The BIDS framework for storing and sharing neuro-imaging data
Common data recording standards are necessary to promote data sharing. If data sharing is not supported and embraced, the valuable information about human brain dynamics contained in many large- and smaller-scale neuro-imaging data sets, recorded with care at considerable expense, is at strong risk of being lost to science unless the data are made available to other researchers.
The new Brain Imaging Data Structure (BIDS) standard has recently emerged to fill the gap in data recording standards. Even if the format is just a few years old, several data archives have already have adopted it, for example: OpenNeuro (previously OpenfMRI), FCP INDI, SchizConnect, the Developing Human Connectome Project, and the Omega archive. Currently, the BIDS validator, an online resource that can check dataset integrity, supports MRI and magnetoencephalography (MEG), and will soon support electroencephalograpy (EEG), intracranial electroencephalography (iEEG), and positron emission tomography (PET).
The BIDS standard is a simple, already broadly-supported framework developed to organize functional Magnetic Resonance Imaging (fMRI) data files and experiment meta information. Files are organized into folders with standard name formats. Experiment information is stored in simple column tabular files and JSON description text files and does not rely on a complex database infrastructure. Event files are tabular text files possibly containing links into an external resource. Because the default BIDS data storage framework is so simple – unformatted or minimally formatted text file, binary data files within folder following a specific naming convention - and is not linked to specific software, it is unlikely to become soon obsolete.
This new standard also allows to expand our horizon of what is possible. For example BIDS apps are containerized solutions (a type of precompiled virtual machine) that can be applied to BIDS datasets, and used seamlessly on either supercomputing resources or laptop computers. BIDS derivative datasets are BIDS datasets that have already been processed and contain derived data. Search engines capable of searching for BIDS datasets on multiple databases are also being developed and major neuroimaging software are developing solution to directly import and export BIDS datasets.
The community is encouraged to participate in the continued development of the BIDS set of standards, and to actively take advantage of the valuable resources that are being developed to make the sharing of data and results easier and more complete.
- Arno Delorme, NITRC Domain Expert