Official Pytorch implementation of Medical Image Segmentation via Cascaded Attention Decoding, WACV 2023.
GAIN 2023 best poster award
Md Mostafijur Rahman, Radu Marculescu

The University of Texas at Austin

Python 3.8
Pytorch 1.11.0
torchvision 0.12.0

Please use pip install -r requirements.txt to install the dependencies.

• Synapse Multi-organ dataset: Sign up in the official Synapse website and download the dataset. Then split the 'RawData' folder into 'TrainSet' (18 scans) and 'TestSet' (12 scans) following the TransUNet's lists and put in the './data/synapse/Abdomen/RawData/' folder. Finally, preprocess using python ./utils/preprocess_synapse_data.py or download the preprocessed data and save in the './data/synapse/' folder. Note: If you use the preprocessed data from TransUNet, please make necessary changes (i.e., remove the code segment (line# 88-94) to convert groundtruth labels from 14 to 9 classes) in the utils/dataset_synapse.py.

• ACDC dataset: Download the preprocessed ACDC dataset from Google Drive of MT-UNet and move into './data/ACDC/' folder.

• Polyp datasets: Download training and testing datasets Google Drive and move them into './data/polyp/'.

You should download the pretrained PVTv2 model from Google Drive, and then put it in the './pretrained_pth/pvt/' folder for initialization.

Download Google pretrained ViT models (R50-ViT-B_16, ViT-B_16, ...) from Google Cloud or use wget https://storage.googleapis.com/vit_models/imagenet21k/{MODEL_NAME}.npz, and then put them in the './pretrained_pth/vit/imagenet21k/' folder for initialization.

cd into CASCADE

For Polyp training run CUDA_VISIBLE_DEVICES=0 python -W ignore train_polyp.py

For Synapse Multi-organ training run CUDA_VISIBLE_DEVICES=0 python -W ignore train_synapse.py

For ACDC training run CUDA_VISIBLE_DEVICES=0 python -W ignore train_ACDC.py

cd into CASCADE 

For Polyp testing run CUDA_VISIBLE_DEVICES=0 python -W ignore test_polyp.py

For Synapse Multi-organ testing run CUDA_VISIBLE_DEVICES=0 python -W ignore test_synapse.py

For ACDC testing run CUDA_VISIBLE_DEVICES=0 python -W ignore test_ACDC.py

We are very grateful for these excellent works PraNet, Polyp-PVT and TransUNet, which have provided the basis for our framework.

@InProceedings{Rahman_2023_WACV,
    author    = {Rahman, Md Mostafijur and Marculescu, Radu},
    title     = {Medical Image Segmentation via Cascaded Attention Decoding},
    booktitle = {Proceedings of the IEEE/CVF Winter Conference on Applications of Computer Vision (WACV)},
    month     = {January},
    year      = {2023},
    pages     = {6222-6231}
}