AttributeError: 'NoneType' object has no attribute '_fixed_uid' about raidionics

[dbouget]

Hi @remogg, which pipeline were you trying to run and for which tumor type?

If you can try to use the test data provided in Raidionics, and run either Preoperative reporting or Preoperative segmentation from the Actions tab and select glioblastoma as tumor type?

From the error message, it looks like a T1-w MR image is expected for timestamp 0 (which is assumed to be preoperatively), but hasn't been provided. It is unlikely an issue with DICOM headers, as long as you can see all the images you've loaded into Raidionics appearing inside the right-hand side, each tagged with the proper sequence type.

[andreped]

Hello, @remogg!

I noticed that you closed the Issue. Did you mean to do that? If yes, could you report how you solved the issue, as it might be of interest to others.

Do you have an MRI in NIfTI format we could get access to in order to reproduce the issue?

Regardless, @dbouget should be able to assist you further.

[remogg]

Hello @andreped,
i accidentally opened 2 tickets with the same content so decided to close this one.

"Do you have an MRI in NIfTI format we could get access to in order to reproduce the issue?" - yes, i could send you an example.

[andreped]

i accidentally opened 2 tickets with the same content so decided to close this one.

Oh, OK. In that case, to keep the conversation, I closed the other one and reopened this one :)

"Do you have an MRI in NIfTI format we could get access to in order to reproduce the issue?" - yes, i could send you an example.

You can see my email from my GitHub user profile if you wish to send me a NIfTI file over email. If it is a public dataset, you could upload it to a google drive and give us access through a URL here. That way @dbouget who likely will debug this further, can get access directly. Otherwise, I can redirect the sample file to him.

---

EDIT: Perhaps @dbouget can follow this up.

[remogg]

Thank you for the fast reply @dbouget.

From the error message, it looks like a T1-w MR image is expected for timestamp 0 [...].

I provided T1w, T2w, FLAIR, T1ce. All Sequences were recognized correctly, except the T1ce, so I selected the correct sequence type manually for it. Then i ran preopreative reporting and selected Gliobloastoma.

If you can try to use the test data provided in Raidionics, and run either Preoperative reporting or Preoperative segmentation from the Actions tab and select glioblastoma as tumor type?

When trying to run the test data provided i get error message 2 with ValueError: No radiological volume for {'timestamp': 0, 'sequence': 'T1-w', 'labels': None, 'space': {'timestamp': 0, 'sequence': 'T1-w'}}.

This is the same error message i get when using my DICOM-data converted into niftii using mricron.

[dbouget]

Tumor sub-compartments segmentation

I am assuming that you are trying to get the segmentation for the tumor sub-compartments (i.e., tumor, necrosis, and edema), which is the reason why you would need to provide a T1w MRI as input, right? So you have selected Tumor, Necrosis, Edema as Output classes in the Settings under Settings > Preferences > Processing - Segmentation?

As it stands right now, and as I might have hastily released a new version of the software to support the segmentation of sub-compartments, the reporting pipeline should be executed only when running a model that segments just the tumor (i.e., Output classes settings set as Tumor).

That being said, the glioblastoma segmentation that you attempted with your own data should have worked, and it seems that the issue is being triggered during the co-registration of the T1-w MR scan to the T1-CE MR scan. I will look into this issue further as I have so far used only data that were already co-registered, which is a bit lazy of me!

Running the test data

The provided test data are just compatible with the tumor segmentation only models, hence the lack of all four input MR sequences, which explains the error message when trying to run a tumor sub-compartments segmentation model.
I have made a new tutorial video showcasing how to use the test data to generate a glioblastoma segmentation, which assumes that the Output classes settings is set as Tumor.