Improvement of Information Transfer Rates Using a Hybrid EEG-NIRS Brain-Computer Interface with a Short Trial Length:...

Area of Science:

• Neuroscience
• Biomedical Engineering
• Brain-Computer Interfaces

Background:

• Non-invasive neuroimaging methods like Electroencephalography (EEG) and Near-Infrared Spectroscopy (NIRS) record brain activity.
• Hybrid EEG-NIRS brain-computer interfaces (hBCIs) combine EEG and NIRS to enhance performance over unimodal systems.
• Existing hBCIs often require long trial lengths (≥10s) due to hemodynamic delays, limiting the information transfer rate (ITR).

Purpose of the Study:

• To develop a more practical hBCI by utilizing intuitive mental tasks (mental arithmetic and word chain) within a shorter trial length (5s).
• To assess the suitability of the word chain task for BCI applications.
• To improve the ITR of hBCIs by overcoming limitations of long trial durations.

Main Methods:

• Simultaneously recorded EEG and NIRS data during mental arithmetic (MA) and word chain (WC) tasks, and a baseline (BL) condition.
• Tasks were performed in short 5-second trials without prior training.
• Classified brain activations related to MA or WC tasks against BL activations using offline/pseudo-online analyses.

Main Results:

• Achieved high classification accuracies: 90.0 ± 7.1% for MA vs. BL and 85.8 ± 8.6% for WC vs. BL.
• Results were significantly higher than unimodal EEG- or NIRS-BCI in most cases.
• Demonstrated substantial improvements in average ITRs: >96.6% for MA vs. BL and 87.1% for WC vs. BL compared to previous studies.

Conclusions:

• Validated the feasibility of a practical hBCI using intuitive tasks and short trial lengths.
• The word chain task is suitable for BCI applications.
• The proposed hBCI approach significantly enhances classification accuracy and ITR, offering a more efficient brain-computer interface.