Quantifying temporal correlations: a test-retest evaluation of functional connectivity in resting-state fMRI

Area of Science:

• Neuroimaging
• Cognitive Neuroscience
• Data Science

Background:

• Resting-state functional magnetic resonance imaging (fMRI) studies frequently analyze functional connectivity (FC) using temporal correlations between brain regions.
• Various metrics exist for quantifying these temporal correlations, but their stability and reliability across different sessions are not consistently evaluated.

Purpose of the Study:

• To systematically investigate the test-retest stability of commonly used functional connectivity (FC) metrics in resting-state fMRI.
• To compare the performance of cross-correlation, partial cross-correlation, cross-coherence, and autoregressive model parameters using a public dataset.

Main Methods:

• Utilized a public-domain resting-state fMRI dataset with three sessions, including one with an 11-month interval and another with a 1-hour interval.
• Performed ROI-level and voxelwise seed-based analyses to assess the stability of different FC metrics.
• Evaluated cross-correlation, partial cross-correlation, cross-coherence, and autoregressive model parameters.

Main Results:

• Cross-correlation demonstrated superior stability compared to other metrics at both ROI and voxel levels.
• Partial cross-correlation showed a tendency to shrink towards zero for negatively correlated regions, impacting stability.
• FC stability was greater between sessions with a shorter interval (1 hour) than a longer interval (11 months).

Conclusions:

• Cross-correlation is recommended as a more stable metric for quantifying functional connectivity in resting-state fMRI.
• The choice of FC metric significantly impacts the reliability and interpretation of brain connectivity findings.
• Temporal proximity of scanning sessions influences the observed stability of functional connectivity measures.