Date of Award
Doctor of Philosophy in Biomedical Engineering - (Ph.D.)
Bryan J. Pfister
Eric J. Lang
Based on Center for Disease Control and Prevention report 2016, around 39.5 million people in the United States suffer from motor disabilities. These disabilities are due to traumatic conditions like traumatic brain injury (TBI), neurological diseases such as amyotrophic lateral sclerosis (ALS), or congenital conditions. One of the approaches for restoring the lost motor function is to extract the volitional information from the central nervous system (CNS) and control a mechanical device that can replace the function of a paralyzed limb through systems called Brain-Computer Interfaces (BCI).
One of the major challenges being faced in BCIs and also in general neural recording field is the limitations of the microelectrodes. In this study, as the first aim, a custom-made micro-electrode array (MEA) using carbon fibers is developed. After ex vivo testing, they are implanted into the paramedian lobule (PML) of the rat cerebellum to record the multi-unit activity from its cortex. Following animal termination, tissue samples are examined with histological techniques for the assessment of tissue damage caused by the electrodes.
Another challenge in the BCI field is extracting the control information regarding the intended motor function from the CNS. The way the cerebellar cortex encodes sensorimotor information and contributes to motor coordination has been a topic of discussion for decades. Recent studies have revealed high correlations between Purkinje cell simple spikes and the forelimb kinematics in experimental animals. However, tracking single spike activity in long-term implants with multi-channel electrodes has well-known challenges. Therefore, as the second aim of this study, the correlation of multi-unit neural signals from the paramedian lobule (PML) of the cerebellar cortex to the forelimb muscle activities (EMG) in rats during behavior was investigated. Linear regression is performed to predict the EMG signal envelopes using the cerebellar activity for various time shifts of the data (±10, ±50, ±100, and ±200 ms) to determine if the neural signals are primarily motor or sensory. The highest correlations (~0.6 on average) between neural and EMG envelopes are observed when the EMG signals are either shifted only about ±10 ms or not shifted at all with respect to the neural signals. There were however still correlations above the chance level for larger shifts in time. The results suggest that PML cortex contains both motor and sensory information in relation to the forelimb activity, and also that the extraction of motor information is feasible from multi-unit neural recordings from the cerebellar cortex. Increased prediction success was observed in reaching and retrieval phases compared to grasping phase when predictions were tested on three phases of the behavior separately. When EMG and neural signal envelopes were clustered, they showed patterns of surges of activity in all three phases. The neural signals showed higher activity in the reaching phase. The 300-1000Hz components of neural signals contributed to the predictions more than the other frequency bands. The results of this study supports the feasibility of a BCI based on MUA extracted from the cerebellar cortex using MEAs.
Cetinkaya, Esma, "Sensorimotor content of multi-unit activity in the paramedian lobule of the cerebellum" (2022). Dissertations. 1616.