marp | title | theme | size | paginate | _paginate | footer |
---|---|---|---|---|---|---|
true |
A primer to BIDS |
default |
58140 |
true |
false |
Ghislain Vaillant | A primer to BIDS | Inria Digital Health Meetup 2022 |
A 10-minute (or so) introduction to the Brain Imaging Data Structure for busy engineers.
- Understand what BIDS is
- Introduce the core concepts
- Overview of available tools
A simple and intuitive way to organize and describe your neuroimaging and behavioral data.
The Brain Imaging Data Structure (BIDS) is a standard specifying the description of neuroimaging data in a filesystem hierarchy and of the metadata associated with the imaging data.
BIDS is a community-driven effort.
The specification is opensource and contributions are accepted through submission of issues and requests to the GitHub repository.
Governance is driven by a formal leadership structure, which ensures strict application of contribution policies and the code of conduct.
Requests for improvement are handled through BIDS Extension Proposals (BEP).
BEPs are incubated within special interest groups, prior to proposal for inclusion to the specification.
Once accepted, the BEP is officially included in the next release of the specification.
Notable BEPs: EEG, MEG, PET, Microscopy WIP: DWI derivatives, PET derivatives, Provenance
- Datasets
- File collections
- Metadata
- Source, raw and derivative datasets
- Minimal requirements
- Storage convention
Raw dataset
- A BIDS-compliant dataset containing the raw data of the study.
Source dataset
- A non BIDS-compliant dataset containing the source data of the study before normalisation to BIDS.
Derivative dataset
- A BIDS-like dataset containing artifacts processed from a raw dataset and other derivatives possibly.
{
"Name": "This dataset",
"BIDSVersion": "1.8.0",
"DatasetType": "raw" | "derivative",
}
Embedded (in-tree)
rawdata/
dataset_description.json
derivatives/
pipeline-foo/
dataset_description.json
pipeline-bar/
dataset_description.json
sourcedata/
Standalone (out-of-tree)
sourcedata/
rawdata/
dataset_description.json
derivatives/
pipeline-foo/
dataset_description.json
pipeline-bar/
dataset_description.json
- File naming convention
- Components and entities
- Derivatives
sub-P01/
ses-M00/
anat/
sub-P01_ses-M00_T1w.nii.gz
sub-P01/
ses-M00/
anat/
sub-P01_ses-M00_run-01_T1w.nii.gz
sub-P01_ses-M00_run-02_T1w.nii.gz
Entities: sub
, ses
, run
Datatype: anat
Suffix: T1w
Extension: .nii.gz
Supported entities by modality.
sub-P01/
ses-M00/
anat/
sub-P01_ses-M00_desc-preproc_T1w.nii.gz
Source entities: sub
, ses
Derivative entities: desc
BIDS derivatives are still a work-in-progress. Consider checking for in-progress BEP first, before reaching for your own implementation.
- Modality-agnostic metadata
- Modality-specific metadata
- Inheritence principle
Modality-agnostic metadata are defined in tabular format and serialised to TSV.
participants.tsv # participant-level metadata
sub-P01/
sub-P01_sessions.tsv # session-level metadata
ses-M00/
sub-P01_ses-M00_scans.tsv # scan-level metadata
Example: participants.tsv
participant_id | age | sex | handedness | group |
---|---|---|---|---|
sub-P01 | 36 | M | n/a | control |
Modality-specific metadata are defined in dictionary format and serialised to JSON.
sub-P01/
ses-M00/
anat/
sub-P01_ses-M00_T1w.nii.gz
sub-P01_ses-M00_T1w.json # Modality-specific metadata
Example sidecar JSON metadata:
{
"MagneticFieldStrength": 3, /* DICOM tag 0018, 0087 */
"MRAcquisitionType": "3D", /* DICOM tag 0018, 0023 */
"EchoTime": 0.00298 /* DICOM tag 0018, 0081 */
}
T1w.json # Dataset-wide definition
sub-P01/
sub-P01_T1w.json # Participant-specific overrides
ses-M00/
anat/
sub-P01_T1w.nii.gz
sub-P01_T1w.json # Modality-specific overrides
Modality-specific metadata computed by dictionary merging from root to leaf.
- Publicly available datasets
- Official tools and libraries
- Starter kit
- dcm2niix: DICOM to NIfTI converter
- HeuDiConv: heuristic-centric DICOM converter
- Clinica: software platform for clinical neuroimaging studies
Thank you for your attention.