The predictive role of pre-cue EEG rhythms on MI-based BCI classification performance

Area of Science:

• Neuroscience
• Biomedical Engineering
• Signal Processing

Background:

• Motor imagery-based brain-computer interfaces (BCI) face performance variability among users.
• Understanding the reasons for low user performance is crucial for BCI development.
• Predicting user performance can aid in optimizing BCI system design.

Purpose of the Study:

• To introduce a novel coefficient derived from pre-cue electroencephalography (EEG) rhythms.
• To investigate the coefficient's ability to predict motor imagery (MI) classification performance in BCI users.
• To explore the relationship between pre-cue EEG spectral power and BCI accuracy.

Main Methods:

• A new coefficient was computed using spectral power of pre-cue EEG data across different brain regions.
• The coefficient integrates both spectral and spatial information, unlike previous predictors.
• Feasibility was tested by correlating the coefficient with cross-validation accuracies in 17 healthy subjects.

Main Results:

• A significant positive correlation (r=0.53, p=0.02) was found between the proposed coefficient and MI classification accuracy.
• Subjects with higher BCI performance exhibited significantly higher coefficient values.
• The coefficient effectively captures individual differences in MI-based BCI performance.

Conclusions:

• Higher frontal theta and lower posterior alpha power before MI may enhance BCI performance.
• The findings suggest a potential method for user-specific BCI training.
• This coefficient offers a promising avenue for developing experiments to improve user motor imagery capabilities.