The document discusses the Brain Imaging Data Structure (BIDS), a standardized method for structuring neuroimaging data which addresses metadata retention and improves data interoperability. BIDS has expanded to include various neuroimaging modalities, including fNIRS, and is designed to facilitate analysis, sharing, and reproducibility in research. The document emphasizes the importance of following BIDS guidelines for effective data organization and management in fNIRS research.

1. The Brain Imaging Data Structure
and its use for fNIRS research
--- to be presented at the NIRS toolkit at the Donders ---
Robert Oostenveld
r.oostenveld@donders.ru.nl

2. What is the Brain Imaging Data Structure
 http://bids.neuroimaging.org
 https://bids-specification.readthedocs.io
 Not a file format, but a way of structuring your data and providing metadata
– It specifies which file formats are to be used (i.e. NIfTI, json, tsv)
– It specifies the naming convention for files and directories
– It addresses the problem of metadata getting lost while doing your research
– Details from DICOM headers are lost in converting DICOM to NIfTI
– Details about the participants that are not documented digitally
– Details about the cognitive task an/or experimental manipulations

3. https://xkcd.com/927/
Rather than as a “standard” (see below),
you should consider it a widely supported “best practice”

4. Brain Imaging Data Structure
 BIDS was initiated at Stanford to address the challenges of OpenFMRI.org (now OpenNeuro.org)
 Main focus is on “brain” imaging, i.e. neuroimaging, which is often done in cognitive neuroscience
 It not only focusses on the imaging aspect
– Also on the cognitive part of the research (i.e. behavior)
– Also on other measures of activity (i.e. physiology)
 In 2016 BIDS was specified for MRI and fMRI, in 2018 also MEG, in 2019 also for EEG and iEEG,
and more extensions are in the works
 BIDS is used by default at the DCCN for most (f)MRI data and most MEG data
 It is a style of research data management (RDM) that you can also use elsewhere
– … for modalities that are not in the specification (fNIRS, eyetracker, video, motion capture)
– … for data that is not from brain imaging experiments

5. An example of raw data with MRI, MEG and EEG
data/README
CHANGES
dataset_description.json
participants.tsv
/sub-01/anat/…
/sub-01/meg/…
/sub-01/eeg/sub-01_task-auditory_eeg.edf
/sub-01/eeg/sub-01_task-auditory_eeg.json
/sub-01/eeg/sub-01_task-auditory_channels.tsv
/sub-01/eeg/sub-01_task-auditory_events.tsv
/sub-01/eeg/sub-01_electrodes.tsv
/sub-01/eeg/sub-01_coordinates.json
Actual EEG data
Directory
structure
Metadata

6. Project background
 BIDS is the result of the ongoing collaboration of people like you!
 BIDS consists of the specification, of tools, and of the communication platforms
 The specification is the result of sharing knowledge, discussion, and consensus
– Email discussion list
– Shared Google docs
– GitHub
 Well-specified governance structure
 BIDS extensions projects are called “BEPs” and have a leader
 Steering group
– Guimar Niso, Russ Poldrack, Kirstie Whitacker, Melanie Ganz, Robert Oostenveld

7. BIDS Extension Proposals, i.e. BEPs
 BIDS is an open and growing standard
– 2016: fMRI
– 2018: MEG
– 2019: EEG and iEEG
– Soon (2020): PET, ASL, Structutal MR with multiple contrasts
– Multiple Contrast MR = BEP001, see here
– PET = BEP009, see here
– ASL = BEP005, see here
– There are (have been) 27 extensions proposals so far (up to BEP027)
– Proposals to extend BIDS have to follow certain rules, as explained in the BEP Guide
– It has to fit in the bigger picture
– It has to represent a wide community consensus

8. BIDS for fNIRS, i.e. BEP028?
 Benefits of sharing and reusing data are (hopefully) obvious
 A common specification
– Improves data interoperability and reusability
– Improves scientific quality and reproducibility
– Facilitates shared tool and pipeline development
– Reduces stress, since you don’t have to come up with everything yourself
 Society for functional Near Infrared Spectroscopy – http://fnirs.org
– fNIRS conferences in Boston (2010), London (2012), Montreal (2014), Paris (2016), Tokyo
(2018), Boston (Oct 2020)
– Shared Near Infrared Spectroscopy Format (SNIRF) is designed by the community in an
effort to facilitate sharing and analysis of NIRS data.

9. SNIRF format
 File format for NIRS data, like OXY3 or OXY4 from Artinis, or like EDF, CNT, VHDR for EEG
 Open format, based on the HDF5 “container” file format with an explicit description of the
content and structure
 Readers available for MATLAB and Python
 HDF5 libraries available for C/C++ and many other languages
 BIDS is more than a file format, also
– how to organize multiple recordings, from multiple subjects
– how to document the details of the task, stimuli and responses
– how to document sensor positions
– how to document subject details and demographics
– …
metadata

10. BIDS for fNIRS, i.e. BEP028?
 Actual raw data in the SNIRF format
 File naming and directory organization according to BIDS
 Metadata according to BIDS
– Tab-separated files
– JSON files
 Some metadata is shared with other modalities in BIDS
– Subject and task details
 Some metadata is fNIRS specific
– Optode configuration (receivers, transmitters)
– Optode positions
– Wavelengths
– …

11. fNIRS data is usually not recorded in isolation
 The unique selling points for fNIRS are that it is mobile, widely applicable in many
experimental settings, etc.
 In a typical fMRI experiment you only get DICOM files and a presentation log file
– These are the “sourcedata” and combined in a BIDS dataset
 As an example: in an experiment like the PROMPT project from Helena we get
– fNIRS data
– Xsens IMU motion capture data
– Video data (3x)
– Stimulus, response and task data (i.e. triggers and annotations)
– 3D scans with optode positions
– questionnaire data

12. Example – the PROMPT experimental setup
NIRS
EXP
XSENS

13. Video @25 Hz
camera1
camera2
camera3
NIRS @50 Hz
Accelerometer @50 Hz
IMU @xx Hz
Audio @44.1 KHz
Trigger @50 Hz

14. Data management - organizing and converting data
 Phase 1 (organize the recorded files)
– Collect all files from all devices
– Convert files in proprietary formats to open formats (oxy4->oxy3, mvn->c3d)
– Scan the paper lab notes into a pdf file
– Rename all files following BIDS
– Place all files in directory organization following BIDS
 Phase 2a (add the metadata files)
– Extract the timing of the events from each individual recording file as “events.tsv”
according to BIDS
– Determine the relative timing of all recordings to each other and store as ”scans.tsv”
according to BIDS
– Determine the timing of the experimental runs/blocks/trials in each of the recordings
 Phase 2b (minimal pre-processing)
– Annotate the video data (freezing of gate), add it to “events.tsv”
– Determine optode positions from 3D structure sensor scans
– Remove all files with identifying information (video, 3D scans)
 Phase 3 (analyze)
– Share the anonymous data with project collaborators
– Analyze the data
– Publish your results and the anonymous data
this gets archived as the
“source data”
this gets archived as the
“raw BIDS data” and shared
with collaborators

15. INCOMING
└── sub-03
├── xsens
├── nirs
├── stim
├── quest
├── structuresensor
└── video
├── acq-begin
├── acq-end
└── acq-mobile
SOURCEDATA
├── sub-01
├── sub-02
└── sub-03
├── xsens
│ ├── sub-01_bodydimensions.mvna
│ ├── sub-01_rec-01_take-01_imu.c3d
│ ├── sub-01_rec-01_take-01_imu.mvn
│ ├── sub-01_rec-01_take-02_imu.c3d
│ ├── sub-01_rec-01_take-02_imu.mvn
│ └── ...
├── nirs
│ ├── DAQ
│ │ ├── Screenshot (10).png
│ │ ├── Screenshot (11).png
│ │ └── Screenshot (12).png
│ ├── sub-01_rec-01_nirs.oxy3
│ ├── sub-01_rec-01_nirs.oxy4
│ ├── sub-01_rec-test_nirs.oxy3
│ └── sub-01_rec-test_nirs.oxy4
├── labnotes_and_questionaires
│ ├── sub-01_quest1.xlsx
│ ├── sub-01_quest2.xlsx
│ └── sub-01_labbook.pdf
├── stim
│ ├── sub-01_rec-01_sync-labrecorder.xdf
│ ├── sub-01_rec-01_sync-log.ascii
│ └── sub-01_rec-01_sync-stim.mat
├── structuresensor
│ ├── sub-01_model.jpg
│ ├── sub-01_model.mtl
│ └── sub-01_model.obj
└── video
├── sub-01_rec-01_acq-begin.mp4
├── sub-01_rec-01_acq-end.mp4
└── sub-01_rec-01_acq-mobile.mp4
BIDS_RAW
├── dataset_description.json
├── participants.tsv
├── README
├── sub-01
├── sub-02
└── sub-03
├── scans.tsv
├── xsens
│ ├── sub-01_rec-01_take-01_imu.c3d
│ ├── sub-01_rec-01_take-01_imu.json
│ ├── sub-01_rec-01_take-01_events.tsv
│ └── ...
├── nirs
│ ├── sub-01_rec-01_nirs.oxy3
│ ├── sub-01_rec-01_nirs.json
│ ├── sub-01_rec-01_nirs_events.tsv
│ ├── sub-01_optodepositions.tsv
│ └── ...
├── labnotes
│ └── sub-01_labbook.pdf
└── video
├── sub-01_rec-01_acq-begin_events.tsv
├── sub-01_rec-01_acq-end_events.tsv
└── sub-01_rec-01_acq-mobile_events.tsv
move copy
closed
archive
used for
analysis

16. Zooming out - The value of BIDS for your fNIRS research
 Following the (pseudo-) BIDS organization and the proposed data management flow, you will
perform your fNIRS analysis on a dataset that is
– Complete
– Well organized
– Re-usable by others (and your “future self”)
 You can develop and test your analysis pipeline on another similarly organized fNIRS
dataset (e.g. pilot measurements with less experimental complexity), i.e. use previous data
to leverage your analysis
 Your analysis pipeline can be reused on future datasets, i.e. use previous analyses to
leverage knowledge in the research team/lab

17. The Brain Imaging Data Structure
and its use for fNIRS research
--- to be presented at the NIRS toolkit at the Donders ---
Robert Oostenveld
r.oostenveld@donders.ru.nl