Data Acquisition Considerations for More Reproducible Neuroimaging Research

• Adopt standards-based data representation from the get go.
• Use ReproIn to automate conversion of your imaging data to BIDS.
• Annotate your metadata (all experimental details) as you collect it.
  • There are a number of ways to annotate your data. The idea with annotations are to provide unambigous markup of your data, whether it be imaging, behavioral, clinical, etc. One strategy is to annotate the variables you intend to collect during the study design process. This can occur by choosing data structures that are publically available and already annotated (e.g. NDAR) or by using the tools below. Regardless of the approach, the important aspect is to unambigously describe what the variables collected mean, their ranges, min/max, units, etc. For more information on the desired properties for annotations, see: Federated Data Elements
    • BrainVerse is a tool being developed which can help to annotate your variables by directly incorporating data structures from NDAR.
    • PyNIDM is both a Python library for creating metadata documents using the Neuroimaging Data Model (NIDM) but also contains tools such as "bidsmri2nidm" and "csv2nidm" which can be used to interactively annotate both BIDS datasets (via participants.csv file) and general CSV-formatted data files containinga header row of variables which require annotation. The annotation process requires you to have a SciCrunch account and API-access key which are freely available on the site. The tools iteratve over your variables and query the InterLex information resource looking for terms that match your variables. If no term is found you are asked to contribute a new term definition which will be used in the InterLex and thus provide persistent annotations.
    • Use a version control system for all files (data, code, environments, etc.) - i.e. DataLad.