If you would like to evaluate the data on the segmentation model, you can download the segmentation code using the following command:
git clone https://github.com/MrGiovanni/DiffTumor/tree/main/STEP3.SegmentationModel
You will need to modify the appropriate command-line arguments in STEP3.SegmentationModel\main.py and adjust the image preprocessing steps as needed. Additionally, you should update the dataset loading process in STEP3.SegmentationModel\main.py to fit your own dataset.
In this codebase, there are three models available for selection: UNET, SwinUNETR, and nnUNet. Our evaluation is based on SwinUNETR and nnUNet, but you are free to choose any model that suits your segmentation needs. We did not use pre-trained weights in our evaluation; however, if you wish to use a pre-trained SwinUNETR model, you can download the corresponding pre-trained model and place the weights in STEP3.SegmentationModel\pretrained_model.