'Segmentations oblique to the acquisition plane of the source data are not yet supported'

[heart project]

Dear All,

I recently had a problem importing a mask into ohif-xnat. The error message is: 'Segmentations oblique to the acquisition plane of the source data are not yet supported'.  Can anyone help me solve the problem, or some tip? 

Is there any pre-processing that I can apply to the images so that they are accepted by ohif-xnat?!

 Any information would be very welcome.  In attach the screen error.

Best regards,

Pedro.

[heart project]

It's interesting to note that for some exams (from the same vendor) I'm able to import the masks, while others generate this error

[Mo Al Sa'd]

Hi Pedro,

When a mask segmentation is imported, the viewer checks its orientation and attempts to match that with the orientation of the reference scan image. This procedure is used in the viewer to fix orientation issues that might have been introduced by tools used to generate the segmentation. However, the alignment of segmentation data that is oblique to the reference scan is not supported yet.

Since you have mentioned that some segmentations import successfully, there could be a bug in the orientation checking algorithm that we use. I wonder if it would be possible to share the links for some sample data via email?

Kind regards,

Mo

[heart project]

Dear Mo,

Attached are the images.

These scans pertain to short-axis cardiac cycles. The exams have been categorized into two folders. One contains segmentations that are successfully read by OHIF-XNAT (No Problem), while the other set encounters issues (with Problem). The represent end-systole (ES) and end-diastolic (ED) frames. The masks are provided separately, and the segmentations are specific to ED and ES only.

For instance, within the 'noProblems' folder:

Files:
    • 13SGoEFcFgNWmhlFItebSh.dcm
    • 30KYUtPiKpZWWkmldnyGms .dcm

pertain to frames 13 and 30, respectively.

To display the overlapping masks, I separated them into a folder in XNAT (assessors) as in the attached image (folders.png)
As this data is sensitive, I kindly ask you to avoid sharing it. 

Best regards,

Pedro.

[heart project]

[heart project]

[heart project]

Hi Mo,

Were you able to download the data ?

Any news or suggestions on how I can also try to help identify the possible adjustment in the plugin ?

Best regards,

Pedro.

Em segunda-feira, 4 de março de 2024 às 17:21:26 UTC-3, heart project escreveu:

[Mo Al Sa'd]

Hi Pedro,

Firstly, there are inconsistencies in both datasets (the ones with and without import issues):

• The SEG ImagePositionPatient values do not match that of the reference image instances
• The SEG image is flipped horizontally relative to the reference image

I have found a bug in the Viewer where it attempts to reorient the SEG image. This bug will be fixed and included in the next release of the Viewer.

Kind regards,

Mo

[Chidi Ugonna]

Hi Mo,

Thank you and the team for all your work on the OHIF Viewer - and of course by extension everyone involved in XNAT in all contexts - The XNAT ecosystem is absolutely critical to our work.

I am just following up on this thread to see if the the issue you mentioned here in the viewer has been addressed in OHIF 3.7.0. I  seem to be experiencing this in a different context in XNAT 1.8.10.1 + OHIF 3.7.0

I am doing some exploratory work with MONAILabel and running a remote MONAILabel server to segment CT and MRI images based on some of the great examples on the MONAI Label github. The CT segmentations work fine in OHIF Viewer but I am getting an error with the segmentation model I am running on MRI MPRAGE images. The client shows a red dialog with an error like "g is undefined" - when I look in the web developer tools on firefox I see "Segmentations oblique to the acquisition plane of the source data are not supported"

The CT example looks like this:

Any insights you have you would be much appreciated!

Thank you

Chidi

[Mo Al Sa'd]

Thank you, Chidi.

There could be various possible reasons behind that behaviour;

1. The MR scan is 4-D, has frames of different orientation, or is multi-frame (Enhanced IOD). The latter case should have been resolved in v3.7.1
2. MONAILabel generated the NIfTI file with wrong orientation - could happen but less likely

Have you tried to create and export a mask manually on the MR scan, and then import it back?

Kind regards,

Mo

[Chidi Ugonna]

Hi Mo,

Thanks for the reply - So I upgraded to OHIF 3.7.1 in XNAT 1.8.10 and tested the workflow again. I was able to create, export and re-import masks successfully with both standard and enhanced Dicoms.

This is a short summary of my experience using various models from the monai zoo ( https://monai.io/model-zoo.htm ) and images from the cancer imaging archive ( https://www.cancerimagingarchive.net.)

1) LUNG1-001 CT image (NSCLC-Radiomics from https://www.cancerimagingarchive.net)  , tested on Spleen CT Segmentation  -> This ran successfully

2) Prostate-MRI-US-Biopsy-0001 T2W Image (Prostate-MRI-US-Biopsy, Jun 28, 2009, MR PROSTATE WO+W CONTRAST, t2_spc_rst_axial obl_Prostate) tested on model  prostate MRI anatomy  -> This ran successfully

3)  UPENN-GBM-00017 T1W (study id ...62967, UPENN-GBM,  Jun 15, 2001, BrainTumor, T1 SAG MPRAGE : Processed_CaPTk ) tested on     Whole brain large UNEST segmentation  -> This did not complete and gave me the error "Error: Segmentations oblique to the acquisition plane of the source data are not yet supported).

The dicoms for  UPENN-GBM-00017 above are not in enhanced mode and are not 4D. I do not think the frames are in different orientations (I am not sure the best way to confirm this but I looked at the ImageOrientationPatient field in all 160 slices and they were the same).

 I also tested UPENN-GBM-00017  in  the  Whole brain large UNEST segmentation  workflow in the ohif viewer that comes with monailabel and which is accessible at localhost:8000/ohif (the images are uploaded to an orthanc server as described here https://github.com/Project-MONAI/tutorials/blob/main/monailabel/monailabel_radiology_spleen_segmentation_OHIF.ipynb ) and the segmentation completed successfully.

I tested the Whole brain large UNEST segmentation   workflow with some enhanced T1w dicoms in the XNAT OHIF plugin that I had available and this time the tab crashed - screenshots further below show what I captured in the developer tools in firefox. The enhanced Dicoms worked successfully in the monai ohif instance.

I am not sure whether this is something intrinsic to the way I have things setup. The XNAT has been run using the docker-compose instance.

Thank you for any insights or help and any workaround suggestions.

Best

Chidi