MEG Channel Selection Using Copula Entropy-Based Transfer Entropy for Motor Imagery BCI.

Area of Science:

• Neuroscience
• Biomedical Engineering
• Signal Processing

Background:

• Magnetoencephalography (MEG) offers high spatiotemporal resolution crucial for motor imagery (MI)-based brain-machine interfaces (BCIs).
• Optimal BCI performance necessitates identifying and utilizing the most informative MEG channels, as not all channels contribute equally.

Purpose of the Study:

• To develop and validate a novel channel selection method for MI-BCIs using copula entropy-based transfer entropy (CTE).
• To enhance BCI classification accuracy and reduce the number of required MEG channels by selecting task-relevant channels.

Main Methods:

• A novel channel selection technique employing copula entropy-based transfer entropy (CTE) was developed.
• The proposed method quantifies causal relationships between channels during MI tasks to identify task-relevant ones.
• Experiments were conducted using a publicly available MEG dataset.

Main Results:

• Channel selection based on CTE significantly improved single-session classification accuracy (p < 0.05) compared to using all channels.
• The CTE method substantially reduced the number of MEG channels required for effective BCI operation.
• Cross-session classification performance using CTE-selected channels surpassed that of competing methods.

Conclusions:

• The proposed CTE-based channel selection method is effective for enhancing MI-BCI performance.
• This approach offers a significant reduction in channel requirements while improving classification accuracy.
• CTE provides a valuable tool for optimizing channel selection in MEG-based BCIs.