Page MenuHomePhabricator

Implement specific UID strategy for mitk::Image
Closed, WontfixPublic


mitk::Image should use its own UID strategy and not the default own.
Reasoning: UID should be as close to dicom as possible.


  • From now all mitk::Image should have an filled DICOM.0008.0018 (SOP Instance UID) property for all its z-position and timesteps. (If it does not exists through the loading process it will be generated (this will be done by T23746))
  • The content of DICOM.0008.0018 will be hashed to generate the ID
  • If, for any reason (evil!!), DICOM.0008.0018 does not exist, one would fall back to default uid generation

Event Timeline

After the final discussion round, this task is for me, a candidate for deletion. I see the problem, pointed out by @nolden, that we provoke an anti pattern if we stared to define UID strategies below BaseData. As far as I see it, after having T23766 and proper helper to configure data dicom conform (see T23748 and T23774) this not needed anymore. @metzger wants to evaluate it for her use cases. If she has no arguments against it, I would spite this task and sub task