Enhancing Classification Performance of fNIRS-BCI by Identifying Cortically Active Channels Using the z-Score Method

Area of Science:

• Neuroscience
• Biomedical Engineering
• Signal Processing

Background:

• Brain-computer interface (BCI) systems integrate multiple stages, including signal acquisition, noise reduction, channel selection, feature extraction, and classification.
• Channel selection in functional near-infrared spectroscopy-based BCI (fNIRS-BCI) is crucial for identifying relevant brain regions and enhancing classification accuracy.

Purpose of the Study:

• To propose and evaluate a novel z-score method for channel selection to improve fNIRS-BCI performance.
• To compare the efficacy of the z-score method against the conventional t-value method and a no-channel-selection approach.

Main Methods:

• The z-score method utilizes cross-correlation to assess the similarity between desired and recorded brain activity signals.
• It involves calculating the z-score for the maximum correlation coefficients of each channel, selecting channels with positive z-scores.
• The method was applied to open-access datasets for mental arithmetic, motor imagery, and finger/foot tapping tasks.

Main Results:

• The z-score method achieved significantly improved classification accuracies: 87.2 ± 7.0% for left motor imagery vs. rest, 88.4 ± 6.2% for right motor imagery vs. rest, and 88.1 ± 6.9% for mental arithmetic vs. rest (p < 0.0167).
• Validation on a separate dataset confirmed the enhanced performance over the t-value method.

Conclusions:

• The proposed z-score method offers a significant advancement in channel selection for fNIRS-BCI systems.
• This method effectively improves classification performance, contributing to the development of more sophisticated fNIRS-BCI applications.