The Brain Imaging Data Structure (BIDS):

an open community standard for neuroscience

Some materials reused from slides by

• Ariel Rokem, The University of Washington
• Rémi Gau, McGill

Yaroslav O. Halchenko

Dartmouth College

http://centerforopenneuroscience.org

bit.ly/bids-nsf-pose-2024

My Goals

• introduce BIDS project
• facilitate adoption of BIDS (as whole) or some ideas or elements
• potentially gain recommendations and contributors
• Make neuroscience a better science

​

​

Brain Imaging Data Structure (BIDS) is …

• a standard for organizing and describing neuroscience data
• a set of tools and services that use this standard
• a community that develops, maintains and extends BIDS

• Researcher A collects a dataset of Magnetic Resonance Imaging data and publishes a paper in the Annals of Anomalous Arcana
• Researcher B requests the data after reading the paper
• The following could happen:
• Researcher A sends researcher B their data
• The data are organized and described in a particular way
• But researcher B does not know/understand how they are organized
• Researcher B gives up and does not follow up on the original finding

Why do we need standards in science?

BIDS Standard Features

• Modular composition�related: YODA principles, github.com/myyoda
• Consistent folder organization
• Expressive file naming schema�related: �“Naming Things” by Jenny Bryan, �“Project structure” by Danielle Navarro,� DataLad RDM Course: How to name a file
• Standard file formats
• Documented Metadata
• Human and Machine readable data
• Human readable specification�(Markdown, HTML, PDF)
• Machine readable schema
• Constant evolution

Figure from: The brain imaging data structure, a format for organizing and describing outputs of neuroimaging experiments

By Gorgolewski et al. Scientific Data, volume 3 (2016) https://doi.org/10.1038/sdata.2016.44

Modularity (explained with pasta)

From: Remi’s slides 10.5281/zenodo.5863713

​

* as expected from reading the method section

Modularity (explained with pasta, frequent reality)

From: Remi’s slides 10.5281/zenodo.5863713

​

Modularity (with BIDS)

MyBidsProject/

├── sourcedata/

│ ├── ingredient-1/

│ ├── ingredient-2/

│ └── raw/

├── derivatives/

│ ├── pasta-1/

│ └── pasta-2/

├── README.md

└── dataset_description.json

​

• scales from a � “data dataset” to “project dataset”
• each module can have the life of its own
• separates source and derived data
• helps reproducibility
• facilitates reuse
• reduces cognitive load

Folder organization

MyBidsProject/

├── sourcedata/

│ └── raw/

│ ├── sub-01/

│ │ ├── [ses-01/]

│ │ │ ├── anat/

│ │ │ ├── func/

│ │ │ ├── beh/

│ │ │ ├── eeg/

│ │ │ └── …

│ │ └── [ses-02/]

│ ├── sub-02/

│ └── dataset_description.json

├── derivatives/

├── README.md

└── dataset_description.json

​

• reduced cognitive load:�“seen one - seen them all”�explore datasets.datalad.org/?dir=/openneuro
• just the right amount of folder nesting
• easy to extend to new data types
• WiP BIDS 2.0: researcher decides �github.com/bids-standard/bids-2-devel/issues/54�“... folders layout to be other than sub-{label}/[ses-{label}/]”

File naming

key1-value_key2-value_suffix.extension

• suffix preceded by an underscore
• entities are composed of key-value pairs separated by underscores
• key and value separated by hyphen
• value could be a label (e.g. sam, 01) or a 0-padded index (e.g., 1, 01)
• keys, labels, and suffixes can only contain letters and / or numbers
• indices can only contain numbers

raw/

├── sub-01/

│ ├── [ses-01/]

│ │ ├── anat/

│ │ └── …

…

“Software X says that this is not a BIDS valid filename.

What is a BIDS way to name files?”

File naming

• sub-*[_ses-*] duplicated
• most entities are �data type specific
• see (per data type) in an �“Entity Table”
• some are REQUIRED
• most are OPTIONAL:�use primarily to disambiguate
• native/standard file formats (also data type specific)

sub-01/

└─ ses-01/

├─ anat/

│ ├─ sub-01_ses-01_T1w.json

│ └─ sub-01_ses-01_T1w.nii.gz

└─ eeg/

├─ sub-01_ses-01_task-olfloc_run-01_eeg.edf

├─ sub-01_ses-01_task-olfloc_run-01_eeg.json

├─ sub-01_ses-01_task-olfloc_run-01_events.tsv

└─ sub-01_ses-01_task-olfloc_run-01_channels.tsv

MRI anatomical (anat/) filename template

sub-<label>[_ses-<label>][_task-<label>][_acq-<label>][_ce-<label>][_rec-<label>][_run-<index>][_part-<mag|phase|real|imag>]_<suffix>.json

sub-<label>[_ses-<label>][_task-<label>][_acq-<label>][_ce-<label>][_rec-<label>][_run-<index>][_part-<mag|phase|real|imag>]_<suffix>.nii[.gz]

​

EEG (eeg/) filename template

sub-<label>[_ses-<label>]_task-<label>[_acq-<label>][_run-<index>]_eeg.<extension>

sub-<label>[_ses-<label>]_task-<label>[_acq-<label>][_run-<index>]_eeg.json

sub-<label>[_ses-<label>]_task-<label>[_acq-<label>][_run-<index>]_events.json

sub-<label>[_ses-<label>]_task-<label>[_acq-<label>][_run-<index>]_events.tsv

File formats (extensions)

• Microscopy
• .ome.tif
• .ome.btf
• .ome.zarr
• .png
• .tif

​

• Near infra-red spectroscopy
• .snirf

​

• …

• MRI
• Volume: .nii[.gz] (NIfTI)
• Surface: .gii[.gz] (GIfTI)

​

• PET
• .nii[.gz] (NIfTI)

​

• EEG / MEG / iEEG
• European data format (.edf)
• BrainVision Core Data Format (.vhdr, .vmrk, .eeg)
• EEGLAB (.set)
• Neurodata Without Borders (.nwb)
• …

BIDS strives to use already existing and used in the community standards, again following 80/20 rule (Pareto principle):

Metadata

• Provided in a “side-car” .json file

​

sub-control01_task-nback_bold.nii - data

sub-control01_task-nback_bold.json - metadata

sub-control01_task-nback_events.tsv - data

sub-control01_task-nback_events.json - metadata

​

• Might be extracted (duplicate) from data file or “supplemental”
• Typically RECOMMENDED and for some REQUIRED

https://bids-specification.readthedocs.io/en/stable/glossary.html lists all metadata fields, which are originally defined in https://github.com/bids-standard/bids-specification/blob/master/src/schema/objects/metadata.yaml

Example:� sub-control01_task-nback_bold.json

{

"TaskName": "N Back",

"RepetitionTime": 0.8,

"EchoTime": 0.03,

"FlipAngle": 78,

"MultibandAccelerationFactor": 4,

"ParallelReductionFactorInPlane": 2,

"PhaseEncodingDirection": "j",

"DeviceSerialNumber": "11035",

}

Common metadata

MyBidsDataset

├─ sub-control01

│ └─ func

│ └─ sub-control01_task-nback_bold.nii

├─ sub-control02

│ └─ func

│ └─ sub-control02_task-nback_bold.nii�└─ task-nback_bold.json

“Don’t repeat yourself” (DRY) principle, In BIDS: Inheritance principle

“I have 40 subjects, with 2 sessions and 4 runs per subject per session.

I have 320 JSON files to edit? HELP!”

MyBidsDataset

├─ sub-control01

│ └─ func

│ ├─ sub-control01_task-nback_bold.json

│ └─ sub-control01_task-nback_bold.nii

└─ sub-control02

└─ func

├─ sub-control02_task-nback_bold.json

└─ sub-control02_task-nback_bold.nii

Tabular (meta)data - TSV files

• .tsv with (meta)data, e.g. participants.tsv:

​

​

​

​

​

• complemented with .json to describe fields

MyBidsDataset

├─ sub-control01/

│ └─ func/

│ └─ sub-control01_task-nback_bold.nii

├─ sub-control02/

│ └─ func/

│ └─ sub-control02_task-nback_bold.nii

├─ participants.tsv

└─ participants.json

participant_id\tage\tsex

sub-control01\t31\tM

sub-control02\t25\tM

{

"participant_id": {

"LongName": "patient identification",

"Description": "anonymized, BIDS compatible identification",

"Annotations": {

"IsAbout": {

"TermURL": "nb:ParticipantID",

"Label": ""

},

"Identifies": "participant"

}

},

"sex": {

"Levels": {

"M": "Male",

"F": "Female"

},

...

}

added by NeuroBagel

Standards beyond file formats: BIDS is a meta-standard

Figure from: BIDS Starter Kit https://bids-standard.github.io/bids-starter-kit/#philosophy

Why do we need standards in science?

• Researcher C develops a new method for analyzing neuroimaging data and publishes it in Memoranda of Magical Methods
• Researcher D would like to use the method to analyze their own data
• The following could happen:
• Researcher D downloads C’s code and tries to run it
• But the code expects the data to be organized in a particular way
• The code requires knowledge of certain acquisition parameters
• Those are no longer available to researcher D, since they did not save those parameters
• The software crashes in ways that researcher D doesn’t understand and they give up

The FAIR principles for data management and stewardship

Findable

Accessible

Interoperable

Reusable

The FAIR Guiding Principles for scientific data management and stewardship. Scientific Data. Wilkinson et al. (2016)

FAIR Data archives embrace (and help develop) BIDS

The OpenNeuro resource for sharing of neuroscience data. eLife. Markiewicz et al. (2022)

https://doi.org/10.7554/eLife.71774

BIDS enables uniform sharing of a large amount of data

It is estimated that measurements from more than 200,000 individuals are currently stored in BIDS format in research centers and data archives

From: The past, present, and future of the brain imaging data structure (BIDS)

By Poldrack et al. Imaging Neuroscience (2024) 2: 1–19. https://doi.org/10.1162/imag_a_00103

https://identifiers.org/DANDI:000108

The ABCD Study

>11,000 subjects, measured at 3 time-points (and 2 more time-points to come!)

The ABCD Community Collection provides BIDS-compliant data and processing derivatives

Not created at once!�BIDS is demand-driven/self-motivating

Not created at once!

80/20. Not only for MRI!

From: The past, present, and future of the brain imaging data structure (BIDS)

By Poldrack et al. Imaging Neuroscience (2024) 2: 1–19. https://doi.org/10.1162/imag_a_00103

BIDS Extension Proposals

Currently in the specifications

BIDS Extension Proposals

Ongoing BEPs

BIDS Extension Proposals: Funding?

Multiple tools facilitate using BIDS

https://bids.neuroimaging.io/benefits.html includes list of converters and various tools

An eco-system of standards-compliant software

More on apps in BIDS apps: Improving ease of use, accessibility, and reproducibility of neuroimaging data analysis methods.

By Gorgolewski et al. PLoS Comput Biol (2017). https://doi.org/10.1371/journal.pcbi.1005209

BIDS Apps and Modularity: YODA

From https://github.com/ReproNim/containers

BIDS-Apps 2.0

Governance

https://bids.neuroimaging.io/governance.html

Governance: a democratically elected steering group

Current members:

• Ariel Rokem, The University of Washington, USA
• Yaroslav O. Halchenko, Dartmouth College, USA
• Cyril Pernet, Copenhagen University Hospital, Denmark
• Dora Hermes, Mayo Clinic, USA
• Camille Maumet, INRIA, France

• Responsible for approving and advancing BEPs
• Provides overall direction for the community
• Helps facilitate decision making
• Elected by members of the community
• Serve for two years in non-overlapping terms

Governance

Volunteer BIDS Maintainers Group

• Researchers and Research Software Engineers

​

• A range of different activities:
• Development and maintenance of the specification
• Surrounding tools
• Extension proposals
• Documentation
• Community management
• …

Governance

https://bids-specification.readthedocs.io/en/stable/appendices/contributors.html

>300 contributors from many countries and institutions

Governance

>10,000 unique visitors to the BIDS specification website in the first half of 2023

​

Summary: Brain Imaging Data Structure (BIDS) is …

• a standard for organizing and describing neuroscience data with
• modular composition
• consistent folder organization
• expressive file naming schema
• standard file formats
• documented metadata
• human and machine readable data
• supported by a wide range of FAIR resources
• constantly evolving through BEPs
• a set of tools and services that use this standard
• a community that develops, maintains and extends BIDS