodological and modeling procedures have important implications in the optimization of current stimulation paradigms and the development of stimulation protocols for new applications.