Arterial Spin Labeling (ASL) perfusion Magnetic Resonance Imaging (MRI) is a non-invasive imaging technique that measures the blood flow to the brain. However, the quantification and analysis of ASL images is often complicated by incomplete or inaccurate reporting of acquisition parameters. This stems from the fact that ASL images are commonly retrieved from different MRI scanners (from different vendors) and handled with a different data management strategy. Such hetereogeneity of data management results in data being stored in a different format, naming, dataset structure, and with a different availability of acquisition metadata. Sharing such a dataset with collaborators or even analyzing it locally at a later time point raises issues about the unclarity of the dataset structure and its metadata. These issues delay correct image analysis or even precludes performing absolute quantification if the metadata cannot be recovered any more, which can severely reduce the comparability between ASL perfusion MRI studies.
When downloading imaging data from an MRI scanner, data is usually stored in the Digital Imaging and Communications in Medicine (DICOM) format. DICOM is a very broad image standard for the storage of medical data and information and includes anything from an X-ray of the toe to an ultrasound of the abdomen. Its broad usability and flexible approach to storing application-specific metadata makes it very challenging to automatically check metadata completeness within the DICOM for specific neuroimaging applications, such as ASL perfusion MRI. Additionally, DICOM is not optimized for storing multi-subject and multi-session data for population. Finally, many parameters for advanced MRI techniques are not standardized or incorporated in DICOM making data anonymization and parameter retrieval complicated.
In 2016, the Brain Imaging Data Structure (BIDS) was established as a community effort, striving for a common way to organize datasets across labs and countries. The BIDS specification was created as a data organization standard describing a storage structure of datasets and a standardized description of metadata. This includes a standardized naming convention of directories and files, supporting longitudinal and multiple measurement study designs; and BIDS fields convention containing metadata, such as the acquisition parameters, stored in a JSON file, accompanying the image data stored in Neuroimaging Informatics Technology Initiative (NIfTI) data format. To assure a high adoption rate, existing standards, such as the National Electrical Manufacturers Association (NEMA) DICOM standard, were incorporated. Additional to this specification, BIDS-compliant example datasets were shared and a BIDS validator application was developed. Furthermore, several DICOM to NIfTI converters and image processing pipelines started supporting the BIDS standard. Initially, the BIDS specification covered anatomical, functional, and diffusion MRI; many extensions have been developed since, e.g., for PET, MEG, EEG, behavioral and genetic data, and other extension proposals are in development.
While BIDS is proving its value for MRI data storage in general, its philosophy is even more crucial for absolute quantitative methods such as ASL. The clinical usefulness of ASL often depends on the validity of its quantification. With many existing sequences, incorporating different labeling strategies, timing parameters, and readout methods, knowledge on each ASL acquisition parameter is important to correctly quantify and evaluate perfusion data.
A steering group of ASL experts initiated an ASL-specific BIDS extension proposal, which would accommodate ASL-specific BIDS fields and data structure, and shared it with the international ASL community for feedback. This steering group encountered several challenges. First, the ASL community consists mainly of physicists and neuroradiologists. Unlike the usual BIDS community, which mainly consists of researchers at the interface of neuroimaging methods and cognitive neuroscience, ASL does not mainly focus on functional imaging. Second, ASL-BIDS had to be revisited multiple times to reach consensus among ASL experts that were familiar with different ASL implementations.
Recently, we released the BIDS extension for ASL-MRI, focussing on the ASL approaches recommended in the 2015 ASL consensus paper. More advanced ASL approaches, mainly used in research and developmental settings, are planned to be included in later versions. This extension was tested and endorsed by the majority of the ASL community, and is currently being adopted in many DICOM-to-BIDS conversion tools and ASL processing software packages. Also, the Open Source Initiative for Perfusion Imaging (OSIPI; osipi.org) incorporated the ASL-BIDS in their ASL lexicon, ASL challenges, and Digital Reference Objects and Phantoms. In the future, ASL-BIDS – in conjunction with OSIPIS’s ASL lexicon – may encourage NEMA and the MRI vendors to include missing BIDS parameters as DICOM fields.
For more details on the development of ASL-BIDS: https://www.nature.com/articles/s41597-022-01615-9
Listen to our experiences when developing ASL-BIDS: https://www.youtube.com/watch?v=mMTI_2In7xY
Get started with BIDS: https://bids-standard.github.io/bids-starter-kit/
The BIDS specification: https://bids-specification.readthedocs.io/en/stable/
Contribute to one of the ongoing extension proposals: https://bids.neuroimaging.io/get_involved.html
Please sign in or register for FREE
If you are a registered user on Research Data at Springer Nature, please sign in
ASL-BIDS is a fantastic example of the kind of community-based reporting guideline that we describe as an essential component of data FAIRness in our recent Research Data Community posting and in the associated paper in Scientific Data. ASL-BIDS offers an excellent means to formalize the attributes of a class of neuroimages stored in research repositories. One can imagine embedding ASL-BIDS metadata within other, more comprehensive metadata to describe not only the images and the imaging techniques, but also the rationale for the experiment, the subjects, and the experimental protocol that led to the images in the first place. This degree of metadata "richness"—which goes beyond the acquired images and includes a description of the complete experiment— is needed for datasets related to ASL images truly to to be FAIR in accordance with the FAIR principles.