Time-varying functional connectivity as Wishart processes.

Area of Science:

• Neuroimaging
• Computational Neuroscience
• Statistical Modeling

Background:

• Functional magnetic resonance imaging (fMRI) measures brain activity.
• Time-varying functional connectivity (TVFC) captures dynamic brain region interactions.
• Estimating TVFC is crucial for understanding brain function.

Purpose of the Study:

• To evaluate the utility of Wishart processes (WPs) for TVFC estimation in fMRI.
• To benchmark WP performance against established TVFC methods.
• To assess WP's practical viability for neuroimaging applications.

Main Methods:

• Developed a comprehensive benchmarking framework for TVFC methods.
• Included simulations, phenotype prediction, test-retest, brain state analysis, and stimulus prediction tasks.
• Utilized scalable approximate inference and open-source libraries for WP implementation.

Main Results:

• Wishart processes (WPs) demonstrated competitive performance across all benchmarks.
• WPs outperformed sliding window and DCC-MGARCH methods in external stimulus prediction.
• WPs showed reduced susceptibility to false positives in TVFC null models.

Conclusions:

• Wishart processes are a viable and effective tool for estimating time-varying functional connectivity from fMRI data.
• WPs offer advantages over traditional methods, particularly in predictive modeling and statistical robustness.
• The study validates WPs as a powerful approach for dynamic brain connectivity analysis.