The fNIRS Reproducibility Study Hub (FRESH): Exploring Variability and Enhancing Transparency in fNIRS Neuroimaging Research

In neuroimaging research, efforts to enhance replication and reproducibility have increased the focus on improving transparency, particularly in the complex data analysis processes. We conducted a multi-lab collaborative study involving 38 international teams that analyzed two functional Near-Infrared Spectroscopy (fNIRS) datasets. These teams tested seven group-level and forty individual-level hypotheses, and they submitted detailed reports on their analysis pipelines and testing outcomes. The results showed significant variability in hypothesis testing outcomes due to differences in analytical approaches. There was greater consensus in group-level analyses compared to individual-level analyses. Factors such as the pruning method, hemodynamic response function model and estimation, and statistical analysis space partly account for the variability in hypothesis testing outcomes. Additionally, we have found higher similarity in hypothesis testing outcomes across the researchers who reported higher confidence in their analysis skills. This study underscores the importance of complying with best practices in fNIRS analysis methodologies and the need for standardized analysis protocols to improve reliability and credibility.