Brain Imaging Data Structure (BIDS)

Research Data Management for Psychology and Neuroscience
Course at Julius-Maximilians-Universität Würzburg, RTG 2660: Approach-Avoidance
Slides | Source

Dr. Lennart Wittkuhn

lennart.wittkuhn@tutanota.com

14:30

Agenda

• 1 Introduction to BIDS
• 2 BIDS Principles
• 3 BIDS Directory Structure
• 4 Data Types and Modalities
• 5 Metadata and Documentation
• 6 BIDS Ecosystem
• 7 Getting Started with BIDS
• 8 Examples
• 9 Extensions and Future
• 10 Summary
• 11 Exercises

Schedule

Day	Date	Time	Title
1	2026-05-07	09:00 - 09:30	Welcome and Introduction to Research Data Management
1	2026-05-07	09:30 - 10:30	Project & Data Organization
1	2026-05-07	10:30 - 11:30	Data Management Plans (DMPs)
1	2026-05-07	11:30 - 12:30	Lunch Break
1	2026-05-07	12:30 - 13:30	Command Line
1	2026-05-07	13:30 - 14:30	Best practices for rectangular data
1	2026-05-07	14:30 - 16:00	Brain Imaging Data Structure (BIDS)

This session: Brain Imaging Data Structure (BIDS)

Image taken from Gorgolewski et al. (2016)

Objectives

💡 You understand what BIDS (Brain Imaging Data Structure) is and why it’s important for neuroimaging.
💡 You can explain the core principles of BIDS and how it solves common data organization problems.
💡 You can organize neuroimaging data according to BIDS directory structure standards.
💡 You understand the role of JSON metadata files and TSV data files in BIDS datasets.
💡 You know how to validate BIDS datasets using the BIDS validator.
💡 You understand the benefits of using BIDS for collaboration, reproducibility, and data sharing.

1 Introduction to BIDS

What is BIDS?

The Brain Imaging Data Structure (BIDS) is a standard to organize and describe neuroimaging and behavioral data.

Image taken from Gorgolewski et al. (2016)

Key characteristics:

• Standardized file organization and naming conventions
• Community-driven specification
• Modality-agnostic (fMRI, EEG, MEG, PET, behavior)
• Human and machine readable
• Growing ecosystem of compatible tools

The Problem BIDS Solves

Before BIDS 😞

• Every lab organized data differently
• Time wasted understanding data structure
• Difficult collaboration and data sharing
• Scripts broke when data moved
• No common metadata standards

With BIDS 😊

• Standard organization across labs
• Immediate understanding of data structure
• Easy collaboration and sharing
• Tool interoperability
• Structured metadata

Standards …

… but probably different with BIDS?

XKCD: Standards

2 BIDS Principles

Core BIDS Principles

1. Simplicity: Based on common file formats (NIfTI, JSON, TSV)
2. Consistency: Standardized naming and folder structure
3. Flexibility: Supports multiple neuroimaging modalities
4. Extensibility: Community-driven extensions (BEPs)
5. Validation: Tools to check compliance
6. Documentation: Self-documenting through metadata files

File Naming Convention

BIDS uses entities to create standardized file names:

sub-<label>[_ses-<label>][_task-<label>][_acq-<label>][_run-<label>]_<suffix>.<ext>

Examples:

• sub-01_task-rest_bold.nii.gz (fMRI resting-state)
• sub-02_ses-001_task-faces_run-1_bold.nii.gz (task fMRI)
• sub-01_T1w.nii.gz (structural MRI)
• sub-03_task-nback_eeg.bdf (EEG data)

3 BIDS Directory Structure

Basic BIDS Dataset Structure

my-dataset/
├── dataset_description.json
├── README
├── CHANGES
├── participants.tsv
├── participants.json
└── 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
        └── sub-01_task-rest_events.tsv

Required Files

Every BIDS dataset must include:

1. dataset_description.json - Dataset metadata (name, version, authors)
2. README - Human-readable description
3. At least one subject folder (sub-<label>/)
4. Proper file naming following BIDS conventions

Optional but Recommended

• participants.tsv - Subject demographics and characteristics
• participants.json - Metadata for participants file
• CHANGES - Version history
• LICENSE - Data usage license
• code/ - Analysis scripts
• derivatives/ - Processed data

4 Data Types and Modalities

Supported Neuroimaging Modalities

Core modalities:

• Anatomical MRI (anat/)
• Functional MRI (func/)
• Diffusion MRI (dwi/)
• Fieldmaps (fmap/)

Electrophysiology:

• EEG (eeg/)
• MEG (meg/)
• iEEG (ieeg/)

Other modalities: PET, microscopy, behavioral data, genetics, physiological recordings, and more!

Example: fMRI Task Data

sub-01/
└── func/
    ├── sub-01_task-faces_bold.nii.gz        # Brain imaging data
    ├── sub-01_task-faces_bold.json          # Imaging parameters
    ├── sub-01_task-faces_events.tsv         # Task events/stimuli
    └── sub-01_task-faces_events.json        # Events metadata

Key files:

• .nii.gz - Brain imaging data
• .json - Technical metadata (TR, TE, slice timing, etc.)
• _events.tsv - Task timing and conditions

5 Metadata and Documentation

JSON Metadata Files

BIDS uses JSON files to store technical parameters and metadata:

{
  "TaskName": "face perception",
  "RepetitionTime": 2.0,
  "EchoTime": 0.03,
  "FlipAngle": 90,
  "SliceTiming": [0, 0.5, 1.0, 1.5],
  "TaskDescription": "Participants viewed faces and objects"
}

Also see section on “Data Dictionaries” in session on “Rectangular Data”

TSV Files for Tabular Data

Tab-separated values for events, participants, and other tabular data:

onset   duration  trial_type    response_time
0.0 2.5 face  1.2
3.0 2.5 object  0.9
6.0 2.5 face    1.1

• Machine-readable for analysis
• Human-readable for inspection
• Version-control friendly (plain text)

6 BIDS Ecosystem

BIDS-Compatible Tools

Data organization:

• Heudiconv
• BIDScoin
• dcm2bids

Preprocessing:

• fMRIPrep
• MRIQC
• QSIPrep

Analysis:

• Nilearn
• SPM12
• FSL
• AFNI

Data sharing:

• OpenNeuro
• BIDS Validator

BIDS Validator

Automatic validation of BIDS compliance:

• Web version: bids-standard.github.io/bids-validator/
• Command-line tool: bids-validator /path/to/dataset
• Checks: File naming, required files, metadata consistency

Benefits:

• Catch errors early
• Ensure compatibility with BIDS tools
• Prepare for data sharing

7 Getting Started with BIDS

Steps to BIDSify Your Data

1. Plan ahead: Consider BIDS during data collection
2. Learn the specification: Read the BIDS specification
3. Use conversion tools: dcm2bids, Heudiconv, etc.
4. Validate: Use BIDS validator regularly
5. Document: Create comprehensive README and metadata
6. Share: Consider uploading to OpenNeuro

BIDS Starter Kit

The BIDS Starter Kit provides:

• Tutorials for getting started
• Templates for common files
• Best practices and examples
• Community support and resources

Benefits of Using BIDS

• Collaboration: Easier to share and understand data
• Reproducibility: Standardized organization aids replication
• Tool compatibility: Access to BIDS-aware analysis pipelines
• Data sharing: Ready for repositories like OpenNeuro
• Future-proofing: Organized data survives lab transitions
• Quality control: Built-in validation catches errors

8 Examples

Example Dataset: Task fMRI Study

bids-dataset/
├── dataset_description.json
├── README
├── participants.tsv
├── participants.json
├── task-faces_bold.json          # Task-level metadata
├── sub-01/
│   ├── anat/
│   │   ├── sub-01_T1w.nii.gz
│   │   └── sub-01_T1w.json
│   └── func/
│       ├── sub-01_task-faces_run-1_bold.nii.gz
│       ├── sub-01_task-faces_run-1_bold.json
│       ├── sub-01_task-faces_run-1_events.tsv
│       ├── sub-01_task-faces_run-2_bold.nii.gz
│       ├── sub-01_task-faces_run-2_bold.json
│       └── sub-01_task-faces_run-2_events.tsv
└── sub-02/
    └── ... (similar structure)

Common Challenges and Solutions

Challenge: Complex naming

• Solution: Use BIDS tools for conversion
• Start with simple cases first

Challenge: Legacy data

• Solution: Gradual conversion
• Document any deviations

Challenge: Multiple sessions

• Solution: Use session folders (ses-<label>)
• Consistent session naming

Challenge: Non-standard acquisitions

• Solution: Use acquisition entity (acq-<label>)
• Document in README

9 Extensions and Future

BIDS Extension Proposals (BEPs)

BIDS evolves through community-driven extensions:

Recent additions:

• EEG/MEG data organization
• PET imaging protocols
• Microscopy data standards
• Motion data capture
• Genetics information

The Future of BIDS

• Broader adoption across neuroscience
• New modalities being standardized
• Enhanced metadata schemas
• Integration with analysis platforms
• Machine learning dataset preparation

10 Summary

Key Takeaways

1. BIDS standardizes neuroimaging data organization
2. Benefits everyone: researchers, collaborators, future you
3. Ecosystem growth: tools, databases, and community support
4. Start now: Plan BIDS organization during data collection
5. Resources available: Specification, starter kit, validator, community
6. Future-proof your research data management

Next Steps

For your next neuroimaging project:

1. Read the BIDS specification
2. Explore the BIDS Starter Kit
3. Try the BIDS validator
4. Join the BIDS community
5. Consider your data collection protocol

Resources

• BIDS Specification: bids-specification.readthedocs.io
• BIDS Starter Kit: bids-standard.github.io/bids-starter-kit/
• BIDS Validator: bids-standard.github.io/bids-validator/
• OpenNeuro: openneuro.org
• Community Forum: neurostars.org/tag/bids

References

• Gorgolewski, K.J., Auer, T., Calhoun, V.D., et al. (2016). The brain imaging data structure, a format for organizing and describing outputs of neuroimaging experiments. Scientific Data, 3, 160044. doi:10.1038/sdata.2016.44

11 Exercises

Exercise 1

Exercise 1: Explore BIDS datasets on OpenNeuro

Objective: Familiarize yourself with BIDS-formatted datasets and understand how standardized organization facilitates data sharing and reuse.

Instructions:

1. Navigate to OpenNeuro:

   • Visit https://openneuro.org/
   • Browse the available datasets on the homepage

2. Select and explore a dataset:

   • Choose one dataset that interests you (e.g., neuroimaging study related to your research area)
   • Click on the dataset to view its details

3. Examine the BIDS structure:

   • Click “Browse” or “Files” to view the dataset’s file structure
   • Explore the following directories (if present):
     • Root level files (dataset_description.json, README, participants.tsv)
     • Subject folders (sub-01, sub-02, etc.)
     • Session folders (if applicable: ses-01, ses-baseline, etc.)
     • Modality folders (anat, func, dwi, fmap, etc.)

4. Explore the dataset:

   • What type of neuroimaging data does it contain?
   • How many subjects are included?
   • What file naming patterns do you observe?
   • Are there any derivative datasets associated with it?

5. Compare with raw data organization:

   • Think about how this standardized structure compares to typical “raw” data organization
   • What advantages does this standardization provide for:
     • Data sharing?
     • Automated processing?
     • Reproducibility?

Discussion points:

• How does the BIDS structure make it easier to understand the dataset without detailed documentation?
• What challenges might researchers face when converting their data to BIDS format?

Exercise 2

Exercise 2: Hands-on BIDS practice

Objective: Learn how to create a BIDS dataset using heudiconv and use BIDS validation.

Access the hands-on tutorial: https://lennartwittkuhn.com/rdm-bids/

Questions & Discussion

Discussion time:

• Have you encountered challenges organizing neuroimaging data?
• Which BIDS modalities are relevant to your research?
• What barriers do you see to adopting BIDS?
• How might BIDS improve your workflow?

References

Gorgolewski, K. J., Auer, T., Calhoun, V. D., Craddock, R. C., Das, S., Duff, E. P., Flandin, G., Ghosh, S. S., Glatard, T., Halchenko, Y. O., Handwerker, D. A., Hanke, M., Keator, D., Li, X., Michael, Z., Maumet, C., Nichols, B. N., Nichols, T. E., Pellman, J., … Poldrack, R. A. (2016). The brain imaging data structure, a format for organizing and describing outputs of neuroimaging experiments. Scientific Data, 3(1). https://doi.org/10.1038/sdata.2016.44.