Looking for cortical patterns of successful motor imagery-based BCI learning

Marie-Constance Corsi (1) , Mario Chavez (2) , Denis P Schwartz (2) , Nathalie George (2) , Laurent Hugueville (2) , Ari E. Kahn (3) , Sophie Dupont (2) , Danielle S. Bassett (3) , Fabrizio de Vico Fallani (1)

1 ARAMIS - Algorithms, models and methods for images and signals of the human brain = Algorithmes, modèles et méthodes pour les images et les signaux du cerveau humain [ICM Paris]
2 ICM - Institut du Cerveau = Paris Brain Institute
3 University of Pennsylvania

Non-invasive Brain-Computer Interfaces (BCIs) based on motor imagery (MI) tasks represent a valuable tool both from a societal and a clinical perspective. Nevertheless, performances vary inconsistently across subjects and the mechanisms underlying a successful skill acquisition is poorly understood. In this longitudinal study performed with the electroencephalography (EEG), we show that BCI training can be characterized by patterns that rely on the neurophysiology. We observed that the desynchronization effect increased significantly over the sessions within the alpha and beta sub-bands for subjects who showed a significant improvement of their BCI scores. Notably, we observed that they also presented a decrease of the functional connectivity in regions beyond those targeted during the BCI experiments, whereas the subjects who did not improve their performances did not show any significant change over sessions. Taken together, these results give additional insights about the skill acquisition process during MI-based BCI trainings.

Domaines: Traitement du signal et de l'image [eess.SP]; Neurosciences

Marie-Constance Corsi, Mario Chavez, Denis P Schwartz, Nathalie George, Laurent Hugueville, et al.. Looking for cortical patterns of successful motor imagery-based BCI learning. 8th Graz Brain-Computer Interface Conference 2019, Sep 2019, Graz, Austria. ⟨hal-02123582⟩