INCORPORATION OF NATURAL SENSORY FEEDBACK TO BRAIN MACHINE INTERFACE WITH HAND EXOSKELETON

The feasibility of a brain-machine interface (BMI) system that links the brain to an external device has been demonstrated for both non-human and human subjects. However, BMI controlled robotic arm movements are usually slow, jerky and imprecise. Many of these problems can be attributed to a lack of somatosensory (tactile, proprioception, etc.) feedback. For a large segment of potential users who have motor impairments but intact sensation, use of an exoskeleton as the external device could provide the natural sensory feedback to improve BMI control. This is especially true for the hand with its incredibly rich sensory innervation. Currently, however, there is no hand exoskeleton available for non-human primate BMI research. In this dissertation, a hand exoskeleton platform was developed for an index finger - thumb precision grip task. The system was first employed as a scientific apparatus to explore the sensory responses in primary motor cortex (M1) to sinusoidal inputs of position and force. Neural firing rate patterns were found to be strongly entrained to the sinusoidal stimulus, with a predominance of neurons responding to joint movement rather than fingertip force. The phase-locking patterns to sinusoidal stimuli were much clearer and more stable for joint movement than for fingertip force as well. Second, the hand exoskeleton system was validated in a real-time BMI-controlled isometric grip force task. Prompted by cues on a computer screen, the monkey used cortical signals to control the grip force produced by the exoskeleton. The exoskeleton drove either the monkey’s own hand (natural sensory feedback condition) or an artificial hand (visual feedback only condition). Although improvements in performance for both conditions were observed over the relatively short training period, it was difficult to differentiate between the efficacy of the two conditions. Interestingly, for both conditions the monkey tried to use a similar neural strategy in controlling grip force as to the one used in natural grip behavior, in which a majority of the neurons recorded exhibited temporal reduction of the firing rates during the force production phase. Overall, the hand exoskeleton platform proved to be not only a powerful platform in BMI research but also an important tool for investigating sensory processing. This new platform should facilitate future experiments which will further insight into BMI design and the neural mechanisms underlying movement control.Ph.D. in Biomedical Engineering, May 201