Section: New Results

Brain Computer Interfaces

Combining ERD and ERS features to create a system-paced BCI

Participants : Maureen Clerc, Théodore Papadopoulo, Joan Fruitet, Eoin Thomas.

An important factor in the usability of a brain computer interface (BCI) is the setup and calibration time required for the interface to function accurately. Recently, brain-switches based on the rebound following motor imagery of a single limb effector have been investigated as basic BCIs due to their good performance with limited electrodes, and brief training session requirements. Here, a BCI is proposed which expands the methodology of brain-switches to design an interface composed of multiple brain-buttons. The algorithm is designed as a system paced interface which can recognise 2 intentional-control tasks and a no-control state based on the activity during and following motor imagery in only 3 electroencephalogram channels. An online experiment was performed over 6 subjects to validate the algorithm, and the results show that a working BCI can be trained from a single calibration session and that the post motor imagery features are both informative and robust over multiple sessions.

This work, which was partially presented at the EMBS conference [33] , is currently under revision for the Journal of Neuroscience Methods.

Bandit algorithms for faster task selection in BCI

Participants : Maureen Clerc, Joan Fruitet, Alexandra Carpentier [Sequel Project-Team, Inria Lille, France] , Rémi Munos [Sequel Project-Team, Inria Lille, France] .

BCIs based on sensorimotor rhythms use a variety of motor tasks, such as imagining moving the right or left hand, the feet or the tongue. Finding the tasks that yield best performance, specifically to each user, is a time consuming preliminary phase to a BCI experiment. This study presents a new adaptive procedure to automatically select, online, the most promising motor task for an asynchronous brain-controlled button.

We develop for this purpose an adaptive Upper Confidence Bound algorithm based on the stochastic bandit theory, and design an EEG experiment to test our method. We compare (offline) the adaptive algorithm to a naive selection strategy which uses uniformly distributed samples from each task. We also run the adaptive algorithm online to fully validate the approach.

By not wasting time on inefficient tasks, and focusing on the most promising ones, this algorithm results in a faster task selection and a more efficient use of the BCI training session. More precisely, the offline analysis reveals that the use of this algorithm can reduce the time needed to select the most appropriate task by almost half without loss in precision, or alternatively, allow to investigate twice the number of tasks within a similar time span. Online tests confirm that the method leads to an optimal task selection.

This study is the first one to optimize the task selection phase by an adaptive procedure. By increasing the number of tasks that can be tested in a given time span, the proposed method could contribute to reducing “BCI illiteracy”.

This work is the result of the collaboration between Sequel and Athena within the ANR CoAdapt. It has been published in NIPS [21] and is accepted in the Journal of Neural Engineering  [52] .

An analysis of performance evaluation for motor-imagery based BCI

Participants : Maureen Clerc, Matthew Dyson [Laboratoire de Neurosciences Cognitives, Aix-Marseille Université, France] , Eoin Thomas.

In recent years, numerous brain computer interfaces (BCIs) have been proposed which incorporate features such as adaptive classification, error detection and correction, fusion with auxiliary signals and shared control capabilities. Due to the added complexity of such algorithms, the evaluation strategy and metrics used for analysis must be carefully chosen to accurately represent the performance of the BCI. In this article, metrics are reviewed and contrasted using both simulated examples and experimental data. Furthermore, a review of the recent literature is presented to determine how BCIs are evaluated, in particular focusing on the correlation between how the data are used relative to the BCI subcomponent under investigation. From the analysis performed in this study, valuable guidelines are presented regarding the choice of metrics and evaluation strategy dependent upon any chosen BCI paradigm.

This work was supported by the ANR Co-Adapt and is currently under revision for the Journal of Neural Engineering.