Large-Scale Cortical Network Analysis and Classification of MI-BCI Tasks Based on Bayesian Nonnegative Matrix...

Area of Science:

• Neuroscience
• Cognitive Science
• Biomedical Engineering

Background:

• Motor imagery (MI) is crucial for clinical rehabilitation and brain-computer interfaces (BCIs).
• Decoding MI tasks and understanding its neural basis remain challenging, hindering clinical applications.
• Current research lacks detailed analysis of large-scale cortical network differences between left- and right-hand MI.

Purpose of the Study:

• To investigate the neural mechanisms underlying left- and right-hand motor imagery tasks.
• To construct and analyze large-scale cortical networks associated with MI tasks.
• To develop a classification model for MI-BCI tasks based on network properties.

Main Methods:

• Combined EEG source reconstruction and Bayesian nonnegative matrix factorization (NMF).
• Constructed large-scale cortical networks for left- and right-hand MI tasks.
• Analyzed functional network connectivities (FNCs) and network properties (clustering coefficient, efficiency, path length).

Main Results:

• Left-hand MI showed significantly increased FNCs within and among visual (VN), sensorimotor (SMN), and other right-hemisphere networks in the beta and all frequency bands.
• Left-hand MI exhibited higher clustering coefficient, global efficiency, and local efficiency, with decreased characteristic path length compared to right-hand MI.
• A classification model based on FNC and network properties achieved 78.2% accuracy in cross-subject two-class MI-BCI tasks.

Conclusions:

• Left-hand MI necessitates greater modulation of widespread cortical networks, particularly in the right hemisphere.
• Identified distinct network patterns and properties differentiating left- and right-hand MI.
• Proposed a potential network biomarker for identifying MI-BCI tasks, advancing BCI development.