Depending on how your data will be used, you can upload it to a socket of type Image or DICOMImageSet.

Data uploaded to Image sockets is converted to SimpleITK (mha) fileformat for radiology or TIFF for pathology. In challenges, this format significantly reduces the workload for participants because the images are provided in an easy-to-use, analysis-ready format.

Data uploaded to DICOMImageSet sockets is not converted but stored as DICOM, following a rigorous de-identification process. In challenges, this means participants will need to handle DICOM parsing and interpretation themselves. Choosing this format may be appropriate when:

1. You want to run a challenge that evaluates algorithms using DICOM as their direct input.
2. You want to publish an algorithm that uses DICOM as their input.
3. You need to use DICOM tag information in your challenge or algorithm.

Note: De-identification removes or replaces many DICOM tags with dummy values (see this readme). If you need metadata that will be stripped during this process, you can provide it separately in JSON format. The data owner is solely responsible for ensuring that the combination of imaging data and metadata keeps the subject de-identified.

Note: The most common DICOM SOP Classes (i.e. 'CT Image Storage') are supported on the platform. See the documentation for a list. If you would like to use a SOP Class that is not supported yet, please contact support to request its addition.