Skip to content

A short essay for adopting the BIDS structure at uiowa

Jump to bottom
James Kent edited this page Mar 5, 2018 · 2 revisions

BRAIN IMAGING DATA STRUCTURE

BIDS (Brain Imaging Data Structure) is an open standard for organizing imaging data in an intuitive and accessible format for both data consumers (i.e. people who analyze the data/model the data) and data producers (i.e. people who download/curate the data). I’ve been in contact with Jatin Vaidya, Michelle Voss, Joel Bruss, Hans Johnson, and Gene Zeien about BIDS, and they have all expressed interest in adopting the standard.

BENEFITS

Structure encourages systematic documentation about the dataset

  • Why this is important:

    • Imagine a student collects some neuroimaging data, analyzes it, publishes a paper, graduates, and leaves academia. Several years later, the principal investigator assigns another student to analyze the same data with new tools, but they have no idea how the data are named or stored. The original student is not available to contact, and many hours are spent trying to decipher where the files are and not being confident what stage of processing each file is in.

  • How BIDS helps:

    • There are defined locations and filenames with specifications for what types of information should go into each (such as a README at the top level giving a project name, funding source, perhaps information about unusable data, etc.), giving someone who may not be well versed in neuroimaging a guide to know what they need to know. BIDS also specifies information about the scans themselves including important information necessary for field-map corrections, or other preprocessing techniques that require information that is not stored (reliably) in the NIfTI header.

Encourages data processing separation

  • Why this is important:
    • When some researchers (including me, previously) collect their data (in dicom format) and transform the data into NIfTIs; the other preprocessing steps (skull stripping, motion correction, intensity normalization, etc.) may also occur in the same directory, making it difficult to determine what is the original file without any preprocessing applied. This can lead to improper methods of analysis by applying motion correction to an already motion corrected scan.
  • How BIDS helps:
    • BIDS separates the freshly transferred NIfTIs from any subsequent processing. All other processing using the original NIfTIs occurs under a “derivatives” subdirectory, neatly packed into their own processing pipeline directory such as “restingStateAnalysis”. Thus if a researcher wants to share their data with another researcher (within or across) institutions, the original NIfTIs are available (there can be one more processing step to deface the anatomicals before sharing to protect the identity of the subjects).

Makes neuroimaging more accessible

  • Why this is important:
    • Often when learning how to process imaging data, a new trainee has to learn how the data are organized according to the idiosyncrasies of the lab (often with little to no documentation available) and in order to learn how to process imaging data they either pass it off to someone else or blindly hope the custom in-house scripts work for their data. In addition, important information about the data may be included in an haphazard manner (if at all). Thus, to even complete an analysis, the trainee may have to figure out that they need to ask the radiologist to get the necessary information for an already collected dataset. This takes an unnecessarily long time to learn.
  • How BIDS helps:
    • Giving a student a well documented dataset (with the BIDS manual as a reference); they can learn as they go. Software tools also exist that natively recognize the BIDS structure allowing trainees to read about well documented tools for quality assurance (MRIQC) and preprocessing (FMRIPREP) without having to understand all of the code that is used to implement it. With a simple commandline interface, these tools produce informative html files with pictures demonstrating the quality of the data and how successfully they were processed. There is a growing community of developers supporting this standard making it simpler to process data, lowering the barrier of entry for trainees.

WHO USES BIDS?

ADOPTING BIDS AT UIOWA

There are several changes we can adopt to make imaging at the university more BIDS friendly and encourage researchers to be more proactive about their data management plans.

  • Require researchers to submit a pdf to the MRI technicians detailing what the name of each of the scans in the protocol should be named (according to the BIDS standard).
  • Edit the participant intake form so that the MRI technicians name the subjects and sessions in a BIDS compliant manner
  • Construct easy to use software to query xnat to download dicom data in a BIDS friendly manner.
  • Disseminating information about BIDS and having multiple departments to adopt the standard.
    • Presenting a poster about BIDS at the Informatics Showcase
    • I’m setting up monthly meetings with Principal Investigators/Post Docs/Graduate Students that are interested in or doing neuroimaging.
    • I’m presenting at the April INC meeting.
    • With the sponsorship of Hans Johnson, setting up a local brainhack event in May.
Clone this wiki locally