Collective sparse symmetric non-negative matrix factorization for identifying overlapping communities in...

Area of Science:

• Neuroscience
• Network Science
• Data Science

Background:

• Resting-state functional magnetic resonance imaging (rs-fMRI) is crucial for studying spontaneous brain activity and network organization.
• Existing group-level analyses often overlook overlapping community structures and individual variations in brain networks.
• Recent interest highlights the importance of overlapping community structures in complex networks, including the brain.

Purpose of the Study:

• To develop a novel method for identifying overlapping community structures in brain functional networks.
• To simultaneously account for group-level organization and inter-subject variability in brain networks.
• To enhance the analysis of brain functional networks using rs-fMRI data.

Main Methods:

• Proposed a new method: collective sparse symmetric non-negative matrix factorization (cssNMF).
• Applied cssNMF to identify group-level overlapping communities across subjects.
• Preserved individual variations in brain functional network organization.
• Validated the method using simulated and real rs-fMRI datasets.

Main Results:

• cssNMF accurately and stably identifies group-level overlapping communities.
• The method effectively captures individual differences in brain network organization.
• Results demonstrate neurophysiologically meaningful interpretations.
• Validated performance through comparisons on diverse datasets.

Conclusions:

• cssNMF offers a robust approach to analyzing brain functional networks.
• The method advances understanding of common community structures and individual differences in brain organization.
• This research provides new insights into the complexity of brain functional networks.