Markov Model-Based Method to Analyse Time-Varying Networks in EEG Task-Related Data.

Area of Science:

• Cognitive Neuroscience
• Computational Neuroscience
• Neuroimaging Analysis

Background:

• Understanding the dynamic nature of functional brain networks is crucial in cognitive neuroscience.
• Existing methods for analyzing time-varying brain networks in EEG/MEG data are limited.
• Novel analytical approaches are needed to capture rapid changes in neural interactions during cognitive tasks.

Purpose of the Study:

• To propose and validate a novel pipeline for characterizing time-varying functional brain networks in single-subject EEG data.
• To evaluate the pipeline's effectiveness on both simulated and experimental datasets, including a Brain-Computer Interface (BCI) task.
• To demonstrate the utility of a Markov Model (MM) framework for analyzing dynamic neural activity.

Main Methods:

• Pre-processing EEG data to remove channel- and trial-wise variations.
• Estimating functional networks using sparse Multi-Variate Auto-Regressive (sMVAR) models within short time windows.
• Identifying functional states via k-means clustering and describing trial sequences using Markov Models.

Main Results:

• The pipeline successfully discriminated between experimental conditions using simulated EEG data.
• Application to a P300 oddball task demonstrated statistically significant discrimination between target and non-target trials using only Markov Model parameters.
• Identified functional network states showed high similarity between individuals.

Conclusions:

• The proposed pipeline offers a valid and novel approach for analyzing time-varying functional brain networks in EEG/MEG data.
• The Markov Model framework provides a powerful tool for inferring dynamic network changes, orthogonal to conventional methods like ERP averaging.
• This work serves as a proof-of-concept for tracking rapid neural interaction patterns during task performance.