A brain functional network feature extraction method based on directed transfer function and graph theory for MI-BCI...

Area of Science:

• Neuroscience
• Biomedical Engineering
• Computer Science

Background:

• Brain-Computer Interface (BCI) technology holds significant promise across various domains.
• Enhancing the accuracy of BCI decoding algorithms through effective feature extraction from electroencephalographic (EEG) signals is a key research focus.

Purpose of the Study:

• To propose a novel method (CDGL) for extracting brain functional network features using directed transfer function (DTF) and graph theory.
• To integrate these network features with the Common Spatial Pattern (CSP) algorithm to improve motor imagery (MI) classification performance.

Main Methods:

• Utilized 32-channel EEG signals from 26 healthy participants.
• Constructed brain functional networks in Alpha and Beta bands using DTF, extracting node degree (ND), clustering coefficient (CC), and global efficiency (GE) via graph theory.
• Fused DTF and graph theory features with CSP, filtered redundant features using Lasso, and classified using a Support Vector Machine (SVM).

Main Results:

• The CDGL method achieved 89.13% accuracy in the Beta band with 8 electrodes, significantly outperforming the traditional CSP method by over 14%.
• Performance was superior with 8 channels compared to 4 channels and better in the Beta band than the Alpha band.
• Optimal performance was confirmed on two public EEG datasets.

Conclusions:

• Feature fusion of DTF network and graph theory significantly enhances CSP algorithm performance for MI classification.
• Increased channel count improves EEG signal feature extraction, boosting model sensitivity and discriminative ability.
• Beta band functional brain network features provide superior performance improvements compared to Alpha band features.