Real-time motor brain-computer interface based on electrocorticography in humans.

Disabled patients can be assisted in their communication needs with a BCI: a device that decodes brain activity directly, bypassing the need for muscular activity. The journal Science selected in 2012 BCI as one of ten runner-ups for “Breakthrough of the year”. However, the main thrust of current BCI research is with the decoding of EEG recordings acquired from the subject’s scalp or invasive ones obtained from deep brain implants. Both have their benefits and drawbacks: invasive recordings yield a superior decoding performance, for a minimal training effort, but the required surgery may cause irreversible damage to the tissue or its vasculature and the recording stability is a recurring concern. To overcome these drawbacks but retain the benefits, a less invasive approach based on electrocorticography (EcoG) was recently proposed. Whereas its potential for BCI was already demonstrated, it still needs to be shown how to build a BCI that operates in real-time, completely based on movement imagery of fingers and other small joints, and that accounts for the nature of ECoG signals. This also summarizes the project’s aim. Our hypothesis is centered on the requirement of a realistic visual feedback for closing the loop between the intended and the decoded movement so that both the subject and the algorithm can be aligned to the possibilities and limitations of the BCI set-up. The experiments will be performed on patients with an ECoG implant as part of their medical treatment.