Localizing and Characterizing Cognitive Activity from Local Field Potentials in the Primate Prefrontal Cortex

Abstract

For most of us, the seemingly mundane act of grabbing a glass, filling it with water, and quenching thirst is a task performed effortlessly. However, for individuals facing physical disabilities, this simple act can be impossible without the help of others. Despite remarkable technological advancements, such as robots using artificial intelligence to perform complex tasks, the potential to restore lost physical functions through brain-computer interface (BCI) systems remains an area demanding exploration. Recent intracortical BCI investigations have predominantly examined the generation of movement from spiking activity within the premotor and primary motor cortices, coupled with the stimulation of associated somatosensory areas. This approach holds the potential to facilitate functional electrical stimulation of muscles, empowering BCI users to command their limbs, therefore restoring lost connections. In contrast, this thesis investigates a different approach: identifying the goal of the BCI user directly from the prefrontal cortex. This intention could be used to control assistive devices, such as a robot arm, allowing them to choose the most suitable actions to complete the goal. Decoding cognitive functions from features based on local field potentials (LFPs), as opposed to single-unit spiking rate, will be explored and whenever possible utilized as these signals are considered more stable over time. The success of a goal based BCI system depends on the quality and stability of the recorded signals from one or more implants. The recorded neural activity from small intracortical implants is contingent on the brain area they covered. The process of localizing and characterizing cognitive activity associated with overarching goals before implantation holds the potential to assist in determining the most suitable location for a BCI implant. This research first validates that LFPs from the prefrontal cortex can be used to decode important factors related to goals, such as saccade target locations and spatial positions. Furthermore, it introduces a methodology for localizing and characterizing cognitive activity, offering valuable insights for strategically planning optimal implant locations.