Decoding responses to periodic visual motion stimulation using multiway regression for self-paced brain computer interfacing

Brain-Computer Interfaces (BCIs) decode neural activity with the aim to establish a communication channel. As traditional neural pathways are bypassed, they have been hailed as a solution for patients with impaired muscular control. BCIs based on brain implants yield superior decoding performance but they require surgery and the observed loss in signal quality is a recurring concern. Non-invasive BCIs, primarily EEG-based ones, curb these drawbacks but at the expense of being slow compared to Assistive Technology such as eye-tracking. The most performant EEG-BCIs rely on responses recorded in sync with the onset of a displayed stimulus (“cue”) to which the subject pays attention. In contrast, asynchronous or self-paced BCIs operate independently of a cue and some even can detect when the user wants to use the BCI. Self-paced operation is a decisive step in moving BCIs out of the lab into the real world, but demonstrations are still rare and there have been no reports on patients. Another development is the quest for less taxing visual stimulation paradigms. Both observations also mark our scientific objective: to develop a new visual stimulation paradigm, based on periodic motion-onset stimuli instead of flickering stimuli, and a multiway regression-based decoder that accounts for self-paced operation in real-time. Our technical objective is to test our self-paced BCI on healthy subjects as well as on patients and to integrate it into a commercial bedside terminal.