[MRIQC 2] Brain Imaging Data Structure (BIDS)

20 May 2024     #bio #brainImaging

Brain Imaging Data Structure (BIDS)

The Brain Imaging Data Structure (BIDS) was created to streamline the organization and sharing of neuroimaging and behavioral data. The driving force behind BIDS is the need for a standardized format in neuroimaging research to prevent misunderstandings, eliminate the time spent on data reorganization, and improve reproducibility. By offering a straightforward and intuitive structure for data, BIDS aims to promote collaboration, speed up research, and make neuroimaging data more accessible to a diverse range of scientists.

BIDS provides detailed guidelines on how to format and name files, ensuring consistency across studies. It supports various neuroimaging modalities, including MRI, MEG, EEG, and iEEG, and is extensible, allowing for the integration of new data types and metadata. Additionally, BIDS is supported by a growing ecosystem of tools and software that facilitate data validation, analysis, and sharing, further enhancing its utility in the research community.

BIDS Format

BIDS was inspired by the format used by the OpenfMRI repository, which is now known as OpenNeuro. The BIDS format is essentially a method for organizing data and metadata within a hierarchical folder structure. It makes minimal assumptions about the tools needed to interact with the data, allowing for flexibility and broad compatibility. This structure helps standardize data organization, facilitating easier data sharing, analysis, and collaboration within the neuroimaging research community.

Folders

There are four levels of the folder hierarchy, and all sub-folders except for the root folder have a specific structure to their name. The format and the example names can be described as:

Project/
└─ Subject (e.g. 'sub-01/')
  └─ Session (e.g. 'ses-01/')
    └─ Datatype (e.g. 'anat')

• Project: contains all the dataset. can have any name.
• Subject: contains data of one subject. One folder per subject. A subject has a unique label.
  • Name format: sub-PARTICIPANT LABEL

• Session: represents a recording session. Each subject may have multiple sessions if the data is gathered from several occasions.

  • If there’s only a single session per subject, this level may be omitted.
  • Name format: ses-SESSION LABEL

• Datatype: represents different types of data.

  

  • anat: anatomical MRI data
  • func: functional MRI data
  • fmap: fieldmap data
  • dwi: diffusion MRI data
  • perf: arterial spin labeling data
  • eeg: electroencephalography data
  • meg: magnetoencephalography data
  • ieeg: intracranial EEG data
  • beh: behavioral data
  • pet: positron emission tomography data
  • micr: microscopy data
  • nirs: near-infrared spectroscopy data
  • motion: motion capture data

Files

Three main types of files:

• .json file: contains metadata
• .tsv file: contains tables of metadata
• Raw data files: e.g. .jpg, .nii.gz

Standardized ways of naming files:

• Do not include white spaces in file names
• Use only letters, numbers, hyphens, and underscores.
• Do not rely on letter case: Some operating systems regard a as the same as A
• Use separators and case in a systematic and meaningful way.
  • CamelCase or snake_case

Filename template

Example:

Dataset/
 └─ participants.json
 └─ participants.tsv
 └─ sub-01/
   └─ anat/
     └─ sub-01_t1w.nii.gz
     └─ sub-01_t1w.json
   └─ func/
     └─ sub-01_task-rest_bold.nii.gz
     └─ sub-01_task-rest_bold.json
   └─ dwi/
     └─ sub-01-task-rest_dwi.nii.gz

Reference

• BIDS Official Website
• BIDS Starter Kit
• BIDS Github