A reproducible and generalizable software workflow for analysis of large-scale neuroimaging data collections using...

Area of Science:

• Neuroimaging
• Computational Neuroscience
• Data Science

Background:

• Neuroimaging research faces reproducibility challenges due to large datasets and complex processing.
• Brain Imaging Data Structure (BIDS) Apps improve data organization but require full audit trails for reproducibility.
• Existing frameworks for reproducible large-scale data processing are often proofs-of-concept and lack generalizability.

Purpose of the Study:

• Introduce the BIDS App Bootstrap (BABS), a user-friendly Python package for scalable and reproducible neuroimaging data processing.
• Address the need for a generalizable tool to manage audit trails for large-scale neuroimaging datasets.
• Facilitate the application of BIDS Apps to large datasets while ensuring full data processing traceability.

Main Methods:

• Leverage DataLad for data management and version control.
• Integrate the FAIRly big framework for reproducible processing.
• Automate script preparation for data processing and version tracking on High Performance Computing (HPC) systems.
• Support job submission and auditing on Sun Grid Engine (SGE) and Slurm HPC environments.

Main Results:

• Demonstrated the scalability of BABS by applying it to the Healthy Brain Network (HBN) dataset (n = 2,565).
• Successfully tracked the full audit trail of data processing in a scalable manner.
• Provided a user-friendly and generalizable solution for reproducible neuroimaging analysis.

Conclusions:

• BABS enables reproducible and scalable neuroimaging data processing.
• The package enhances research reproducibility by ensuring full audit trails.
• BABS is broadly extensible through an open-source development model, promoting wider adoption and contribution.