Cortical surface recurrent brain-computer interfaces

Abstract

The output of a "traditional" brain computer-interface (BCI) is the operation of an effector mechanism, like a cursor or a prosthetic arm. In contrast, the output of a recurrent brain-computer interface (rBCI) is electrical stimulation delivered directly into the central nervous system (CNS). Recurrent BCIs have been used to artificially bridge two separate sites in the CNS whose communication may have been interrupted. They have also been used to associate activity of a site in the CNS with stimulation of another site, to produce synaptic plasticity between the two sites. To date, rBCIs have utilized intracortical implants to record neural activity and deliver electrical stimuli, which have problems that limit their clinical applicability. These limitations can be addressed by cortical surface electrodes, subdural or epidural, that can capture electrocorticography (ECoG) signals and deliver electrical cortical surface stimulation. We first examine the recording capabilities of cortical surface arrays. We study the relationship of ECoG signals with motor behavior and EMG activity from upper extremity muscles. We demonstrate that EMG activity can be decoded from multichannel ECoG, and document the gradual decrement in decoding performance over several months of recording. Second, we examine the stimulation capabilities of cortical surface arrays. We characterize the effects of repetitive stimulation on the electrode-tissue interface, by measuring electrode impedance. We determine the impact of stimulation on cortical excitability, by measuring stimulus-evoked motor responses. Finally, we examine the effect of stimulation on spontaneous cortical activity, as evidenced by ECoG power at different frequencies. Third, we investigate a cortical surface rBCI system to study the role of sensorimotor beta oscillations in synaptic plasticity. Stimulation at a cortical site was triggered from specific phases of beta (15-25 Hz) oscillatory episodes of ECoG recorded from a different site. The effects of conditioning stimulation on cortical connectivity were determined through cortically-evoked potentials and ECoG phase coherence. We document a short-term change in cortical connectivity following beta oscillations that is mediated by synaptic modification and follows a Hebbian-like rule. Our findings on properties of cortical surface recording and stimulation will promote translation of these techniques to clinical applications. Our demonstration of changes in cortical connectivity induced by a cortical surface rBCI furthers our understanding of cortical oscillations and provides a paradigm for activity-dependent cortical plasticity using these less invasive implants.