Improved motor imagery training for subject's self-modulation in EEG-based brain-computer interface.

Area of Science:

• Neuroscience
• Biomedical Engineering
• Human-Computer Interaction

Background:

• Motor imagery (MI) training is crucial for electroencephalogram (EEG)-based brain-computer interface (BCI) systems.
• Current MI training protocols receive less attention than machine learning algorithms.
• Effective MI training requires subjects to actively modulate brain activity during calibration.

Purpose of the Study:

• To propose and evaluate a novel trial-feedback paradigm for improving MI training.
• To compare the effectiveness of the trial-feedback paradigm against a non-feedback paradigm.
• To enhance subjects' self-modulation abilities for better MI task performance.

Main Methods:

• A within-subject design comparing a trial-feedback and a non-feedback paradigm across two sessions.
• Real-time topographic map visualization and qualitative evaluation after each MI trial in the feedback paradigm.
• Post-calibration feature distribution visualization and quantification.
• Electrooculogram (EOG) signal monitoring to discard distracted trials.

Main Results:

• The trial-feedback paradigm demonstrated superior spatial filter visualization compared to the non-feedback paradigm.
• Higher average offline and online classification accuracies were achieved with the trial-feedback session.
• The trial-feedback approach showed greater utility in promoting subject self-modulation and MI task performance.

Conclusions:

• The proposed trial-feedback paradigm significantly enhances MI training effectiveness for BCI systems.
• Real-time feedback and visualization aid subjects in understanding and adjusting their brain activity.
• This approach offers a promising direction for improving BCI performance through optimized user training.