DILRAN for medical image fusion

''An attention-based Multi-Scale Feature Learning Framework for Multimodal Medical Image Fusion'' Paper link

The extended version of this work is accepted by IEEE BIBM 2024. The code is based on this repo and will be released at Here

Usage

You may want to change to your own dataset. If you have a 3-channel PET or SPECT image, you may want to change the dataset_loader.py file

To train the network, run

python3 ./train_with_val.py --batch_size 4 --epochs 100 --lambda1 0.2 --lambda2 0.2

To see the full list of parameters, run

python3 ./train_with_val.py -h

To evaluate the results, run

python3 ./inference.py

If you are using a different model, you may have to modify a little bit of the code.

Comment out anything related to wandb in the code if you do not want to use it to visualize the result.

Citation

@article{zhou2022attention,
  title={An Attention-based Multi-Scale Feature Learning Network for Multimodal Medical Image Fusion},
  author={Zhou, Meng and Xu, Xiaolan and Zhang, Yuxuan},
  journal={arXiv preprint arXiv:2212.04661},
  year={2022}
}