Failed to Fetch Sample from MONAI Label Server

[landynfrancis]

Hello all,

I have previously used MONAI Label several times for Active Learning and assisted annotation of 3D images within 3D Slicer. I am revisiting my previous setup, now working with 2D TIFF files.

When I try to retrieve a sample from the MONAI Label server, 3D Slicer reports that it failed to fetch the sample. After some initial research I thought it was the file format, so I converted my files to NRRDs then to NIfTIs. I am still getting the same error, see below:

There is more than one file, use the vtkITKArchetypeImageSeriesReader instead


Exception from vtkITK MegaMacro: 
itk::InvalidRequestedRegionError (0x17392150)
Location: "unknown" 
File: /work/Stable/Slicer-0-build/ITK/Modules/Core/Common/src/itkDataObject.cxx
Line: 367
Description: Requested region is (at least partially) outside the largest possible region.





Algorithm vtkITKArchetypeImageSeriesScalarReader (0x144065b0) returned failure for request: vtkInformation (0x173a3990)
  Debug: Off
  Modified Time: 211076
  Reference Count: 1
  Registered Events: (none)
  Request: REQUEST_DATA
  FROM_OUTPUT_PORT: 0
  ALGORITHM_AFTER_FORWARD: 1
  FORWARD_DIRECTION: 0




vtkMRMLStorageNode::ReadData: Failed to read node FILENAME (vtkMRMLMultiVolumeNode1) from filename=PATH


ReadData: This is not a nrrd file


vtkMRMLStorageNode::ReadData: Failed to read node FILENAME  (vtkMRMLDiffusionWeightedVolumeNode1) from filename=PATH


vtkMRMLVolumeArchetypeStorageNode::ReadDataInternal: Cannot read file as a volume of type DiffusionTensorVolume [fullName = PATH]: FileFormatError. Number of files listed in the node = 0. File reader says it was able to read 530 files. File reader used the archetype file name of PATH [reader 0th file name = PATH].


vtkMRMLStorageNode::ReadData: Failed to read node FILENAME (vtkMRMLDiffusionTensorVolumeNode1) from filename=PATH


ReadData: This is not a nrrd file


vtkMRMLStorageNode::ReadData: Failed to read node FILENAME (vtkMRMLVectorVolumeNode1) from filename=PATH


vtkMRMLVolumeArchetypeStorageNode::ReadDataInternal: Failed to instantiate a file reader


vtkMRMLStorageNode::ReadData: Failed to read node FILENAME (vtkMRMLVectorVolumeNode2) from filename=PATH


vtkMRMLVolumeArchetypeStorageNode::ReadDataInternal: Cannot read file as a volume of type Volume [fullName = PATH]: FileFormatError. Number of files listed in the node = 0. File reader says it was able to read 530 files. File reader used the archetype file name of PATH [reader 0th file name = PATH].


vtkMRMLStorageNode::ReadData: Failed to read node FILENAME (vtkMRMLScalarVolumeNode1) from filename=PATH

I have redacted full filenames and file paths for privacy reasons, but they are all files named as follows {NAME}-{133-650}. It seems like Slicer is trying to load the entire series of images in the folder instead of the single one MONAI Label is providing.

MONAI Label's output looks like:

[2025-01-16 09:07:24,825] [28630] [MainThread] [INFO] (monailabel.endpoints.activelearning:44) - Active Learning Request: {'strategy': 'random', 'client_id': 'user-xyz'}

[2025-01-16 09:07:24,840] [28630] [MainThread] [INFO] (monailabel.tasks.activelearning.random:47) - Random: Selected Image: FILENAME; Weight: 1737036444

[2025-01-16 09:07:24,846] [28630] [MainThread] [INFO] (monailabel.endpoints.activelearning:60) - Next sample: {'id': 'FILENAME', 'weight': 1737036444, 'path': PATH, 'ts': 1737036220, 'name': FILENAME}

Any help with solving this bug would be appreciated, as MONAI Label is an incredibly useful tool, made useless if I can't fetch samples.

[diazandr3s]

Hi @landynfrancis,

Thanks for sharing this.

It seems this error comes from the image conversion:

ReadData: This is not a nrrd file

Could you explain how this conversion was done?

Also, are you trying to use the Radiology App for this? Default Segmentation model within the Radiology App is for 3D images.

You could use SAM2 instead once the image conversion is fixed. Here is a similar thread: #1809

Hope this helps,

[landynfrancis]

Hello @diazandr3s

When I was trying to conversion it was done through Python script that would extract the data array and use other libraries (tifffile, pynrrd, etc.) to save the new file.

As far as the app goes, I am using my own app, loosely based on the Radiology App. My app has multiple models available to use under it's config, for either 2D or 3D models based on how we launch the server itself.