A Multiscale Consensus Method Using Factor Analysis to Extract Modular Regions in the Functional Brain Network.

Area of Science:

• Neuroscience
• Network Science
• Computational Biology

Background:

• Brain functional connectivity is typically analyzed using correlation matrices derived from regions of interest (ROIs).
• Functional segregation, a key organizational principle of the brain, is usually extracted via edge-filtering and community detection algorithms.
• Existing methods often require network sparsification, which can lead to information loss.

Purpose of the Study:

• To propose a novel method for extracting functional segregation from brain correlation matrices using exploratory factor analysis (EFA).
• To address the limitations of traditional EFA, such as replication and generalizability issues, by employing a multiscale consensus approach.
• To avoid network sparsification by directly applying EFA to the correlation matrix.

Main Methods:

• A multiscale approach using EFA for node-partitioning was developed to overcome the challenge of determining the optimal number of factors.
• Consensus aggregation was used to combine EFA results across different scales.
• The influence of the scale interval on performance was analyzed.
• The proposed method was compared against state-of-the-art techniques in a case study.

Main Results:

• The multiscale consensus EFA method demonstrates comparable performance to current state-of-the-art methods for extracting functional segregation.
• The approach successfully identifies modular brain regions without requiring edge-filtering or binarization of the network.
• The study provides insights into the appropriate scale selection and the impact of scale intervals.

Conclusions:

• The proposed multiscale consensus EFA offers a robust and effective alternative for identifying functional segregation in brain networks.
• This method facilitates the study of spontaneous brain activity and modular organization in resting-state networks.
• The findings suggest potential for improved analysis of brain functional architecture in clinical and research settings.