Multiframe Evolving Dynamic Functional Network Connectivity Motifs (Evodfncs) from Continuity-Preserving Planar...

Area of Science:

• Neuroscience
• Computational Neuroscience
• Medical Imaging

Background:

• Brain network connectivity is increasingly studied as a dynamic, time-varying property.
• Current methods often focus on discrete static states, limiting the understanding of evolving connectivity patterns.
• Leveraging large resting-state functional magnetic resonance imaging (rs-fMRI) datasets requires advanced techniques to capture temporal dynamics.

Purpose of the Study:

• To develop a flexible, data-driven framework for identifying group-level dynamic functional network connectivity (dFNC) states.
• To visualize complex, high-dimensional brain connectivity dynamics in a lower-dimensional space.
• To characterize differences in temporal connectivity patterns between patient groups and healthy controls.

Main Methods:

• Utilized uniform manifold approximation and embedding (UMAP) for dimensionality reduction of whole-brain connectivity dynamics.
• Developed a framework for analyzing multiframe (movie-style) dFNC states.
• Applied the method to a large rs-fMRI dataset from a study of schizophrenia.

Main Results:

• Successfully extracted naturalistic, fluidly-varying connectivity motifs from rs-fMRI data.
• Identified distinct dFNC states that differ between schizophrenia patients (SZs) and healthy controls (HC).
• The UMAP embedding preserved key features of high-dimensional connectivity dynamics, including trajectory continuity.

Conclusions:

• The proposed framework enables robust identification and visualization of group-level dFNC states.
• This approach offers a powerful tool for understanding brain dynamics in neurological and psychiatric disorders.
• The findings highlight the potential of dynamic connectivity analysis in schizophrenia research using rs-fMRI.