Time-dependence of graph theory metrics in functional connectivity analysis
Sharon Chiang1, Alberto Cassese2, Michele Guindani3
1Department of Statistics, Rice University, Houston, TX, USA.
Neuroimage
|November 1, 2015
Summary
Brain network analysis using graph theory often assumes stable connectivity. This study introduces methods to quantify dynamic brain network stationarity, revealing more robust measures and improving disease detection in epilepsy.
Area of Science:
- Neuroscience
- Network Science
- Computational Biology
Background:
- Graph theory is widely used to model brain connectome topology.
- Functional connectivity analyses typically assume temporal stationarity, which recent evidence challenges.
- Dynamic fluctuations in brain connectivity are increasingly recognized.
Purpose of the Study:
- To investigate the temporal stationarity of brain network topology.
- To develop and validate novel estimators for quantifying network stationarity.
- To assess the impact of temporal stationarity on discriminating between healthy and diseased brain states.
Main Methods:
- Utilized a Bayesian hidden Markov model (HMM) approach to estimate dynamic graph theoretical measures.
- Developed and applied two novel temporal stationarity estimators: the S-index and N-index.
- Analyzed resting-state functional MRI data from healthy controls and temporal lobe epilepsy patients.
Main Results:
- Identified specific graph theory measures (e.g., small-world index, global integration, betweenness centrality) exhibiting higher temporal stationarity.
- Demonstrated that incorporating subject-level stationarity differences enhances the discrimination power between disease states.
- Confirmed the dynamic nature of functional connectivity and the robustness of certain topological measures.
Conclusions:
- Brain network topology is dynamic, challenging the assumption of stationarity in traditional graph theory analyses.
- Novel stationarity indices (S-index, N-index) offer quantitative insights into temporal dynamics.
- Accounting for dynamic functional connectivity improves sensitivity and consistency in brain network research and clinical applications.
