Time-varying dynamic network model for dynamic resting state functional connectivity in fMRI and MEG imaging

Area of Science:

• Neuroscience
• Computational Neuroscience
• Network Science

Background:

• Dynamic resting state functional connectivity (RSFC) reveals temporal fluctuations in brain networks.
• Existing methods (e.g., sliding-window, clustering) for dynamic RSFC analysis have limitations including high dimensionality, poor signal reconstruction, data insufficiency, insensitivity to rapid changes, and limited cross-modality generalizability.

Purpose of the Study:

• To develop a novel, unifying, and adaptive framework called time-varying dynamic network (TVDN) for analyzing dynamic RSFC.
• To overcome the limitations of existing non-adaptive methods for dynamic RSFC estimation.

Main Methods:

• Developed a generative model within the TVDN framework to link low-dimensional dynamic RSFC to brain signals.
• Implemented an adaptive inference algorithm to learn the dynamic RSFC manifold and detect state transitions.
• Validated TVDN's applicability to multiple neuroimaging modalities (fMRI, MEG/EEG).
• Evaluated performance using signal reconstruction capabilities and comprehensive simulations.

Main Results:

• TVDN successfully captures brain activity dynamics and detects brain state switching.
• Demonstrated robust performance in both simulated data and real fMRI and MEG datasets.
• Outperformed existing benchmark methods in capturing dynamic functional connectivity.

Conclusions:

• The TVDN framework offers a significant advancement in analyzing dynamic functional brain connectivity.
• TVDN provides a robust and generalizable approach applicable across different neuroimaging modalities.
• This method enhances our ability to understand dynamic brain states and their transitions.