The Comet Toolbox: Improving Robustness in Network Neuroscience Through Multiverse Analysis

Abstract

In network neuroscience, a broad range of methods for estimating dynamic functional connectivity from fMRI data and subsequent analyses using graph-theoretic approaches have been introduced in recent years. However, in the absence of ground truths regarding the validity of analytical steps in capturing true brain dynamics, researchers are often faced with a multitude of arbitrary, yet defensible, analytical choices, raising concerns regarding the robustness of results. Here, we aim to address this issue by implementing a comprehensive suite of dynamic functional connectivity methods in a unified Python software package, allowing for a diverse exploration of brain dynamics. Anchored in the framework of multiverse analysis, the present work introduces a workflow for systematically exploring different methodological choices. The developed toolbox includes a graphical user interface for ease of use and accessibility for those who wish to operate outside of a script-based pipeline. Comprehensive documentation and demo scripts are included to support adoption and usability. By promoting transparency and robustness, Comet aims to promote best practice in the study of brain dynamics.