Single-Trial NIRS Data Classification for Brain-Computer Interfaces Using Graph Signal Processing.

Area of Science:

• Neuroscience
• Signal Processing
• Biomedical Engineering

Background:

• Near-infrared spectroscopy (NIRS)-based brain-computer interface (BCI) systems commonly rely on slope and mean changes in hemodynamic responses for feature extraction.
• Existing methods often overlook the spatial patterns present across NIRS measurement channels, potentially limiting BCI performance.

Purpose of the Study:

• To introduce a novel feature extraction methodology for NIRS-based BCI systems that incorporates spatial information.
• To evaluate the effectiveness of the proposed method in improving classification accuracy and efficiency.

Main Methods:

• A graph signal processing (GSP) approach, termed graph NIRS (GNIRS), was developed to capture spatial patterns in NIRS signals.
• The GNIRS methodology was applied to a publicly available NIRS dataset recorded during a mental arithmetic task.
• Performance was compared against traditional slope and mean-based feature extraction techniques.

Main Results:

• GNIRS achieved higher classification rates (CRs), reaching up to 92.52%, compared to slope (90.35%) and mean (82.60%) methods.
• The proposed GNIRS method resulted in feature vectors with reduced dimensionality.
• High CRs were observed from the first second of the mental task onset, indicating faster system response.

Conclusions:

• The GSP-based GNIRS methodology effectively captures spatial information in NIRS signals for BCI applications.
• GNIRS offers superior classification performance and reduced feature dimensionality compared to existing methods.
• This approach has the potential to enable faster and more efficient NIRS-based BCI systems.