Detecting Changes in Correlation Networks with Application to Functional Connectivity of fMRI Data

Area of Science:

• Human sciences
• Neuroscience
• Psychology

Background:

• Human science research frequently investigates temporal changes in processes, such as shifts in brain states (fMRI) or psychological states (daily diaries).
• Dynamic processes are often simplified into static networks, potentially losing critical temporal information.
• Quantifying dynamic relations using correlation networks is essential for understanding state changes.

Purpose of the Study:

• To introduce and evaluate three novel data-driven methods for detecting change points in correlation networks.
• To provide methods for assessing significant differences between correlation network patterns from different time segments.
• To compare the performance of these change point detection methods on simulated and empirical data.

Main Methods:

• Described three methods for change point detection in correlation networks: dynamic connectivity regression, max-type method, and PCA-based method.
• Utilized lag-0 pair-wise correlation (or covariance) estimates to represent dynamic relations among variables.
• Employed significance testing approaches to compare network patterns across time segments.

Main Results:

• Presented three distinct data-driven approaches for identifying temporal shifts in correlation networks.
• Demonstrated the application of these methods on both simulated datasets and real functional connectivity fMRI data.
• Compared the efficacy of the proposed change point detection and significance testing methods.

Conclusions:

• The study offers valuable tools for analyzing dynamic processes in human sciences, particularly in neuroimaging and psychological research.
• The developed methods enable a more nuanced understanding of state changes by analyzing temporal network dynamics.
• These techniques can be broadly applied to identify significant changes in network structures across different time blocks.