5k5000/DeepClosing

Deep Closing: Enhancing Topological Connectivity in Medical Tubular Segmentation

The repo of our paper 'Deep Closing: Enhancing Topological Connectivity in Medical Tubular Segmentation'

Environment

conda create -n DeepClosing python=3.10
conda activate DeepClosing
pip install torch==1.12.0+cu116 torchvision==0.13.0+cu116 torchaudio==0.12.0 --extra-index-url https://download.pytorch.org/whl/cu116
pip install pytorch-lightning==1.5.0
pip install monai==0.9.0
pip install scikit-image
pip install wandb
pip install nibabel

Quick Reference:

To begin with, the proposed framework,Deep Closing, consists of two operation:

(1) Deep Dilation:

DeepClosing/DeepClosing.py

Line 289 in c7c5ce9

def DeepDilation(self,T,is_infer_sliding_window=True, sw_roi_size=(224,224),sw_batch_size=4,verbose=False):
(2) Simple Component Erosion:

DeepClosing/DeepClosing.py

Line 259 in 2e7cb06

def Simple_Component_Erosion(self, T, M_T):
(*) DeepClosing = DeepDilation + Simple Component Erosion (Inference):

DeepClosing/DeepClosing.py

Line 313 in 2e7cb06

def DeepClosing(self,T,is_infer_sliding_window=True, sw_roi_size=(224,224),sw_batch_size=4,verbose=False):

The implementation of the proposed Simple Point Erosion Module is presented in the position below:

DeepClosing/DeepClosing.py

Line 313 in 2e7cb06

    
           def DeepClosing(self,T,is_infer_sliding_window=True, sw_roi_size=(224,224),sw_batch_size=4,verbose=False):

Besides, the Masked Shape Reconstruction (Training Stage) is presented in the position below:

DeepClosing/DeepClosing.py

Line 30 in 2e7cb06

def Masked_Shape_Reconstruction(config_path, device = torch.device("cuda:0")):

todo

We plan to provide more detailed information after the acceptance of our paper. Thanks for your constructive comments to help us improve our paper.