The Functional Segregation and Integration Model: Mixture Model Representations of Consistent and Variable...

Area of Science:

• Neuroscience
• Computational Neuroscience
• Brain Imaging Analysis

Background:

• Brain function relies on both segregated (local) and integrated (distributed) processing between specialized cortical regions.
• Functional magnetic resonance imaging (fMRI) is crucial for studying these relationships, but modeling high-dimensional fMRI data remains challenging.
• Conventional Gaussian mixture models (GMMs) for brain parcellation lack flexibility, failing to model interregional connectivity or network variability.

Purpose of the Study:

• To develop an advanced model, the Functional Segregation and Integration Model (FSIM), extending GMMs for robust fMRI data analysis.
• To simultaneously estimate spatial clustering, group functional connectivity, and model network variability.
• To improve the interpretability and predictive power of brain network models.

Main Methods:

• Developed the Functional Segregation and Integration Model (FSIM) as an extension of Gaussian mixture models (GMMs).
• FSIM simultaneously estimates spatial clustering and group functional connectivity, incorporating voxel- and subject-specific network scaling profiles.
• Compared FSIM against standard GMM using simulated and experimental resting-state fMRI data within a predictive cross-validation framework.

Main Results:

• FSIM's enhanced flexibility did not significantly alter parcellation reliability compared to standard GMM.
• Voxel- and subject-specific network scaling profiles significantly improved the prediction of functional connectivity in independent data.
• FSIM provides interpretable parameters characterizing consistent and variable aspects of functional connectivity structure.

Conclusions:

• The FSIM effectively summarizes functional connectivity structure in group-level fMRI data.
• FSIM enables modeling relationships between network variability and behavioral/demographic variables, demonstrated by predicting subject age.
• This model offers a powerful tool for understanding brain network dynamics and individual differences.