README

1st Place Solution to MICCAI 2020 ADAM Challenge in Segmentation Task.

Step 1. Prepare input data and running environment

Put input data into the folder input. The folder structure should be

input
——10001
————orig
——————struct.nii.gz
——————...
————pre
——————TOF.nii.gz # we only use the preprocessed TOF data
——————...

Run

python 3.7.6; pytorch 1.5.0; cuda 10.1

cd nnUNet
pip install -e .
cd ..
# preprocess
python ADAM_Preprocess.py --input_path ./input
cd nnUNet/nnunet

Step 2. Inference

Download the trained models: link (PW: b9oo), and put it in the ADAM2020/nnUNet/nnunet/TrainedModels/nnUNet/3d_fullres

Run

nnUNet_predict -i ../../input_preprocess -o ../../seg_initial -m 3d_fullres -t Task600_ADAM --disable_tta
cd ../../

# postprocess
python ADAM_Postprocess.py --save_path ./output

The final segmentation results are in the folder output.