Jason White

Biomedical Engineering

PhD Defense

Date: January 8th, 2016 (Friday)

Time: 2:30 PM

Location: Georgia Tech, IBB Building (Petit Institute), Room 1128

Co-Advisors:

Steve DeWeerth, PhD, Department of Biomedical Engineering, Georgia Tech

Keith Tansey PhD, MD, Department of Neurology, Emory University

Committee Members:

Lena Ting, PhD, Department of Biomedical Engineering, Georgia Tech

T. Richard Nichols, PhD, Department of Physiology, Georgia Tech

Boris Prilutsky, PhD, Department of Physiology, Georgia Tech

Magnus Egerstedt, School of Electrical and Computer Engineering, Georgia Tech

OPTIMIZING SENSORY STIMULATION TO ASSIST HUMAN WALKING AFTER SPINAL CORD INJURY

Sensory stimulation has shown promise in improving human walking after spinal cord injury (SCI). Previous studies have demonstrated some improvement with open-loop, non-individualized sensory stimulation, but after SCI, there are many unique, individual changes in sensorimotor processing. These changes make a priori identification of the best sensory stimulation pattern difficult for any given individual. Real-time optimization provides a solution to this individuality problem, through optimizing sensory stimulation parameters for a given subject in on-line (in real-time). In this research, I developed an approach to optimize sensory stimulation to maximally assist human walking after incomplete SCI. To do so, I had to develop and validate a novel optimization algorithm for globally-optimizing noisy, time-variant, black-box systems, while maximizing the information gained from each test (experiment). I optimized sensory stimulation across a range of SCI subjects, across multiple sensory stimulation sites, and with different stimulation parameterizations. In all subjects and stimulation sites, the optimal stimulation protocol produced better walking (i.e. less external force assistance was required) than three alternative stimulation protocols: an industry-standard stimulation protocol, a no-stimulation protocol, and a random-stimulation protocol. The optimization approach minimized the total force required from an assistive orthosis, and post-hoc analysis of the optimization sessions produced a better understanding of how stimulation parameters affected specific gait features (e.g. hip forces during swing). Transcutaneous spinal cord stimulation (TSCS) frequency had divergent effects on the stance and swing phases – high frequencies tended to assist with swing, but low frequencies tended to assist with stance. For the two peripheral nerve stimulation sites (posterior tibial and common peroneal nerves), the optimal gait-phase for stimulation was generally after mid-stance and before early swing. There was some variability within this time-range depending on the specific feature under study. Experimental history (i.e. time spent walking/time spent being stimulated) proved to be as important a predictor as any of the stimulation parameters.