The Cybathlon BCI race: Successful longitudinal mutual learning with two tetraplegic users.

This work aims at corroborating the importance and efficacy of mutual learning in motor imagery (MI) brain-computer interface (BCI) by leveraging the insights obtained through our participation in the BCI race of the Cybathlon event. We hypothesized that, contrary to the popular trend of focusing mostly on the machine learning aspects of MI BCI training, a comprehensive mutual learning methodology that reinstates the three learning pillars (at the machine, subject, and application level) as equally significant could lead to a BCI-user symbiotic system able to succeed in real-world scenarios such as the Cybathlon event. Two severely impaired participants with chronic spinal cord injury (SCI), were trained following our mutual learning approach to control their avatar in a virtual BCI race game. The competition outcomes substantiate the effectiveness of this type of training. Most importantly, the present study is one among very few to provide multifaceted evidence on the efficacy of subject learning during BCI training. Learning correlates could be derived at all levels of the interface-application, BCI output, and electroencephalography (EEG) neuroimaging-with two end-users, sufficiently longitudinal evaluation, and, importantly, under real-world and even adverse conditions.

Long-Term Stable Control of Motor-Imagery BCI by a Locked-In User Through Adaptive Assistance.

Performance variation is one of the main challenges that BCIs are confronted with, when being used over extended periods of time. Shared control techniques could partially cope with such a problem. In this paper, we propose a taxonomy of shared control approaches used for BCIs and we review some of the recent studies at the light of these approaches. We posit that the level of assistance provided to the BCI user should be adjusted in real time in order to enhance BCI reliability over time. This approach has not been extensively studied in the recent literature on BCIs. In addition, we investigate the effectiveness of providing online adaptive assistance in a motor-imagery BCI for a tetraplegic end-user with an incomplete locked-in syndrome in a longitudinal study lasting 11 months. First, we report a reliable estimation of the BCI performance (in terms of command delivery time) using only a window of 1 s in the beginning of trials (AUC ≈ 0.8 ). Second, we demonstrate how adaptive shared control can exploit the output of the performance estimator to adjust online the level of assistance in a BCI game by regulating its speed. In particular, online adaptive assistance was superior to a fixed condition in terms of success rate ( ). Remarkably, the results exhibited a stable performance over severalmonths without recalibration of the BCI classifier or the performance estimator.