Graph Analysis and Modularity of Brain Functional Connectivity Networks: Searching for the Optimal Threshold

Area of Science:

• Neuroscience
• Network Science
• Statistical Physics

Background:

• Brain connectivity is often analyzed as complex networks.
• Graph theory and community detection are key tools for understanding brain network structure.
• Network sparsification is crucial but selecting the optimal threshold remains challenging.

Purpose of the Study:

• To introduce percolation analysis as a method for determining the optimal sparsification threshold in brain connectivity networks.
• To evaluate the effectiveness of percolation analysis in preserving community structure information during network simplification.
• To assess the robustness of this method against noise and data variability.

Main Methods:

• Utilized synthetic brain networks with known modular structures and realistic topological properties.
• Applied percolation analysis to identify optimal sparsification thresholds.
• Validated the approach using three community detection algorithms: Newman's modularity, InfoMap, and Asymptotical Surprise.
• Investigated the impact of noise and data variability on threshold selection.

Main Results:

• Percolation analysis successfully identified optimal sparsification thresholds that maximized community structure information.
• The method demonstrated effectiveness across different community detection algorithms.
• The chosen thresholds were robust to varying levels of noise and data variability.

Conclusions:

• Percolation analysis offers a data-driven approach to optimize network sparsification for brain connectivity studies.
• This method enhances the reliability of community detection in neuroimaging data, particularly in patient populations.
• It provides a principled way to balance network simplification with the preservation of crucial structural information.