Goal selection as a control strategy in a brain-computer interface

Abstract

A brain-computer interface (BCI) translates signals recorded directly from the brain into commands that control an external device, such as a computer cursor, wheelchair, or neuroprosthetic. BCIs promise to help the nearly 6 million people who live with paralysis by allowing them to interact with the world in ways they are no longer able. BCIs can also be used by able bodied individuals to extend their capabilities. BCIs differ widely in how they implement the translation from raw brain signal to device command. Two competing control strategies, goal selection and process control, differ in how much the BCI assists the user. In process control, the user controls every step of the process and receives minimal to no assistance from the system. Other terms for process control include low-level control or continuous control. In goal selection, the user only needs to determine the goal and the system executes the process to achieve that goal. Other terms for goal selection include high-level control or shared control. This thesis presents the first studies directly comparing goal selection and process control. We found in these studies that the goal selection based paradigms were easier to learn, had a decreased training period, and provided improved speed, accuracy, and information transfer in both the simple and more complex applications studied. This thesis also extends our understanding of the neurophysiology while using a sensorimotor rhythm based BCI. When individual trial data were analyzed and not averaged as is typically done in the literature, we found that duration of sensorimotor rhythm modulation was more correlated to successful use than amplitude of modulation. Additionally, we found that correct modulation that led to either a single hit or overall high accuracy was the same between the two control strategies. This shows that the improved performance in these studies while using the goal selection based paradigms was more attributable to the difference in device command instead of the difference in raw brain signal. By understanding neurophysiology and applying that knowledge to BCI design, we can make a better BCI.