Datasets:

Modalities:
Text
ArXiv:
License:

Different preprocessing scheme

#4
by TurquoiseA - opened

Hi there, thanks for sharing your wonderful work! I have some questions about your preprocessing scheme. I noticed several differences between the original CT-CLIP preprocessing script and yours and hopefully you can help me understand. Here are some changes I noticed but please feel free to add if I missed anything!

  1. The spacing parameters here: https://huggingface.co/datasets/RadGenome/RadGenome-ChestCT/blob/main/processed_code/preprocess_ctrate_train.py#L104
  2. The addition of monai transformation: https://huggingface.co/datasets/RadGenome/RadGenome-ChestCT/blob/main/processed_code/preprocess_ctrate_train.py#L82
  3. An additional transpose to the axial view: https://huggingface.co/datasets/RadGenome/RadGenome-ChestCT/blob/main/processed_code/preprocess_ctrate_train.py#L119
  4. Exclusion of HU value scale setting.

Thanks again!

RadGenome org

Yes, we made these changes (spacing, transformation..) to enable the CT-rate data to be used as input to SAT model.

Sign up or log in to comment