Computer cursor control by motor cortical signals in humans with tetraplegia

A direct neural interface system (NIS) promises to provide communication and independence to persons with paralysis by harnessing intact motor cortical signals to enable controlling prosthetic devices. An intracortical NIS aims to achieve this by sensing extracellular neuronal signals through chronically implanted microelectrodes and by decoding the spiking activity of neurons into prosthetic control signals. In non-human primate studies, decoding has been performed by finding a relationship between neuronal signals and actual limb movements. However, such decoding approaches face challenges in the case of paralyzed persons where there is no true movement information. Specifically, we have focused on dealing with several key questions in decoding of neural activity in humans with paralysis: what movement parameters should be decoded?; which decoding algorithms lead to more accurate estimation of movement parameters?; how do we train decoding algorithms without observing actual movement parameters?; and how many control parameters can be decoded from a single neural ensemble? In this paper, we summarize our recent studies to address these questions to improve decoding performance, which enables a human with tetraplegia to drive a 2D computer cursor to an arbitrary position and execute a ldquoclickrdquo on the area of interest.