Tutorial for 3D classification with IDC and Google Cloud

Dear all,

In the Radiology Informatics Lab at Mayo Clinic (MN), we have been working on developing a couple of tutorial notebooks to work with IDC database. Like the previous tutorial notebooks my colleagues have already developed, I would like to share another notebook that I have been working on recently:

This repository includes a sample Google Cloud notebook that shows how to train a 3D classification model on Imaging Data Commons (IDC) buckets.
Here are what I have covered in this notebook:

  • How to connect Google Cloud with IDC.
  • How to retrieve the DICOMs metadata from IDC using BigQuery commands.
  • How to download DICOMs from IDC.
  • How to change DICOMs to NIFTIs in the clouds.
  • How to pre-process data and train a 3D model on it using MONAI and Pytorch.

In case of any issues, please contact me in discourse or via this email address:

Hope you find it helpful!


Thanks a lot @po.rouzrokh! It’s on my list to go over the notebooks you guys contributed and look in more detail, but this is definitely very helpful! I now linked this notebook from our docs here: Colab notebooks - IDC User Guide.

If you have any comments based on your experience interacting with IDC, please let us know!