The construction of spatio-temporal functional brain network based on Ising model for EEG classification

Area of Science:

• Neuroscience
• Computational Neuroscience
• Biomedical Engineering

Background:

• Functional brain networks (FBNs) are crucial for brain analysis but often fail to capture individual temporal correlations and multi-term features.
• Existing single-layer, single-scale FBNs limit classification accuracy and generalizability.

Purpose of the Study:

• To develop a multi-scale spatio-temporal FBN that effectively represents temporal variability and spatial distribution from electroencephalography (EEG) data.
• To improve the accuracy and generalizability of brain network analysis for classification tasks.

Main Methods:

• Constructed a multi-scale spatio-temporal FBN using EEG data, incorporating brain field data aggregation via an Ising model.
• Calculated autocorrelation between symbol subsequences to represent temporal brain states.
• Built temporal FBNs with symbol sequence correlations as link weights and spatial FBNs with inter-channel functional connectivities.

Main Results:

• Achieved classification accuracies up to 99% on datasets for fatigue detection, emotion recognition, Parkinson's diagnosis, and motor imagery.
• Demonstrated the effectiveness, efficiency, and generalizability of the spatio-temporal FBN approach.

Conclusions:

• The spatio-temporal FBN effectively represents short-term and long-term individual features.
• This method enables universal recognition across individuals and accurate distinction between categories.