Quantifying and evaluating motor imagery ability using EEG microstates in MI-BCI training.

Brain-computer interfaces (BCIs) based on motor imagery (MI) demonstrate significant potential in device control, gaming, and clinical rehabilitation. However, due to various factors, the successful implementation of the MI-BCI system is not a straightforward process, and many subjects fail to effectively operate the system as expected. To avoid additional training for these subjects, accurate assessment and prediction of MI ability are crucial for adjusting training strategies. To this end, our study investigates the neuroelectrophysiological correlates of MI ability, aiming to establish a multivariate linear evaluation model utilizing electroencephalogram (EEG) microstate parameters that directly reflect cortical dynamics. We collected and analyzed data from 18 participants, identifying possible evaluation factors through correlations between microstate parameters and both classification accuracy and event-related desynchronization (ERD) power characteristics. The results showed that the differences between the subjects were closely related to their sensorimotor ability, attention, and cognitive ability. The established evaluation model exhibited robust assessment accuracy (R2 = 0.77, p < 0.001, MAE = 5.65) even when utilizing limited experimental data (20 trials). Taken together, the microstate-based evaluation approach in this study offers a new perspective on understanding and analyzing the causes of MI ability differences, which can help identify users suitable for BCI systems and promote subsequent personalized training, application promotion, and development.