Analysis of Graded Sensorimotor Rhythms for Brain-Computer Interface Applications

The emerging field of neural engineering is tasked with applying engineering principles towards understanding neuroscience. A by-product of such a venture has been the development of a class of assistive devices known as brain-computer interfaces (BCIs) which link brain activity to actions performed by external devices. One application of this technology is in the rehabilitative sector for individuals with neuromuscular diseases and disorders. Despite tremendous efforts in the last few decades, a reliable signal that reflects fine motor control has yet to be adequately investigated. This gap in knowledge has limited the potential of BCIs to restore movement and communication. To directly address this, I investigated the activation of the sensorimotor cortex surrounding execution of graded movement. Distinguishing between different degrees of effort provides the possibility for more command signals to control a BCI. Instead of discriminating between users attempting or visualizing different simple movements, such as left-hand reach versus right-hand reach, information regarding the extent or intensity of movement would be available for use. As a result, more complex output commands could be interpreted and utilized by the BCI. This allows for feedback to be delivered to the user in real-time to close the loop between intention and action. The primary aim of this dissertation is to investigate whether graded effort associated with a movement task can be distinguished from scalp EEG recordings. This endeavor is motivated by the need for numerous, natural, and intuitive command signals in brain-controlled assistive devices, such as BCIs, to achieve fine control. Measurable changes in brain activity during a graded cognitive task can serve as commands that bridge the gap between intent and fine control and thereby play a vital role in therapeutic protocols aimed at recovery of function. A novel protocol was first established that guides subjects through a graded movement task while providing immediate visual feedback. First in a healthy cohort and then in individuals with chronic stroke, I tested the ability of the protocol to elicit and model levels of effort while the subject attempts the graded movement task. After a wide search into features of the electroencephalogram (EEG) that may highlight the different levels of effort, I arrived at an 8 – 30 Hz rhythm that was differentially attenuated according to the movement. Applying common modeling techniques led to the ability to predict these gradations from EEG with greater than chance accuracy. A forward-looking study concludes this dissertation in which the bridge between mental intent and feedback is closed with real-time prediction of effort from the EEG. This work suggests a model that predicts graded movement in healthy and clinical populations is feasible. The ability to connect intent and execution in those who may have lost function can play a vital role in therapeutic programs aimed at recovery of function and independence