versaFlow: a versatile pipeline for resolution adapted diffusion MRI processing and its application to studying the variability of the PRIME-DE database

• DOI: 10.3389/fninf.2023.1191200
• Publication Date: 2023-08-11T00:49:49Z
• Authors: Alex Valcourt Caron; Amir Shmuel; Ziqi Hao; Maxime Descoteaux
• Keywords: Robustness

The lack of “gold standards” in Diffusion Weighted Imaging (DWI) makes validation cumbersome. To tackle this task, studies use translational analysis where results humans are benchmarked against findings other species. Non-Human Primates (NHP) particularly interesting for this, as their cytoarchitecture is closely related to humans. However, tools used processing and must be adapted finely tuned work well on NHP images. Here, we propose versaFlow, a modular pipeline implemented Nextflow, designed robustness scalability. tailored vivo DWI at any spatial resolution; it allows maintainability customization. Processes workflows using cutting-edge state-of-the-art Magnetic Resonance (MRI) technologies diffusion modeling algorithms, namely Tensor (DTI), Constrained Spherical Deconvolution (CSD), DIstribution Anisotropic MicrOstructural eNvironments Diffusion-compartment imaging (DIAMOND). Using provide an in-depth study the variability metrics computed 32 subjects from 3 sites Primate Data Exchange (PRIME-DE), which contains anatomical T1-weighted (T1w) T2-weighted (T2w) images, functional MRI (fMRI), brains. This dataset includes images acquired over range resolutions, single multi-shell gradient samplings, multiple scanner vendors. We perform reproducibility versaFlow Aix-Marseilles site's data, ensure that our implementation has minimal impact observed subsequent analyses. report very high majority metrics; only gamma distribution parameters DIAMOND display less reproducible behaviors, due absence mechanism enforce random number seed software used. should taken into consideration when future applications performed. show PRIME-DE data exhibits great level variability, similar or greater than obtained human studies. Its usage done carefully prevent instilling uncertainty statistical hints need sufficient harmonization acquisition protocols development robust algorithms capable managing induced differences models and/or