Mental Task Evaluation for Hybrid NIRS-EEG Brain-Computer Interfaces.

Area of Science:

• Neuroscience
• Biomedical Engineering
• Signal Processing

Background:

• Recent electroencephalography (EEG) and near-infrared spectroscopy (NIRS) studies demonstrate distinct neural responses to cognitive tasks like motor imagery and mental arithmetic.
• These findings suggest potential for improved brain-computer interface (BCI) performance through multimodal data fusion.

Purpose of the Study:

• To propose and evaluate a hybrid BCI (hBCI) paradigm integrating EEG and NIRS data.
• To enhance binary classification performance by fusing simultaneous NIRS-EEG signals.
• To identify optimal task pairings and assess the benefits of a multimodal approach for BCI applications.

Main Methods:

• Simultaneous NIRS-EEG data were recorded from nine participants performing seven distinct mental tasks.
• Classifiers were trained using EEG features alone, NIRS features alone, and combined EEG-NIRS features for pairwise task classification.
• Performance was evaluated based on kappa statistics and classification accuracy across different time windows.

Main Results:

• The combined NIRS-EEG approach yielded an average peak kappa increase of 0.03 (1.5% accuracy) using one-second windows.
• A more substantial accuracy increase (10%) was observed when focusing on time windows with high NIRS performance.
• Analysis revealed key brain regions and feature types contributing to classification accuracy.

Conclusions:

• Fusion of NIRS and EEG data in an hBCI framework significantly improves binary classification performance.
• The multimodal approach offers potential for developing more efficient and flexible brain-computer interfaces.
• This study lays the groundwork for future hBCI research leveraging NIRS-EEG integration.