Continuous evaluation of denoising strategies in resting-state fMRI connectivity using fMRIPrep and Nilearn

• DOI: 10.1371/journal.pcbi.1011942
• Publication Date: 2024-03-18T17:56:32Z
• Authors: Hao-Ting Wang; Steven L. Meisler; Hanad Sharmarke; Natasha Clarke; Nicolas Gensollen; Christopher J. Ma...; François Paugam; Bertrand Thirion; Pierre Bellec
• Keywords: Benchmark (surveying); Communication noise

Reducing contributions from non-neuronal sources is a crucial step in functional magnetic resonance imaging (fMRI) connectivity analyses. Many viable strategies for denoising fMRI are used the literature, and practitioners rely on benchmarks guidance selection of an appropriate choice their study. However, software ever-evolving field, can quickly become obsolete as techniques or implementations change. In this work, we present benchmark featuring range strategies, datasets evaluation metrics analyses, based popular fMRIprep software. The prototypes implementation reproducible framework, where provided Jupyter Book enables readers to reproduce modify figures Neurolibre preprint server (https://neurolibre.org/). We demonstrate how such be continuous research software, by comparing two versions fMRIprep. Most results were consistent with prior literature. Scrubbing, technique which excludes time points excessive motion, combined global signal regression, generally effective at noise removal. Scrubbing was effective, but incompatible statistical analyses requiring sampling brain signal, simpler strategy, using motion parameters, average activity select compartments, preferred. Importantly, found that certain behave inconsistently across and/or fMRIPrep, had different behavior than previously published benchmarks. This work will hopefully provide useful guidelines users community, highlight importance methods.