Performance Enhancement of Steady-State Visual Evoked Field-Based Brain-Computer Interfaces Incorporating MEG Source...

Area of Science:

• Neuroscience
• Biomedical Engineering
• Signal Processing

Background:

• Helmet-type magnetoencephalography (MEG) systems, including optically pumped magnetometer (OPM)-based MEG, are gaining traction for brain-computer interfaces (BCIs).
• Steady-state visual evoked field (SSVEF)-based BCIs offer high information transfer rates (ITR) and require minimal calibration.
• Conventional algorithms do not fully leverage MEG's spatial resolution and whole-head coverage.

Purpose of the Study:

• To develop a novel weighting method to enhance SSVEF-based BCI performance using MEG source imaging.
• To fully utilize whole-head MEG recordings by incorporating averaged source locations of SSVEF signals.

Main Methods:

• Developed and applied the averaged source location-based weighting (ASLW) method.
• Integrated ASLW with the filter bank-driven multivariate synchronization index (FBMSI) algorithm (ASLW-FBMSI).
• Evaluated performance using classification accuracy and ITR across various window sizes with 20 participants.

Main Results:

• The ASLW-FBMSI algorithm significantly improved classification accuracy and ITR compared to conventional methods.
• Performance gains included a 13.9% accuracy increase (3-s window) and a 13.1 bits/min ITR increase (2.5-s window).
• The algorithm demonstrated a low processing delay (0.107 s for 4-s data) and was validated in online experiments.

Conclusions:

• The ASLW method effectively enhances SSVEF-based BCI performance by leveraging MEG source imaging.
• ASLW-FBMSI offers a promising approach for improving BCI accuracy and speed.
• The findings support the broader applicability of ASLW in MEG-based BCI research.