Convert Enhanced dicom into plain dicom

[tboutelier]

Hi guys,
First let me thank you for the amazing you have done on pydicom, I really like it and use it regularly.
I was wondering if you plan on developing a functionality to convert enhanced DICOM format into plain DICOM format? I mean by that generating n DICOM files with one slice of the original DICOM file in each, with all the correct tags filled. I know that DCM4CHE proposes this functionality in Java, but I could not find the equivalent in Python. Adding this to pydicom would be great!
Thanks!

[mrbean-bremen]

See my answer in #1691 - here is the gist of it:
I'm not aware of any tool or code that does that out of the box. I think such a functionality does not belong directly into pydicom, but into a higher-order library like highdicom - I actually would ask there, as they already do the opposite (create Enhanced DICOM from single frames).

There is also no generic way to do this, as each SOP class has its own quirks, but most of it is indeed similar, e.g.

• create a new DICOM dataset for each slice
• copy all of the data in the DICOM root
• create a new SOPInstanceUID for each dataset
• flatten the tags in the Shared Functional Group Sequence and copy them
• flatten the tags in the Per Frame Functional Group Sequence that belongs to the current slice and copy them
• read the multi-frame data using pixel_array as a NumPy array, split it into the slice data and copy these into PixelData of the slices

After that, some of the tags have to be removed or adapted, and new tags have to be added, which is mostly specific to the concreate SOP class and to your use case, so some configuration would be needed here, too. You also may want to compress the dataset, which may involve other libraries, as pydicom does not support this yet.

All in all, this is not a trivial thing to do and certianly does not belong directly into pydicom, though it would fit into pydicom-contrib and possibly highdicom, if somebody finds the time to do this. You are welcome to contribute, of course 😄

[tboutelier]

Thanks for the answer.
I agree with the proposed strategy. I think however that all the difficulty is in your last comment, that some tags need to be removed or adapted. Which ones and how, here is the question!
I still think this would fit in the pydicom library, although it is quite high level function!

[mrbean-bremen]

I think however that all the difficulty is in your last comment, that some tags need to be removed or adapted. Which ones and how, here is the question!

True, that is what I meant. And that is exactly the part that does not belong into pydicom in my opinion, as pydicom is a general-purpose DICOM framework and generally does not handle the specifics of individual SOP classes - that would make the maintanence a nightmare, given the available resources. It might make sense to add some convenience function, for example to flatten the content of the functional group sequences, but the specific stuff belongs into a higher-order package.

[pangyuteng]

Just wanted to say I am a happy pydicom user and after having replaced an in-house multiframe-splitting module (and just found out the update module utterly failed - due to not testing it via enough modalities) I fully support pydicom NOT supporting this. 😂

For those that do not want to implement your own modules, you can check out the below few implementations:

• dclunie's dicom3tools dcuncat
• gdcm's gdcmtar
• dcm4chee's emf2sf
• highdicom's pull-request: "Multiframe Conversion #34" (work-in-progress!)

for those that stopped by this page, for entertainment purposes - my barebones failed logic are provided in the gist below:
https://gist.github.com/pangyuteng/803cf2fda3d7568aaa348fdd409937d7

[darcymason]

For those that do not want to implement your own modules, you can check out the below few implementations:

Thanks for the summary and the list of references - that is really useful information for anyone looking for this functionality.