Didn't work with non-dicom data

[pieper]

When I tried running the module on a CT scan from Slicer's SampleData module it failed with this message in the console. It would be better to give a dialog warning that DICOM data is required and then fail gracefully. (or maybe it can work on non-DICOM data?).

Creating SAM predictor ... weights not found
Downloading SAM weights ... Done
Creating SAM predictor ... Done
Traceback (most recent call last):
  File "/home/exouser/Downloads/Slicer-5.3.0-2023-07-08-linux-amd64/slicer.org/Extensions-31860/TomoSAM/lib/Slicer-5.3/qt-scripted-modules/tomosam.py", line 492, in onPushEmbeddingsCreate
    instanceUIDs = volume_node.GetAttribute("DICOM.instanceUIDs").split()
AttributeError: 'NoneType' object has no attribute 'split'

[fedesemeraro]

So far we have only tested TomoSAM by importing 3D tif stacks into Slicer. If you convert DICOM into tifs then it should work. I am a bit short on time this week, but I would love your help if you want to extend this limitation to DICOM files

[pieper]

It looks like the issue is this code:

https://github.com/fsemerar/SlicerTomoSAM/blob/b9f5e6d92974292800cf67d93d3f638c35880f6a/TomoSAM/tomosam.py#L486-L497

When you load data from a file there can be a storage node but no DICOM.instanceUIDs so it should be easy to catch this case.

[pieper]

Just to report: I was able to download your tiff file of the CT and was able to calculate the embeddings. But when I add a point I get the following error. The loaded tiff file is selected as the volume node so I'm not sure what's going on.

Traceback (most recent call last):
  File "/home/exouser/Downloads/Slicer-5.3.0-2023-07-08-linux-amd64/slicer.org/Extensions-31860/TomoSAM/lib/Slicer-5.3/qt-scripted-modules/tomosam.py", line 292, in onMarkupIncludePointPositionDefined
    self.addPoint(caller, self.logic.include_coords)
  File "/home/exouser/Downloads/Slicer-5.3.0-2023-07-08-linux-amd64/slicer.org/Extensions-31860/TomoSAM/lib/Slicer-5.3/qt-scripted-modules/tomosam.py", line 353, in addPoint
    self.logic.get_mask(self.first_freeze)
  File "/home/exouser/Downloads/Slicer-5.3.0-2023-07-08-linux-amd64/slicer.org/Extensions-31860/TomoSAM/lib/Slicer-5.3/qt-scripted-modules/tomosamLib/tomosamLogic.py", line 182, in get_mask
    new_mask, _, _ = self.predictor.predict(point_coords=np.array(include_points + exclude_points),
  File "/home/exouser/Downloads/Slicer-5.3.0-2023-07-08-linux-amd64/lib/Python/lib/python3.9/site-packages/segment_anything/predictor.py", line 134, in predict
    raise RuntimeError("An image must be set with .set_image(...) before mask prediction.")
RuntimeError: An image must be set with .set_image(...) before mask prediction.

[fedesemeraro]

Thank you for reporting this. Does this happen with the precomputed embeddings as well or only when computing them within TomoSAM?

[pieper]

I'm not sure - I only tried computing them, not downloading.