IDC/TCIA session at the RSNA 2021 Deep Learning Lab

This past week IDC had in person (!!!) presence at the 2021 RSNA Annual meeting. We had a poster at the educational exhibit, and I was teaching a session as part of the 2021 RSNA Deep Learning Lab about IDC and TCIA co-organized with Justin Kirby @kirbyju (unfortunately Justin was not able to attend in person). We had great attendance of the session, and it was wonderful to finally be able to present IDC in person!

If you registered for RSNA 2021, you should still be able to access the recorded sessions. But if not - we welcome you to explore the Google Colab notebook that we presented at the session: This notebook evolved from the materials presented earlier at MICCAI 2021 by Dennis Bontempi @denbonte, and was developed in collaboration between Justin, Dennis and me.

This notebook will walk you through the steps how to use Google Colab VM to

  • install a pretrained nnU-Net model for segmenting abdominal structures in CT
  • fetch a chest CT series from IDC and prepare it for processing
  • apply segmentation model, visualize results and compare with the existing segmentations from IDC side-by-side
  • utilize IDC metadata tables to find more chest CT series, and explore the various metadata attributes such as Manufacturer, to investigate generalizability of the tool.

Would be great to hear your thoughts about this notebook! If there are other tools you would like to see applied to IDC data, please let us know.

1 Like