The contents of this archive are files needed to build and train the U-Net model with 3D convolutions, and to make needed visualizations. To run the code, you need the data provided by the Open-KBP challenge, which can be fetched from here. The provided-data folder needs to be in the same directory as the code files. The files found in provided_code directory were provided by the Open-KBP challenge to help with reading in the data.
This is the code that reads in all of the data, defines the U-Net model, and trains it. It also gets the test dose scores (MAE) and makes visualizations of the CT image and ground-truth or prediction dose (relies on mayavi).
This auxiliary code draws the learning curves of some of the configurations tested. It needs the pickled loss files included in the archive.
This auxiliary code fetches a sample patient from the training data and draws its contoured CT image as well as their ground-truth dose (relies on mayavi).