Consistency of network modules in resting-state FMRI connectome data

Area of Science:

• Neuroscience
• Brain Imaging
• Network Science

Background:

• Spontaneous brain activity at rest forms distinct resting state networks (RSNs) with correlated temporal dynamics.
• Spatial independent component analysis (ICA) and graph theory methods consistently identify RSNs, though often at different spatial resolutions.

Purpose of the Study:

• To directly compare the consistency of RSNs identified by ICA and graph theory-based network analyses.
• To evaluate RSN robustness across subjects at a voxel-level resolution using scaled inclusivity (SI).

Main Methods:

• Analysis of resting-state fMRI data from 194 subjects at voxel-level resolution.
• Application of scaled inclusivity (SI) metric to quantify modular partition consistency across subjects.
• Direct comparison of RSNs derived from spatial ICA and graph theory approaches.

Main Results:

• Some RSNs demonstrate robust consistency across subjects, comparable to ICA findings.
• Certain commonly reported RSNs exhibit lower inter-subject consistency.
• This study provides the first direct comparison of these methods at comparable resolutions.

Conclusions:

• The consistency of RSNs varies, with some being more reliable than others across individuals.
• The findings highlight the importance of resolution in network identification and comparison.
• This research offers insights into the reliability of different analytical approaches for resting-state fMRI data.