README.md

3D D-LKA Net

Instructions for the 3D D-LKA Net.

Environment Setup

1. Create a new conda environment with python version 3.8.16:

   conda create -n "d_lka_net_3d" python=3.8.16
   conda activate d_lka_net_3d

2. Install PyTorch and torchvision

   pip install torch==1.11.0+cu113 torchvision==0.12.0+cu113 --extra-index-url https://download.pytorch.org/whl/cu113

3. Install the requirements with:

   pip install -r requirements.txt

4. Install 3D deformable convolutions.

   cd dcn/
   bash make.sh

Model weights

You can download the learned weights of the D-LKA-Net in the following table.

Task	Learned weights
Multi organ segmentation	D-LKA Net
Pancreas	D-LKA Net

Synapse Dataset

1. Download the Synapse dataset from here: Synapse
2. Rename each folder containing 'unetr_pp' to 'd_lka_former'. THIS IS IMPORTANT.
3. Adjust the paths in the run_training_synappse.sh
4. Run the following lines:

   cd 3D
   bash run_training_synapse.sh

5. After the training is finished, run the evaluation:

   run_evaluation_synapse.sh

For further instructions, refer to the nnFormer repository.

Pancreas Dataset

1. Download the pancreas dataset from here: dataset
2. The folder structure should be as follows:

   /pancreas_code
   --/dataset_pancreas
   ----/Pancreas
   ----/PANCREAS_0001.h5
   .
   .
   .
   ----/PANCREAS_82.h5

3. Adjust the paths in the train_pancreas.py file.
4. Run

   cd 3D/pancreas_code
   python train_pancreas.py

5. Test

   python test_pancreas.py