Exception on unknown segmentation algorithm type when loading DCM Seg (e.g. TCIA dataset "qiba_ct_1c")

[rfloca]

Currently DCMQI is not able to load nonconformant DCM Seg e.g. from TCIA dataset "qiba_ct_1c" as those DCM Segs have not algorithm type set.

That leads to an exception in Dicom2ItkConvert.cpp line 485ff.

Looking at the standard DCMQI is correct, as SAT has to be defined. Nevertheless it seems over pendantic, as there is no option to e.g. explicitly allow to load such a data and only bring a worning and replase the value e.g. with manual or a user defined choice. This ultimatly means to dismiss the data, even so the missing meta information is not relevant for all use cases (it would something different, e.g. with geometric information).

I would like to clarify if the current implementation should be the only behavior and if not, how to best adapt it.

Thanks,
Ralf

[fedorov]

Ralf, that's a fair point. I agree this is overly pedantic. I think it would be fine to set algorithmType to empty string in this situation. Would you like to make a PR? This way you could test it against your data, and see if there are any other similar issues in other places.

I have to warn you though that "qiba_ct_1c" is ancient, and I am not sure it was encoded correctly. There may be other issues. Here are OHIF v2 links from Imaging Data Commons for 6 studies from that collection that contain SEG. I tried to open one of those, and failed.

https://viewer.imaging.datacommons.cancer.gov/viewer/1.2.840.113704.1.111.2224.1298053869.2
https://viewer.imaging.datacommons.cancer.gov/viewer/1.2.840.113619.2.267.3.346830868.413.1298911364.892
https://viewer.imaging.datacommons.cancer.gov/viewer/1.2.840.113704.1.111.4032.1298392690.28
https://viewer.imaging.datacommons.cancer.gov/viewer/1.2.392.200036.9116.2.6.1.48.1215591467.1297222370.956181
https://viewer.imaging.datacommons.cancer.gov/viewer/1.2.840.113704.1.111.5740.1298487487.25
https://viewer.imaging.datacommons.cancer.gov/viewer/1.3.12.2.1107.5.1.4.54023.30000011021001533310900000001

[rfloca]

Thanks for the warning. 😄
It is not important, that qiba_ct_1c will work out in the end. But an bug ticket for MITK triggered the investigation, and at least for the raised point, I thought it would do no harm to be less pedantic or at least offer the option to be less pedantic.

I will have a look regarding the PR.