Can sliding-window correlations reveal dynamic functional connectivity in resting-state fMRI?

Area of Science:

• Neuroimaging
• Cognitive Neuroscience
• Data Analysis

Background:

• Research in resting-state functional magnetic resonance imaging (fMRI) increasingly focuses on dynamic functional connectivity (dFC) within sessions.
• Current statistical assessments of dFC are often inadequate or omitted, potentially leading to inaccurate conclusions.

Purpose of the Study:

• To highlight the necessity of robust statistical tests for detecting dFC.
• To provide a methodology for carrying out and assessing dFC measures.
• To address statistical pitfalls in dFC analysis.

Main Methods:

• Simulations and analysis of spontaneous blood-oxygen-level-dependent (BOLD) fMRI data from macaques and humans.
• Focus on sliding-window correlations and a non-linear dFC measure.
• Methodology designed to be generalizable to various dFC measures.

Main Results:

• Simulations indicate that detecting dFC in typical 10-minute resting-state fMRI sessions using sliding-window correlations is nearly impossible.
• This finding was validated in both macaque and human datasets, with no significant dFC detected in individual sessions.
• Session- or subject-averaging significantly increases detection power, revealing that most functional connections are dynamic.

Conclusions:

• Appropriate statistical methods are essential to avoid pitfalls in dFC assessment.
• The study emphasizes the dynamic nature of most functional connections when analyzed with sufficient statistical power.
• Awareness and application of sound statistical practices are crucial for advancing dFC research.