Direct lingam and visibility graphs for analyzing brain connectivity in BCI

Area of Science:

• Neuroscience
• Computational Neuroscience
• Biomedical Engineering

Background:

• Brain-computer interfaces (BCIs) enable direct communication between brain activity and external devices.
• Understanding directed connectivity within the brain is crucial for advancing motor imagery (MI)-based BCIs.
• Existing methods for analyzing brain networks require robust algorithms for complex signal processing.

Purpose of the Study:

• To investigate directed brain connectivity during motor imagery (MI) using two distinct network analysis methods.
• To compare the efficacy of Limited Penetrable Horizontal Visibility Graph (LPHVG) and Direct Lingam for MI classification.
• To evaluate the potential of graph theory in enhancing BCI performance.

Main Methods:

• Motor imagery electroencephalogram (MI-EEG) signals were mapped into networks using Limited Penetrable Horizontal Visibility Graph (LPHVG).
• Directed connectivity was assessed using LPHVG combined with Transfer Entropy (TE) and the Direct Lingam (Bayesian network) model.
• Support Vector Machine (SVM) and Convolutional Neural Network (CNN) classifiers were employed for MI classification.

Main Results:

• The LPHVG method achieved a classification accuracy of 92.7% for distinguishing 4 classes of MI.
• Direct Lingam achieved a classification accuracy of 90.6% for the same task.
• Analysis revealed distinct network structures across different MI classes, with LPHVG demonstrating superior discriminative capability.

Conclusions:

• Graph theory, particularly the LPHVG approach, shows significant potential for improving the efficiency and accuracy of motor imagery-based BCIs.
• The study demonstrates that LPHVG is a more effective method than Direct Lingam for analyzing directed brain connectivity in MI tasks.
• Network analysis of brain signals offers a promising avenue for developing more sophisticated and reliable brain-computer interfaces.