The New Jersey Institute of Technology's
Electronic Theses & Dissertations Project

Title:	Prediction of forelimb muscle activities and movement phases using corticospinal signals in the rat
Author:	Gok, Sinan
View Online:	njit-etd2018-023 (xvi, 82 pages ~ 4.9 MB pdf)
Department:	Department of Biomedical Engineering
Degree:	Doctor of Philosophy
Program:	Biomedical Engineering
Document Type:	Dissertation
Advisory Committee:	Sahin, Mesut (Committee chair) Adamovich, Sergei (Committee member) Foulds, Richard A. (Committee member) Elkabes, Stella (Committee member) Ihlefeld, Antje (Committee member)
Date:	2018-05
Keywords:	Brain computer interfaces Corticospinal tract Micro-electrode array Neural interface Rat Supervised learning
Availability:	Unrestricted
Abstract:	The targeted population for this project is primarily patients with high level spinal cord injury (SCI) and individuals with motor neuron diseases (MND). In both SCI and MND cases motor control is interrupted due to lack of communication between the brain and the musculature, although both sides are otherwise functional. The approach in this project is to use neural engineering techniques to restore the motor function that was lost because of an injury or disease. Brain-computer interfaces (BCIs) attempt to extract the volitional signals from the cortex when the brain's normal outputs to the musculoskeletal system are impaired. However, BCIs that depend on the cortical activities suffer from two main impediments that are intrinsic to the BCI approach itself; firstly, under-sampling of the volitional information due to limited number of recording channels, and secondly, the long-term instability of the neuronal firings that make it difficult to track movement parameters, such as hand kinematics. As an alternative approach, a spinal cord computer interface (SCCI) can address both obstacles by providing means to access neural signals from a relatively smaller yet denser implant area in order to extract low-level movement parameters, such as muscle electromyography (EMG) signals, for prolonged signal stability. Since the descending fibers of the spinal cord influence the lower motor neurons that directly innervate the skeletal muscles, decoding the information in these fibers can provide a way to establish a robust relationship between the neural control signals and the output parameter, that is the EMG signal. The axons carrying the cortical information through the spinal cord are tightly bundled together in the descending tracts that eventually synapse with the inter-neurons and alpha motor neurons located in the spinal grey matter. The corticospinal tract (CST) is one of the descending tracts that carry the forelimb volitional information. In this study, the CST signals are recorded in rats that are implanted with custom-designed flexible multi-electrode arrays (MEAs). The power spectral density of the CST signals during the movement is notably higher than those observed during resting and anesthesia. The average inter-channel coherences up to 1.5 kHz are significantly higher for reach-to-pull task compared to face grooming and resting states, suggesting the presence of volitional information in the recorded CST signals. The results show that the CST signals can be segregated into two or three different classes using the forelimb movement components as guidance criteria with 97% and 71% accuracies, respectively. Predictions with correlation coefficients as high as 0.81 for the biceps EMG are achieved in individual sessions, although the average prediction accuracies vary considerably among rats. These results support the feasibility of an EMG-based Spinal Cord Computer Interface for patients with high level of paralysis.