Mamba? Catch the Hype or Rethink What Really Helps for Image Registration

This is the official Pytorch implementation of the paper @MICCAI2024@WBIR2024: "Mamba? Catch the Hype or Rethink What Really Helps for Image Registration (WBIR2024)"

---

TODOs

• Upload networks code
• Upload configuration files
  • Upload network configuration files
  • Upload data configuration files
  • Upload training configuration files
• Upload training and inference scripts
  • Upload training scripts
  • Upload inference scripts
  • Upload evaluation scripts
  • Test run on all scripts
    • Training scripts
    • Evaluation scripts
    • Numerical problem of Mamba block
• Upload dataloading scripts
• Upload pretrained model weights
• Upload error map and deformation plotting script
• Update README.md

---

Low-level Computational Blocks

• CNN (VoxelMorph)
• Transformer (TransMorph)
• Large-Kernel CNN (LKU-Net)
• Mamba (MambaMorph)

---

High-level Registration-specific Designs

• Dual Stream Encoders
• Motion Pyramid and Warping
• Correlation Layers
• Iterative Optimization

---

Dataset

Training

• OASIS
• ADNI
• IXI

Zero-shot Evaluation

• LPBA
• MindBoggle

---

Pretrained Model

---

---

Prerequisites

---

Training

---

Inference

---

Citation

If you find this repository useful in your research, please consider to cite use in your work by:

@inproceedings{jian2024mamba,
  title={Mamba? Catch The Hype Or Rethink What Really Helps for Image Registration},
  author={Jian, Bailiang and Pan, Jiazhen and Ghahremani, Morteza and Rueckert, Daniel and Wachinger, Christian and Wiestler, Benedikt},
  booktitle={International Workshop on Biomedical Image Registration},
  pages={86--97},
  year={2024},
  organization={Springer}
}

---

Acknowledgement

Many thanks to the following repositories for providing helpful resources to my work:

• VoxelMorph
• TransMorph
• SegMamba
• UMamba
• MONAI
• MMCV

---

Lincense & Copyright

© Bailiang Jian Licensed under the MIT Licensce