Mo-Diff

This repo is the official implementation for: Temporal Differential Fields for 4D Motion Modeling via Image-to-Video Synthesis (MICCAI 2025)

Comparison between our model and existing methods

Dataset Link

ACDC

4D Lung

Preprocess

We follow the preprocessing strategy in UVI-Net to do patch cropping and intensity normalization.

Requirements

• python 3.9
• pytorch 1.8.1
• torchvision 0.9.1
• simpleitk 2.0.2
• monai 0.9.0
• medpy 0.4.0

Usage

If you want to train the model from scratch, please follow the next steps.

1. To acquire the temporal differential fields via the given "temporal_field_synthesis" folder. Please run "python Main.py"
2. To conduct the image-to-video synthesis with temporal fields. Please run "python Main.py" of the "conditional_video_synthesis" folder

Citation

If you use our code in your work or find it is helpful, please cite the corresponding paper:

@article{you2025temporal,
  title={Temporal Differential Fields for 4D Motion Modeling via Image-to-Video Synthesis},
  author={You, Xin and Zhang, Minghui and Zhang, Hanxiao and Yang, Jie and Navab, Nassir},
  journal={International Conference on Medical Image Computing and Computer Assisted Intervention},
  year={2025}
}