ORMIR-MIDS as a Standard Specification

ORMIR-MIDS as a Standard Specification#

The ORMIR-MIDS specifications are a simple and intuitive way to organize MSK imaging data. They are an extension of the specifications of the Brain Imaging Data Structure (BIDS) and the Medical Imaging Data Structure (MIDS)

In this page, you will find:
1. Structure and naming of folders and files
2. Supported imaging modalities

Structure and naming of folders and files#

Folder structure#

The folder structure is hierarchical and composed of 4 levels of nested folders:
Dataset → Subject → Session → Imaging Method

Dataset: root (or parent) folder containing all the MSK data (E.g., Dataset_organized)
Subject: folder(s) containing all the data about a single subject (E.g., sub-01 for the first subject)
Session: folder(s) containing the data of an acquisition session at a specific time point(E.g., ses-01 for baseline acquisitions)
Imaging Method: folder(s) containing the data of a specific acquisition. (E.g., mr-anat for anatomical MR images). You can find the currently supported imaging methods in this table

Filesystem structure#

Within each folder or subfolder, there are specific files:

Participants’ information: definition under construction!
Scans’ information: definition under construction!
Session information: definition under construction!
Image volume: 3D image
Image information: Header of the dicom images in .json format

Folder naming#

Folders have precise names:

Folder names referring to subjects and sessions are composed of two parts:
1. sub or ses, indicating subject (blue in the figure above) or session (purple in the figure above)
2. <ID> of the subject or of the session (e.g., 01)
Note: The two parts are separated by dash (-)
Folder names referring to image modalities (yellow in the figure above) can have various names, such as mr-anat, ct, etc. For a list of the available names, see the second column (called Folder name) in the table below

File naming#

Files have precise names:

Each file name is composed of three parts, each of them containing information about:
1. Subject, such as sub-01 (blue in the figure above)
2. Session, such as ses-01 (purple in the figure above)
3. File content, such as scans, metadata, etc. (green in the figure above). For the images, you can find the currently supported file types in the table below
Note: The three parts are separated by underscore (_)

Supported imaging modalities#

In the following table, you will find the image data types that are currently implemented or under development in ORMIR-MIDS. The table columns specify:

Imaging method: the image acquisition method, such as computed tomography, MR imaging, etc.
Folder name: name of the output folder containing the images (e.g., mr-anat, ct, etc.). See the folder naming specifications for more details
File name suffix: suffix of the file name of the image volume (e.g., mese, megre, etc.). See the file naming specifications for more details
Image volume: dimension (2D, 3D, or 4D) and directions (x, y, z, echo) of the obtained image volume (i.e., nii.gz files)
JSON fields: information that the dicom header must have to be able to transform the data into the ORMIR-MIDS structure

Imaging method	Folder name	File name suffix	Image volume	JSON fields
Computed tomography	ct	N/A	3D (x,y,z)	XRayEnergy in kVp XRayExposure in mAs
Computed / plain radiography	cr	N/A	2D (x,y)	ExposureTime in ms X-RayTubeCurrent in mA
MRI: Multi-echo gradient echo	mr-anat	megre	4D (x,y,z,echo)	EchoTime (array) in ms WaterFatShift in pixels MagneticFieldStrength proposed: ReadoutMode: Monopolar/Bipolar proposed: PrecessionDirection: Counter-/Clockwise
MRI: Multi-echo spin echo	mr-anat	mese	4D (x,y,z,echo)	EchoTime (array) in ms RefocusingFlipAngle in degrees optional: ExcitationProfile and RefocusingProfile (arrays) providing the slice profiles in degrees
MRI: T1w / T1w-FS / T2w / T2w-FS	mr-anat	t1w / t1w-fs / t2w / t2w-fs	3D (x,y,z)
MRI: Quantitative T1 / T2 / wT2	mr-quant	t1 / t2 / wt2	3D (x,y,z)

Current and future work#

We are working on converters for:

More imaging modalities, such as HR-pQCT, MRI Phase contrast, MRI DESS, MRI Diffusion
Derived data, such as segmentation labels

Here are the specs of some imaging modalities we will support:

Phase contrast

NIfTI structure: 4D (x,y,z,t) + multiple suffixes for venc direction
Filename suffix: pc_mag / pc_ph_1 / pc_ph_2 / pc_ph_3
Folder: flow ??
The velocity information is stored as phase data scaled between -π and +π.
JSON required fields:
- Venc in cm/s
- EncodingDirection (3D vector) for each phase volume, indicating the direction of the positive velocity encoding for that volume. TBD: patient coordinate system or image coordinate system.

Diffusion

NIfTI structure: 4D (x,y,z,direction)
Filename suffix: diff
Folder: diff ??
JSON required fields:
- MixingTime in ms
- EncodingDirection (array of 3D vectors). The norm of the vector is the b-value. The normalized vector indicates the direction of the diffusion gradient in patient coordinates.

Segmentation labels

NIfTI structure:
- 3D (x,y,z) with integer values, from 0 to N, each value corresponding to a different label
- 4D (x,y,z,label) stack of binary (0/1) values, each dimension corresponding to a label
Filename suffix: seg ??
Folder: seg ??
JSON required fields:
- Labels (array of strings). List of the labels represented in the masks. The first value in the list corresponds to either a gray level of 0 or to the 1st volume in the fourth dimension. E.g. ["Background", "SOL", "VM", "VL"]
- Note: the string representation of the labels must follow a standardized format. While it is possible that the same anatomical structure is represented by different labels (e.g. SOL or Soleus), the labels must be known. This allows flexibility in the implementation of segmentation tools, while keeping easy interoperability because all values are easily convertible. A list of standardized labels is visible here. Please contact Francesco Santini if you would like to add your own definitions.