ezBIDS: Guided standardization of neuroimaging data interoperable with major data archives and platforms

Data standardization is essential for promoting data sharing, addressing reproducibility issues, and advancing scientific rigor in neuroimaging. The Brain Imaging Data Structure (BIDS) standard provides a common framework for organizing and describing neuroimaging data, facilitating data sharing and replication. However, current BIDS conversion tools often require technical expertise, including coding skills and knowledge of the BIDS specification. This poses a barrier for many researchers. This paper introduces ezBIDS, a novel tool designed to overcome these limitations by providing a user-friendly, web-based platform for BIDS conversion that requires no coding skills or installation. ezBIDS leverages a Software-as-a-Service (SaaS) model, guiding users through the conversion process with semi-automated inference and interactive feedback. The tool aims to improve the accessibility of the BIDS standard to a broader audience, thereby enhancing data sharing and promoting reproducibility within the neuroimaging community. The widespread adoption of BIDS is critical to address the current reproducibility crisis in neuroimaging, particularly given the high cost of acquiring neuroimaging data. A standardized framework allows for easier sharing of data and facilitates the use of standardized analysis pipelines, which helps improve the reliability of research findings.

Existing BIDS conversion tools, while valuable, often present significant hurdles for researchers. Many require command-line interface (CLI) expertise and programming knowledge for installation and usage. Others may have limited functionality, such as lacking support for task event files conversion or robust quality assurance (QA) checks. This necessitates the development of new tools to make BIDS conversion more accessible to a wider range of researchers, regardless of their level of technical expertise. The authors reviewed several existing tools, highlighting their strengths and weaknesses in comparison to the proposed ezBIDS system. Specifically, they mention Osirix, Horos, BIDS Toolbox, fw-heudiconv, pyBIDSconv, Biscuit, and BIDScoin, pointing out the lack of user-friendly interfaces and functionalities in several of them.

ezBIDS employs a Software-as-a-Service (SaaS) model, hosted on brainlife.io. Users upload raw imaging data (DICOM or dcm2niix output) to a secure server. The core of ezBIDS uses a series of Python functions to automatically infer BIDS-relevant information, such as data type, suffix, and entity labels, employing heuristic rules based on metadata extracted from DICOM headers. This information is then presented to the user via a web interface for review and modification. The process involves several key steps: 1. **Data Transformation:** DICOM files are converted to NIfTI format using dcm2niix (if necessary). 2. **BIDS Information Identification:** ezBIDS Core analyzes the data to propose BIDS-compliant metadata, including subject/session IDs, data types, suffixes, and entity labels. The process handles grouping similar images and accounts for minor variations in metadata values. 3. **Dataset Description:** Users provide basic dataset information via the web interface. 4. **Subject and Sessions:** Users review and modify subject and session IDs, using options like PatientName, PatientID, or numerical labels. 5. **Series Mapping:** Users review and modify BIDS information for each series of images. 6. **Events:** ezBIDS handles conversion of task event timing files to the BIDS-specified events.tsv format. Users map columns from uploaded event files to BIDS columns. EzBIDS attempts to automatically link events files to corresponding BOLD files using subject, session, task, and run identifiers. 7. **Dataset Review:** Users review the entire dataset using a web interface, making adjustments as needed and using quality assurance checks (bids-validator). 8. **Pseudo-anonymization and Defacing:** Users can opt to deface anatomical images to remove identifying facial features using Quickshear or pydeface. 9. **Participants Information:** Users can add or modify participant phenotype information. 10. **BIDS Dataset Validation and Download:** The final BIDS dataset is validated using bids-validator and can be downloaded locally or uploaded to OpenNeuro.org or brainlife.io. 11. **ezBIDS Templates:** Users can download a template for future sessions to speed up the conversion process. 12. **Guided BIDS Metadata Capture:** ezBIDS allows users to specify additional metadata fields not handled by dcm2niix. The entire workflow is designed to be intuitive and user-friendly, reducing the technical expertise needed for BIDS conversion.

ezBIDS successfully converts neuroimaging data to the BIDS standard without requiring installation, programming skills, or extensive knowledge of BIDS. The tool's semi-automated approach significantly reduces manual effort by leveraging heuristic rules to infer BIDS information. The web interface allows users to easily review and modify this information, facilitating accurate and efficient conversion. A key advantage of ezBIDS is its unique feature of handling task event timing file conversion, a task often overlooked by other tools. This feature significantly improves the usability of task-based fMRI datasets for replication and analysis. The interoperability with OpenNeuro.org and brainlife.io further enhances the tool's practicality by providing streamlined pathways for data sharing and collaborative analysis. The quality assurance checks and the ability to create and reuse templates improve the efficiency and quality of the conversion process. The tool has been validated across multiple scanners and data types from several institutions and shows effectiveness in handling various imaging modalities, including anatomical (anat), functional (func), field maps (fmap), diffusion-weighted imaging (dwi), and perfusion (perf) data. ezBIDS's user-friendly design and comprehensive functionality addresses a crucial need within the neuroimaging community for accessible data standardization.

ezBIDS addresses the limitations of existing BIDS conversion tools by providing a user-friendly web interface that requires no coding or installation. This greatly enhances the accessibility of the BIDS standard to a wider range of researchers. The key findings demonstrate that ezBIDS effectively converts various neuroimaging data types to BIDS while providing guidance and quality assurance checks. The inclusion of task event conversion significantly improves the usability of task-based fMRI data. Interoperability with major data archives and platforms fosters collaboration and data sharing. The results contribute to the wider adoption of FAIR data principles within the neuroimaging community. While the focus has primarily been on MRI data, future development aims to incorporate additional imaging modalities. The tool's design promotes reproducibility in neuroimaging by providing a standardized framework for data organization and analysis.

ezBIDS offers a significant advancement in neuroimaging data standardization. Its user-friendly interface, automated features, and integration with major data repositories significantly lower the barrier to entry for BIDS adoption. Future work includes expanding support for additional imaging modalities and enhancing the tool's capabilities based on user feedback. ezBIDS's contribution lies in making data standardization more accessible, promoting data sharing, and ultimately furthering reproducibility in neuroimaging research.

While ezBIDS strives for broad compatibility, there may be limitations in handling highly specialized or uncommon data formats or metadata. The current focus on MRI data means that support for other modalities may be less mature. The web-based nature of ezBIDS means that it relies on internet connectivity and may be inaccessible in locations with limited or unreliable internet access. The reliance on dcm2niix for DICOM to NIFTI conversion implies any limitations or specific handling of dcm2niix should be considered as a limitation of ezBIDS.