| Brain-computer interface (BCI) technologies have traditionally been designed under the assumption that BCI users are patients who no longer exhibit any significant motor control abilities, due in part to deficits resulting from neuromuscular disease. Of particular interest is the activity that occurs in motor related areas of the brain during precise, dexterous movements of the hand. Previous studies of the central and peripheral motor nervous systems suggest the presence of synergistic activations of musculoskeletal groups during coordinated movement, however the existing clinical technology has not been sufficient for accurate exploration. Among the many features of the cortical response measured by electrocorticography (ECoG), we find that high gamma (75-200Hz) activity in primary motor cortex shows high spatial preference for individual digit movements during overt finger flexions. In contrast, the average spatial activity during object grasping appears to show little unique spatial organization relative to the grasps performed. To investigate this discrepancy, we explore the difference of spatial distribution of activity during dexterous hand movement between traditional (10mm-spacing) and high resolution (3mm) ECoG grids. We find that the informational density present at the higher scale underscores the potential of synergistic dimensionality reduction as a possible model of prosthetic control. Complicating the gains from resolution, however, is the remnant hand motor function present in a large population of patients interacting with BCIs. This presents a challenge for real-world implementation of these systems for therapy and rehabilitation. Thus, an important question that had yet to be systematically studied within the BCI control architecture is: can subjects use BCIs simultaneously coordinated with overt motor activity? We explored this question by testing if patients could effectively learn to perform and coordinate bimanual control. Patients implanted with ECoG electrodes were trained to interact with a 2-D BCI center-out cursor task in which one dimension of control was modulated by neural activation related to motor imagery and the other dimension was controlled by overt natural movement. Over the course of multiple sessions and days, the subjects gained levels of proficiency and demonstrated accurate control. In addition, this system was implemented for a test case subject with lifelong hemiparesis and significantly impaired bilateral motor abilities brought on by a perinatal stroke. Especially in the stroke case, the subject demonstrated significant proficiency and accurate control, despite confounding ipsilateral and contralateral neural activations. As control improved, activation patterns changed across related cortical areas but without any significant observable changes in the overt motor behavior. In gestalt, this work provides steps and a framework towards adapting ECoG-based BCIs for patients with similar cortical deficits for the purpose of dexterous coordinated hand control. |
