pyradiomics
pyradiomics copied to clipboard
AIM-Harvard
General features contained complex numbers
Describe the bug We ran pyradiomics extraction on segmentations produced by TotalSegmentator on NLST CT data. While rounding and saving general features, we found that there are complex numbers in a few cases. We do not necessarily know if is normal to have complex numbers or not.
To Reproduce Steps to reproduce the behavior: A colab notebook with the image series in question can be found here: https://colab.research.google.com/drive/1BzgkUhdeSNmZWnswHkde9z2On8tlEcGR?usp%3Dsharing&sa=D&source=docs&ust=1709739967232788&usg=AOvVaw09YmF5IMwwrCuIIN3S9NSP
Version (please complete the following information):
- OS: Ubuntu
- Python version: 3.11
- PyRadiomics version 3.1.0
@vkt1414 can you be a little more specific? The notebook you linked to is very long. To help you get help, can you narrow down a specific example of a feature that results in a complex number? Maybe include a link to the cell?
@pieper I had the same comment but held it back - this was in the context of TotalSegmentator application to NLST, and I think the easiest would be to wait for ~2 weeks or so so that we have segmentations in IDC, and after that we can have a very minimal example to replicate the issue. I do not think there is anyone eager to jump into investigating this at the moment anyway. I encouraged Vamsi to document the issue just so that we have a record of this happening.
No rush on my end. I was just curious to see if there was an easy answer and ran into this issue so I thought I'd give some feedback.
Thanks @pieper! I do not know which general feature contains the complex number yet. I can drill it down and will update the issue in a few weeks as @fedorov suggested.