Date of Award
Doctor of Philosophy (PhD)
This dissertation addresses the research for the development of spinal cord-computer interface (SCCI). The main objective of SCCI is to generate voluntary motor control signals for individuals with spinal cord injury (SCI).
In the neuroscience aspect, organization of the fibers in the descending tracts of the dorsolateral funiculus of the cervical spinal cord was investigated in cats. The spinal cord was penetrated with silicon substrate microelectrodes at 400 μm intervals in the medio-lateral direction at the C5/C6 and C6/C7 segmental borders. The stimulus consisted of a 20 ms train of charge-balanced biphasic pulses at 330 Hz. The evoked activities from selected forelimb muscles were acquired into computer. The muscle contractions were usually in the form of short twitches. In both segmental borders, the activation threshold was relatively higher in the middle of the dorsolateral funiculus. The majority of the muscles studied had a dorsal or ventral concentration of the activation points. The distal muscles were mostly activated in the ventro-lateral aspect of the funiculus, while the elbow muscle maps spread to both dorsal and ventral sides. These results show a functional organization in both cervical segments although there is an extensive overlap between the areas dedicated for each forelimb muscle.
In the neural signal processing aspect, the feasibility of increasing the channel separation for a neural interface was investigated using the blind source separation (BSS) technique. Multi-contact spinal cord recordings were assumed to be a linear mixture of independent source signals inside the spinal cord. The results from simulated multi-channel recordings show a perfect channel separation. Further investigation was performed on real spinal cord recordings by eliminating the secondary sources, using the FastICA algorithm. The results suggest that the information rate of a spinal cord interface can be improved by separating the neural recordings into its independent components and selecting the ones with the largest distance between them. Comparison between ICA and PCA reveals that ICA is more suitable for this application.
This study constructs the first step in the development of SCCI. The results demonstrate that SCCI is feasible both in the neuroscience and signal processing perspectives.
Tie, Yanmei, "" (2005). Dissertation. 580.