When compressing with Lossless, does DICOM mandate a new SOPInstanceUID?

[Simon]

Hi all,

I'm pretty sure compressing instances with lossy compression requires a new SOPInstanceUID to be generated, as the image can be substantially different.

But if I'm compressing lossless, is a new UID required? I understand it's desirable, but for my use case that has some practical drawbacks.

Your input is appreciated.

Simon

[David Clunie]

For lossless (reversible) compression, it is neither desirable nor required to assign new UIDs; lossless compression is considered transparent to the interpretation of the image pixel data (no different than changing from little to big endian representation, etc.), so assigning new UIDs would be counter-productive.

Only lossy (irreversible) compression requires new UIDs, since it potentially changes the "professional interpretation" of the pixel data. See:

http://dicom.nema.org/medical/dicom/current/output/chtml/part03/sect_C.7.6.html#sect_C.7.6.1.1.3

One can debate whether or not this is a good design, but that's what it says.

David

[Simon]

Thanks for clearing that up David.

Cheers!

[Muralidhar Chowdary Nagineni]

Hi David,

I had one more doubt, what happens to all Non-image objects(PR,SR,SEG,KO etc) for a given study if we compress the reference images with lossy compression technique.

I feel ideally we also need to change the old reference sop instance to new sop instance by parsing inside the image specified every where to maintain the link. Otherwise they(non-image dicom object associated with image) will get dangling when sent to other dicom node.

Does all commercial softwares follow this way ? or do they just show a warning to user saying this action(convert images to lossy) might leads to dangling non-image objects.

Also one more use case like: when users perform the QA operation by splitting the series or deleting the images ,do we need to correct the reference sop instances in non-image objects or split these non-image objects?

Thanks & Regards,
Muralidhar

[SpreadTooThin]

I found am AGFA PACS that insisted in changing the SOP Instance UID when sending a lossless compressed images. I wondered if that meant about the source image. Was it lossy?