Replicability of Functional Brain Networks: A Study Through the Lens of Seven Resting-State Networks

The study of brain networks is essential for improving our understanding of how the human brain functions. Functional connectivity (FC) analysis is a widely used approach for studying co-activating patterns among brain regions by estimating their temporal dependencies and constructing an undirected network. Data processing is critical before estimating a subject's functional network, but the absence of a standardized procedure serves as a source of heterogeneity in results, especially in multi-site studies. Commonly studied functional networks include the default mode, sensorimotor, visual, salience, dorsal attention, frontoparietal, and language networks. These networks are stable and still exhibit intrinsic activation when an individual is at rest, making them ideal networks to focus on for studying how processing choices affect the replicability of functional connectivity networks. We use the aforementioned seven networks to assess the impact of various processing choices, including preprocessing pipeline, band-pass filtering, and brain parcellation, on the replicability of functional connectivity estimates for multi-site resting-state fMRI (rs-fMRI) data from the Autism Brain Imaging Data Exchange (ABIDE). Finally, we provide some practical recommendations for how researchers should proceed with processing choices in the face of these effects.