Alpha Correlates of Practice During Mental Preparation for Motor Imagery

In this study, we quantified performance variations of motor imagery (MI)-based brain-computer interface (BCI) systems induced by practice. Two experimental sessions were recorded from ten healthy subjects while playing a BCI-oriented video game for 2 weeks. The analysis focused on the exploration of electroencephalographic (EEG) changes during mental preparation between novice and practiced subjects. EEG changes were quantified using global field power (GFP), dynamic time warping (TW), and mutual information (MutInf): GFP represents the strength of the electric field, TW measures signal similarities, and MutInf signals interdependency. Each metric was selected to relate insights extracted from mental preparation to the three experimental hypotheses associating practice with BCI performance. Significant results were identified in lower alpha for GFP and upper alpha for TW and MutInf. GFP in lower alpha during mental preparation assessed not only novice versus practiced variations but also "intrasession" differences. Findings suggest that EEG changes during mental preparation provide a quantitative measure of practice level. These metrics extracted before motor intention could be applied to BCI models targeting MI to monitor a user's degree of training.