Enhancing Auditory BCI Performance: Incorporation of Connectivity Analysis

Area of Science:

• Neuroscience
• Biomedical Engineering
• Machine Learning

Background:

• Auditory brain-computer interface (BCI) technology is advancing, with invasive methods like intracranial electroencephalography (iEEG) offering high signal fidelity.
• Distinguishing between similar auditory stimuli (e.g., speech vs. music) using brain connectivity remains a challenge for current BCI systems.

Purpose of the Study:

• To investigate the efficacy of brain connectivity matrices in classifying auditory stimuli using iEEG data.
• To enhance the precision of auditory BCI systems by adapting noninvasive BCI frameworks for invasive data.
• To identify optimal brain wave bands and connectivity metrics for improved classification accuracy.

Main Methods:

• Analysis of intracranial electroencephalography (iEEG) data to compute brain connectivity matrices.
• Utilized various brain wave bands, including alpha, beta, theta, and gamma.
• Employed connectivity metrics such as Phase Locking Values (PLV) and Coherence.
• Applied a Support Vector Machine (SVM) classifier to brain connectivity data.

Main Results:

• Brain connectivity matrices effectively classified auditory stimuli with high precision.
• The Support Vector Machine (SVM) classifier achieved 97% accuracy in distinguishing brain states.
• Neural activity within the gamma band demonstrated the highest classification performance.
• The proposed methods improved upon previous studies by 9.64%.

Conclusions:

• Brain connectivity analysis, particularly using gamma band activity and metrics like PLV and Coherence, significantly enhances auditory BCI performance.
• The integration of noninvasive BCI methodologies with invasive iEEG data offers a promising pathway for developing more sophisticated auditory BCI systems.
• High classification accuracy achieved with SVM highlights its suitability for processing complex neural connectivity data in BCI applications.