Closed fedorov closed 9 months ago
Check out this pull request on
See visual diffs & provide feedback on Jupyter Notebooks.
Powered by ReviewNB
Sorry for the late review and in the interest of time, I did a couple of things:
if 'google.colab' in sys.modules: !cd /content && wget https://github.com/DCMTK/dcmtk/releases/download/DCMTK-3.6.6/dcmtk-3.6.6.tar.gz !bunzip dcmtk-3.6.6.tar.gz !tar -xvf dcmtk-3.6.6.tar !export DCMDICTPATH=/content/dcmtk-3.6.8-linux-x86_64-static/share/dcmtk-3.6.8/acrnema.dic
I do understand that it is used for validating the DICOM files but I do not see relevance in the production notebook. So, I'll remove the dcm2k
I actually strongly disagree with this. There are important benefits to keep that cell, and really no downsides.
The main issue is that at the moment dsrdump
fails with segmentation fault, and that's why I labeled that cell as WIP. But even with that deficiency, it is better than nothing - it's a starting point to look into why it fails and fix it, which should be feasible.
Instead of encoding algorithm details for each individual feature in radiomics features SR, do this at the level of the measurement group. This helps significantly reduce the size of the SR content sequence, which should help work around the 1MB SQ limit of Google Healthcare.
In the process of implementing this refinement, clean up the code to make it possible to complete the test runs much faster, and make the code a bit more robust.