A Novel Method for Extracting Hierarchical Functional Subnetworks Based on a Multisubject Spectral Clustering...

Area of Science:

• Neuroscience
• Network Science
• Data Science

Background:

• Brain network modularity analysis is crucial for understanding functional integration and segregation.
• Existing methods often focus on single-level modularity and binary network data.
• Hierarchical organization is a known characteristic of brain networks.

Purpose of the Study:

• To develop a novel multisubject spectral clustering technique for extracting hierarchical functional brain network structures.
• To address limitations of existing methods by utilizing full weighted connectivity information.
• To enhance the reliability of hierarchical clustering through ensemble methods.

Main Methods:

• Proposed a group-level network hierarchical clustering (GNetHiClus) technique.
• Employed bootstrap aggregation with majority voting for robust results.
• Evaluated the method on resting-state fMRI data from the Human Connectome Project across various group sample sizes.

Main Results:

• GNetHiClus consistently extracts hierarchical network structures across different sample sizes.
• The method successfully clusters brain networks into specialized subnetworks from high-level cognitive to low-level perceptual.
• Demonstrated hierarchical integration of information from lower to upper network levels for communication efficiency.

Conclusions:

• GNetHiClus provides a reliable approach to uncover hierarchical brain network organization.
• The findings align with principles of network segregation and integration.
• This technique can advance the understanding of brain function from a hierarchical perspective.