Open wasserth opened 1 year ago
Hi Jakob,
First of all, thanks for all your work on TotalSegmentator. Please let me know if you have any ideas for Comp2Comp, which as you know, builds on TotalSegmentator.
Sorry for the error! I think we had different names in different branches that we merged in yesterday and that caused the issue. We need to add some unit tests to make sure this kind of thing doesn't happen.
Best regards, Louis
I tried the contrast_phase classifier and it worked great for a few cases I tried it on. I thought about building something similar, but never found time to do so. The runtime was quite long for me. I think there could be potential to reduce it. Probably a lower number of radiomics features would also work fine without reducing Accuracy too much. You could check which ones are slow and see if you can remove those. I always use pyradiomics for calculating features. Maybe that is faster. But I did not check that.
Connecting @edreisMD who developed this pipeline
Thank you Louis. Hi @wasserth. Glad it was useful for you, thank you so much for the thoughts. We just fixed a bug that was happening in some cases, but this shouldn't reduce the runtime yet.
Hi, I just tried to run Comp2Comp and I experienced the following error:
I fixed it by renaming
get_dicom_nifti_paths_and_num
toget_dicom_or_nifti_paths_and_num
.