Identification of community structure-based brain states and transitions using functional MRI

Area of Science:

• Neuroscience
• Computational Neuroscience
• Network Science

Background:

• Brain function depends on a dynamic balance between network integration and segregation.
• Identifying hidden brain states and their transitions is a significant challenge in neuroscience.

Purpose of the Study:

• To propose a Bayesian method for characterizing latent brain states based on community structure.
• To develop a novel strategy for detecting transitions between community structures in blood oxygen level-dependent (BOLD) time series data.

Main Methods:

• Utilized a latent block model with posterior predictive discrepancy to detect changes in community structure.
• Estimated parameters including latent node-to-community assignments and weighted connectivity within/between communities.
• Validated the method using in-silico evaluations and the Human Connectome Project (HCP) dataset.

Main Results:

• The Bayesian method successfully characterized latent brain states and detected transitions in BOLD time series.
• Empirical validation on HCP task-fMRI data demonstrated appropriate lags between task demands and state changes.
• Distinct community patterns were identified for fixation, low-demand, and high-demand working memory conditions.

Conclusions:

• The proposed Bayesian approach effectively identifies and tracks dynamic brain states and their transitions.
• This method provides a robust framework for analyzing functional brain network reconfiguration during cognitive tasks.
• Findings contribute to understanding the neural mechanisms underlying working memory and cognitive flexibility.