Cascaded U-Net for vessel segmentation

by Pierre Rougé, Nicolas Passat, Odyssée Merveille

Code associated with the paper Cascaded Multitask U-Net using topological loss for vessel segmentation and centerline extraction.

Installation

Install environment with environment.yml

conda env create -f environment.yml

Usage

Clone the repository

git clone https://github.com/PierreRouge/Cascaded-U-Net-for-vessel-segmentation.git
cd Cascaded-U-Net-for-vessel-segmentation

Put the MRA Images in data/Images, the segmentations GTs in data/GT and the skeletons GTs in data/Skeletons
Use one of the training function
1. To pretrain segmentation network
```
  cd train
  python train_segmentation.py
```
1. To pretrain skeletonization network
```
  cd train
  python train_skeletonization.py
```
1. To train cascaded U-Net : put the pretrained weights in respectively pretrained_weights/segmentation and pretrained_weights/skeletonization and run :
```
cd train
python train_cascaded_unet.py
```

Visual results

Future implementations

We are currently working to implement several state-of-the-art methods for vascular segmentation !

Citation

If you use this repository please consider citing :

@article{rouge2023cascaded,
      title={Cascaded multitask U-Net using topological loss for vessel segmentation and centerline extraction},
      author={Roug{\'e}, Pierre and Passat, Nicolas and Merveille, Odyss{\'e}e},
      journal={arXiv preprint arXiv:2307.11603},
      year={2023}
   }

Contact

Pierre Rougé : pierre.rouge@creatis.insa-lyon.fr