Numpy/png/nii segmentation mask(s) to DICOM ROI, how?

[Powell Molleti]

Hi All,

Needless to say I am a bit overwhelmed by DICOM standard. For example I have 22 slices of DICOM images from a study and I have 22 slices of segmentation masks in numpy/nii format. Same dimensions, same spacing and same origin when represented in 3D volume as DICOM. How do I get to see the mask as DICOM ROI in tools Horos or in OsiriX?. 

I have been trying to at-least use dcmqi to get .dcm and see the segmentation overlay on the DICOM images using Slicer. No luck yet.

Seems like a ubiquitous problem and I am missing few key steps or perhaps I went in a wrong direction. Any help or pointers is much appreciated.

thanks

Powell.

[Steve Pieper]

Hi -

Why do you need your segmentation in DICOM exactly?  We are generally big fans of DICOM SEG, but it's not yet common in the community and not required for pyradiomics.

-Steve

--
You received this message because you are subscribed to the Google Groups "pyradiomics" group.
To unsubscribe from this group and stop receiving emails from it, send an email to pyradiomics...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/pyradiomics/ba32b7d5-491b-435e-9784-b17850ff15df%40googlegroups.com.

[Powell Molleti]

Hi Steve,

Essentially the radiologists would like to review the segmentation(which are in NII/PNG/Numpy mask format) in tools like OsiriX, like they already do. They would like to give feedback and if necessary fix the segmentation in the DICOM viewers like OsiriX. It is important to having something integrated into their workflow.

How would I go about doing this?.

Thanks

Powell.

[Steve Pieper]

Does Osirix support DICOM SEG?  I don't use Osirix or Horos myself, but I thought it had its own file formats for annotation.  Would be great if they supported SEG.

In any case, if you want DICOM SEG then dcmqi is indeed the right tool for you.  It should work to make SEG that works with Slicer and other platforms (https://dicom4qi.readthedocs.io/en/latest/).

-Steve

[Powell Molleti]

I am under the impression that DICOM has a concept of ROI ?. Or is it the case that OsiriX and Horos both have their own ROI formats ?.

Thanks again, I am currently focusing on

Slicer to show DICOM SEG as the dcmqi video tutorial does.

Powell.

[Andrey Fedorov]

Powell,

You should contact OsiriX/Horos, and ask them to implement DICOM SEG.
Last time I checked, they did not support it. The more user voices we
have, the more likely they are to listen.

Yes, I believe those platforms use some platform-specific
representation for ROI, but I have never used them extensively, and I
never used that specific feature.

I, unfortunately, don't know more to help you solve your problem (=
have a segmentation in a format that can be visualized and is editable
in OsiriX/Horos).

AF

> To view this discussion on the web visit https://groups.google.com/d/msgid/pyradiomics/CAOAWGOcZ8Cbo%2BDi-0fX1eTqzUHDSbdjG4oXK-YrhxbZBL5gtkQ%40mail.gmail.com.

[Powell Molleti]

Hi Andrey,

Thank you for a very thoughtful reply and also I want to thank you for all the effort in making dcmqi and Slicer tool available, this is an invaluable asset/tool.

I will reach out to OsiriX and Horos and will also contact others who are interested in this to reach out to them to implement DICOM Seg standard.

thanks

Powell.

[Andrey Fedorov]

Thank you Powell - at this point, we have significant evidence this
and other DICOM capabilities are capable to solve a lot of issues
related to data management for annotations and other image-derived
data types, and we need users like you pushing the vendors and
developers of imaging tools to add support in their implementation.
There is no other way.

You can also see summary of support of several relevant objects in
tools here: https://dicom4qi.readthedocs.io/en/latest/, perhaps this
could be helpful too.