A Comparative Analysis of Dynamic Functional Connectivity Methods Using Naturalistic fMRI

Abstract

Introduction

The brain dynamically responds to internal and external stimuli across multiple temporal scales. From this continuous stream of information, people automatically segment experiences into meaningful, discrete events–a process fundamental to perception, memory, and learning. Recent research shows that event segmentation is closely linked to transitions between stable neural states, reflected in dynamic changes in brain network interactions during naturalistic experience [1]. Dynamic functional connectivity (dFC) refers to time-varying patterns of statistical dependencies between brain regions, providing a powerful framework for investigating how neural states evolve during cognitive processes in real-world settings.

Challenges and Motivation

Measuring and interpreting dFC is complicated by its variability across scanning sessions and time scales [2]. Sources of variability include task demands, spontaneous neural state transitions, and non-neural noise (e.g., physiological signals or scanner artifacts), making it difficult to identify the sources and temporal characteristics of dFC changes. Various dFC analysis methods exist—each with distinct advantages and limitations—but their selection depends critically on the research question, experimental design, and stimulus type. Naturalistic stimuli, such as film clips, provide a more ecologically valid approximation of real-world experiences than traditional tasks [3]. Their temporal structure and semantic richness offer clear markers for event segmentation, improving interpretations of dFC to better reflect how neural states and network reconfigurations unfold during natural cognition. Using naturalistic stimuli, researchers can more effectively reveal neural mechanisms segmenting continuous experience into discrete events, bridging experimental findings with everyday brain function.

Objectives and Methodology

This thesis will compare key dFC methods—sliding window correlation, time-frequency analysis, and dynamic graph analysis—using publicly available movie-viewing fMRI datasets. The comparison will assess how these methods differ in detecting transient neural states, identifying event boundaries, and quantifying dynamic changes in network properties such as global efficiency, local efficiency, and clustering coefficient, while maintaining robustness to motion and physiological noise. While neuroimaging techniques like EEG offer higher temporal resolution, this analysis focuses on fMRI due to its superior spatial coverage and suitability for studying large-scale brain networks over extended periods.

Significance

Real-world cognition is dynamic and complex. By systematically comparing dFC methods in the context of event segmentation during naturalistic viewing, this thesis aims to strengthen the methodological bridge between dFC analysis and ecologically valid stimuli. It seeks to clarify how transient neural states and changes in functional connectivity underlie our experience of meaningful, discrete events. Ultimately, these insights will inform method selection for future studies and contribute to a broader understanding of how the brain dynamically organizes cognition in real-world settings.

References

[1] D. Oetringer, D. Gözükara, U. Güçlü, and L. Geerligs, “The neural basis of event segmentation: Stable features in the environment are reflected by neural states,” Imaging Neuroscience, vol. 3, p. imag_a_00432, Jan. 2025. doi: 10.1162/imag_a_00432.

[2] R. M. Hutchison et al., “Dynamic functional connectivity: Promise, issues, and interpretations,” NeuroImage, vol. 80, pp. 360–378, Oct. 2013. doi: 10.1016/j.neuroimage.2013.05.079.

[3] S. Sonkusare, M. Breakspear, and C. Guo, “Naturalistic Stimuli in Neuroscience: Critically Acclaimed,” Trends in Cognitive Sciences, vol. 23, no. 8, pp. 699–714, Aug. 2019. doi: 10.1016/j.tics.2019.05.004.