Two-dimensional movement control using electrocorticographic signals in humans

Abstract

We show here that a brain-computer interface (BCI) using electrocorticographic activity (ECoG) and imagined or overt motor tasks enables humans to control a computer cursor in two dimensions. Over a brief training period of 12-36 min, each of five human subjects acquired substantial control of particular ECoG features recorded from several locations over the same hemisphere, and achieved average success rates of 53-73% in a two-dimensional four-target center-out task in which chance accuracy was 25%. Our results support the expectation that ECoG-based BCIs can combine high performance with technical and clinical practicality, and also indicate promising directions for further research.

Figure 1

ECoG array in situ. A: Exposed brain after craniotomy. B: 8×8 electrode grid on the surface of the brain. C: Lateral x-ray image. The electrode grid and several strips are visible. D: Average brain template and electrode locations co-registered to the x-ray image.

Figure 2

Electrode locations in the five subjects projected onto a standard brain. To facilitate comparison, the electrodes were projected from the left onto the right hemisphere for subjects A, D and E (see asterixes).

Figure 3

Example of an analysis comparing ECoG signals for right hand movement and rest (modified from [36]). A: Color-coded values of r² for all locations and frequencies. B: Average spectra (red for rest, green for right hand movement) (left) and r² as a function of frequency (right) for electrode 15. C: Topographical distribution (black dots indicate electrode locations on the 8×4 grid) for color coded r² values calculated for 20 Hz.

Figure 4

Trial sequence for two-dimensional cursor movement. Initially, the screen was blank (rest period). Then, a target appeared in one of four possible locations on the periphery of the screen (movement preparation). Then, a cursor appeared in the center of the screen and immediately started moving as determined by the subject’s ECoG features (movement period). When the cursor reached the target or the space of one of the three other targets, the screen went blank and then, after a brief rest period, the next trial began.

Figure 5

Learning curves for ECoG control of two-dimensional cursor movement using motor actions or imagery (see text).

Figure 6

Average cursor movement trajectories for the five subjects.

Figure 7

Example topographies of control for subjects D and E, calculated for all locations and for the control signals given by the frequency-band combination used online. These topographies show the color-coded correlation (as r² values) of cortical activity with vertical or horizontal movement, and thus indicate the level of task-related control of different cortical areas. Subject D used actual tongue movements for vertical control and actual hand movements for horizontal control. Subject E used imagined versions of the same actions. The traces below each topography show r² values for the locations used online (indicated by stars). Yellow bars indicate the frequency bands used online. The topographies show activity patterns over locations expected with these motor/imagery tasks (see also [28]), and are similar for actual and imagined tasks. These figures also illustrate that choice of different locations/frequencies could have yielded improved online performance (see text).