Open Data in Musculoskeletal Imaging
Introduction
As musculoskeletal imaging technologies advance, so too does the need for shared, high-quality datasets and standardized image-processing frameworks. Two complementary initiatives — UMUD (Universal Musculoskeletal Ultrasonography Database) and ORMIR-MIDS (Open and Reproducible Musculoskeletal Imaging Research community—Medical Imaging Data Structure) —are aiming to improve how MSK researchers collect, organize, and analyze imaging data. Presented jointly at Love Data Week 2025, these projects underscore the importance of accessible data, standardized formats, and reproducible workflows.
UMUD: Universal Musculoskeletal Ultrasonography Database
Motivation and Aims
Ultrasonography is a versatile imaging modality for assessing muscle and tendon properties, including architecture, cross-sectional area, and tissue stiffness. However, the lack of a centralized, open repository—combined with the absence of standardized metadata descriptors—hampers large-scale analyses and the development of automated image analysis algorithms. UMUD addresses this by curating publicly available musculoskeletal ultrasonography datasets from platforms such as Zenodo and the Open Science Framework.
Key Features
- Aggregated, Searchable Datasets
UMUD provides detailed descriptors for each dataset, covering muscle group, ultrasound device, participant demographics, and publication details, to streamline discovery and foster collaborations. - Benchmark Datasets for Algorithm Development
Researchers can train and validate their analysis pipelines on benchmark datasets that include multi-expert annotations of muscle architecture, panoramic cross-sectional images, and labeled data for deep learning models. - Tool and Algorithm Index
UMUD also curates a list of state-of-the-art automated analysis algorithms, simplifying the selection of resources that best fit each researcher’s objectives.
ORMIR-MIDS: A Standard for MSK Imaging Data
Motivation and Aims
Quantitative MSK imaging is inherently diverse, encompassing multiple imaging modalities, anatomical sites, and vendor-specific file formats. Building on the Brain Imaging Data Structure (BIDS) framework, ORMIR-MIDS introduces a clinically oriented data standard that unifies this heterogeneous landscape. By adopting a common file format and metadata schema, ORMIR-MIDS paves the way for seamless interoperability, reproducible workflows, and powerful cross-platform analyses.
Key Features
- NIfTI + JSON Files
ORMIR-MIDS employs the open NIfTI format for images and JSON for metadata, allowing researchers to easily share and interpret data across software ecosystems. - Folder Structure and Naming Conventions
Data are organized by participant and modality-specific subfolders, ensuring intuitive navigation. Each imaging file is paired with minimal and/or optional JSON headers that document image acquisition and patient-related details. - Interoperability and Anonymization
The two-header approach (minimal vs. sensitive information) simplifies anonymization and supports future conversions to and from other formats.
Conclusion and Future Outlook
At Love Data Week 2025, we highlight how UMUD and ORMIR-MIDS are aiming to increase openness, standardization, and reproducibility in the musculoskeletal imaging landscape. Collectively, they address the current bottlenecks in data accessibility, facilitate rigorous validation of analysis tools, and promote consistent practices in MSK research.
To join the event please register.