A Riemannian Framework for Longitudinal Analysis of Resting-State Functional Connectivity

Area of Science:

• Neuroimaging
• Neuroscience
• Biostatistics

Background:

• Longitudinal resting-state functional connectivity (rs-fMRI) analysis is crucial for understanding brain changes over time.
• Characterizing dynamic functional connectivity across multiple visits remains a challenge in neuroimaging research.

Purpose of the Study:

• To develop a robust framework for characterizing longitudinal changes in functional connectivity using rs-fMRI data.
• To identify group-specific differences in connectivity trajectories, such as sex differences, across multiple scanning sessions.

Main Methods:

• A Riemannian framework was designed to represent functional connectivity as points on a manifold.
• Geodesic regression was employed to model longitudinal connectivity trajectories.
• Lie group actions and latent p-value theory were used to identify group differences in a common tangent space.

Main Results:

• The proposed method successfully identified sex differences in functional connectivity patterns.
• The framework accounts for the inherent uncertainty in analyzing longitudinal neuroimaging data.
• The approach was validated on a cohort of 246 subjects with three rs-fMRI scans each.

Conclusions:

• The Riemannian framework offers a powerful new approach for analyzing longitudinal rs-fMRI data.
• This method enhances the ability to detect subtle, time-dependent changes in brain functional connectivity.
• The findings highlight the potential for this framework in uncovering group-specific neurobiological differences.