Skip to the content.


ORMIR-MIDS is a ‘Medical Imaging Data Structure (MIDS)’ developed by the ‘Open and Reproducible Musculoskeletal Imaging Research (ORMIR)’ community. It comprises two components:

The aim of this standard is to provide a common interface for data postprocessing and analysis tools to read and write data in an interoperable way. The existing DICOM standard has very broad guidelines that result in a large variability among vendors of medical devices, to the point that it can hardly be considered a “standard” at all.

The ORMIR-MIDS standard is inspired by the neuroimaging BIDS specification and its MIDS extension to additional imaging modalities and anatomical regions. ORMIR-MIDS is mostly compatible with BIDS, in the sense that it is intended to subdivide and group the acquired images based on their source modality (e.g. magnetic resonance imaging, computed tomography, plain radiography) and diagnostic role (e.g. anatomical images, quantitative maps, functional images, etc.).

Images are stored as three- or four-dimensional NIfTI datasets paired with three plain-text JSON headers each. These images are stored in a predefined directory and naming structure that allows immediate identification of the role of each dataset.

As integration with medical and PACS systems is a prerequisites for our standard, bidirectional conversion between DICOM and ORMIR-MIDS is ensured.

ORMIR-MIDS is a fork and extension of muscle-BIDS.

The specification

ORMIR-MIDS prescribes that all imaging data should be kept in a NIfTI format inside a predefined directory structure, with a specific file-naming scheme. Associated with each NIfTI file, there are up to three JSON files containing:

  1. The relevant parameters for that specific acquisition (e.g. ‘XRayEnergy’ for CT, multiple ‘EchoTime’ values for multi echo spin echo MRI).
  2. The patient/participant information (which can be omitted for anonymization).
  3. All extra headers needed for the conversion to DICOM (originating from the initial acquisition).

Each imaging modality and acquisition type has a corresponding set of relevant parameters and a naming convention. Further, unlike the original BIDS specification, ORMIR-MIDS also provides a structure for health-related data—including an ‘ehr’ directory per session, containing a table (.tsv) of clinical data and (optionally) the raw clinical data itself.

The full ORMIR-MIDS specification can be found here.

The Python package

A work-in-progress Python package is delivered to implement the conversion between DICOM and ORMIR-MIDS, and the I/O of ORMIR-MIDS-compatible files from Python programs. More information is found here. The package source code and user guidelines are available at

The development of the ORMIR-MIDS specification and package started during the 2nd ORMIR workshop “Sharing and Curating Open Data in Musculoskeletal Imaging Research” and is currently ongoing. ORMIR-MIDS is an extension of muscle-BIDS, which was partly developed during the 2nd ORMIR workshop “Building the Jupyter Community in Musculoskeletal Imaging Research”. Both ORMIR-MIDS and muscle-BIDS are part of the ORMIR community core projects.