Detecting Dynamic Community Structure in Functional Brain Networks Across Individuals: A Multilayer Approach

Area of Science:

• Neuroscience
• Network Science
• Statistical Modeling

Background:

• Existing methods for community detection in brain networks are limited to single subjects or static networks.
• Analyzing dynamic, multi-subject brain networks requires advanced statistical approaches to capture individual and temporal variations.

Purpose of the Study:

• To develop a unified statistical framework for characterizing community structure in multi-subject, dynamic brain functional networks.
• To overcome limitations of existing methods by analyzing both individual differences and temporal evolution of network communities.

Main Methods:

• Proposed a multi-subject, Markov-switching stochastic block model (MSS-SBM) using a multilayer extension of the stochastic block model (SBM).
• Developed a novel fitting procedure based on multislice modularity maximization for simultaneous community partition across subjects.
• Integrated a dynamic Markov switching process to identify recurring temporal states and change points in inter-community interactions.

Main Results:

• Simulations demonstrated accurate community recovery and tracking of dynamic regimes in multilayer networks using MSS-SBM.
• Applied to task fMRI data, revealing group-level core-periphery structures associated with language and motor functions.
• Detected dynamic reconfiguration of modular connectivity and unique connectivity profiles across different task conditions.

Conclusions:

• The proposed multilayer network representation offers a principled method for detecting synchronous, dynamic modularity in brain networks across subjects.
• This framework advances the analysis of dynamic brain functional networks, providing insights into cognitive processes and information integration.