DICOM Segmentations do not display

[fedorov]

Together with @CPBridge, @DanielaSchacherer, @dclunie and @cgorman we have been struggling to understand why we are unable to visualize multi-resolution DICOM SEG encoded as either TILED_SPARSE or TILED_FULL in Slim+Google DICOM Store combination. The samples are available in tests S6 and S7 in the IDC test inventory (access to the document is limited to IDC team members, but these specific samples can be shared at request with anyone interested to investigate the issue).

[igoroctaviano]

I'm getting a few 400s when using GCP:

https://healthcare.googleapis.com/v1/projects/idc-sandbox-000/locations/us/datasets/idc-dicom-test-inventory/dicomStores/test-samples/dicomWeb/studies/2.25.213661963103110408605329613498871186883/series/1.2.826.0.1.3680043.10.511.3.93656756710834591768370015932747426/instances/1.2.826.0.1.3680043.10.511.3.49566294483227479815612054298702676/frames/1683

[{
  "error": {
    "code": 400,
    "message": "no PixelData found in requested instance",
    "status": "FAILED_PRECONDITION"
  }
}
]

There's a possibility that even though no errors were logged during the import in GCP, some segments might have silently failed. I'll retry the import in a separate dicom store to see if I get the same results.

[CPBridge]

I thought @DanielaSchacherer had managed to access these cases successfully over DICOM web using the python client?

[igoroctaviano]

I could load the SEGs successfully locally using dcm4chee. The problem is trying to load them via GCP.

[CPBridge]

Understood, yes I thought she tried loading from GCP

[DanielaSchacherer]

Hi Igor, I was indeed able to retrieve the whole instances from a Google DICOM store using the Google DICCOMweb client and they didn't seem corrupt or anything. What I wasn't able to do was to visualize any kind of annotation with the Slim viewer when connecting it to a Google DICOM store.

What exactly did you do when you received the error mentioned above?

[igoroctaviano]

Hi Igor, I was indeed able to retrieve the whole instances from a Google DICOM store using the Google DICCOMweb client and they didn't seem corrupt or anything. What I wasn't able to do was to visualize any kind of annotation with the Slim viewer when connecting it to a Google DICOM store.

What exactly did you do when you received the error mentioned above?

Have you tried toggling the segmentation and checking the logs? I'm getting a bunch of failed requests related to missing pixel data (segments)

[fedorov]

As discussed elsewhere, this was due to missing support of BitsAllocated=1, which is now available. I confirmed this is working now, at least on the sample I tested.