Frequency-Resolved Dynamic Functional Connectivity Reveals Scale-Stable Features of Connectivity-States

Area of Science:

• Neuroscience
• Connectomics
• Signal Processing

Background:

• Investigating temporal variability in functional connectivity is crucial in connectomics.
• Sliding window techniques on resting-state fMRI (rs-fMRI) time courses are commonly used for dynamic functional connectivity analysis.

Purpose of the Study:

• Introduce frequency-resolved dynamic functional connectivity (frdFC) using multivariate empirical mode decomposition (MEMD).
• Assess the robustness and scale-stability of connectivity states across different frequencies.
• Propose scale-stability as a quality criterion for connectivity-state and model selection.

Main Methods:

• Applied multivariate empirical mode decomposition (MEMD) to rs-fMRI time courses.
• Conducted filter-bank investigations to analyze frequency-specific connectivity.
• Utilized k-means clustering to identify connectivity states.
• Employed null-models, simulations, and theoretical considerations for validation.

Main Results:

• MEMD successfully generated time courses for frdFC analysis.
• Connectivity-state structures were robust across frequency scales, becoming more evident at lower frequencies.
• A drop-off in scale-stability was observed between k=4 and k=5 clusters.
• Filter design proved more sensitive in rs-fMRI data compared to simulated data.

Conclusions:

• Connectivity states represent a multivariate and multiscale phenomenon.
• Scale-stability serves as a valuable quality criterion for selecting connectivity states and models.
• The study provides a foundation for future frdFC research and offers a data repository.