Measuring the Non-linear Directed Information Flow in Schizophrenia by Multivariate Transfer Entropy

Area of Science:

• Neuroscience
• Computational Neuroscience
• Biomedical Engineering

Background:

• Schizophrenia (SCZ) is associated with severe deterioration of brain network function.
• Existing effective connectivity methods struggle to quantify direct non-linear causal interactions in brain activity.
• Electroencephalogram (EEG) data reflects complex, non-linear causal activities within the brain.

Purpose of the Study:

• To introduce and validate Multivariate Transfer Entropy (MTE) for quantifying non-linear causal interactions in EEG.
• To compare the performance of MTE against Granger Causal Analysis (GCA) and Bivariate Transfer Entropy (BVTE).
• To investigate brain network differences between individuals with SCZ and healthy controls (HC) using MTE.

Main Methods:

• Construction of EEG effective networks using MTE.
• Comparative analysis of MTE, GCA, and BVTE performance using simulations under varying signal-to-noise ratios, edge recovery, sensitivity, and specificity.
• Application of MTE to P300 task EEG data from HC and SCZ groups.

Main Results:

• MTE demonstrated superior performance over GCA and BVTE in simulations across multiple metrics.
• MTE successfully identified deteriorated network interactions in SCZ patients compared to HC.
• The study confirmed MTE's capability to capture both linear and non-linear causal relationships.

Conclusions:

• MTE is a robust and effective tool for quantifying non-linear causal interactions in EEG data.
• MTE reveals significant brain network deterioration in schizophrenia.
• This novel approach offers potential for deeper understanding of brain network dysfunction in SCZ.