Epidural electrocorticography and intra-cortical local field potentials in motor cortex during volitional arm movements and their applications to neural prosthetic control

This thesis studied two novel control signals for a brain-computer interface (BCI): the high frequency components of the intracortical local field potential (LFP) and epidural electrocorticography (ECoG), both in the motor cortices of non-human primates. The intracortical LFP, which is the sum of all electrical activity recorded on a penetrating microelectrode, was directionally tuned in the high frequencies (60-200 Hz) of its power spectrum. The tuning was present in a center-out reaching task and remained in a separate circle-drawing task. Furthermore, this high frequency tuning was well correlated to nearby simultaneously recorded single units. The epidural ECoG signals did not exhibit significant directional tuning in any frequency band, but did significantly increase in the high frequencies during movement. A real-time closed-loop BCI was developed and implemented using the epidural ECoG signals. Since the epidural ECoG signals were not directionally tuned, control was achieved entirely from neural adaptation due to biofeedback. Two signals, recorded approximately one cm apart, were used to control horizontal and vertical cursor velocity. Although the two signals initially were correlated, the correlation decreased with each day of closed loop training. In only three to ten training sessions, the subjects performed a 2-D center-out "reaching" task a using only epidural ECoG signals. In addition, one subject used the epidural ECoG signals to perform a circle "drawing" task. The results presented in this thesis demonstrate that the high frequency intracortical LFP can approximate local spiking activity and that epidural ECoG provides a minimally invasive control signal for a brain-computer interface.