Anatomical Variation Modeling for Radiation Therapy

This repository contains data and code for training and testing generative models that predict anatomical variations across radiation therapy fractions, along with their impact on 3D dose distributions.

---

📋 Repository Overview

This project develops generative models to learn realistic anatomical changes that occur during multi-fraction radiation therapy. The models are conditioned on first fraction CT scans and generate plausible anatomical variants that match the distribution of real anatomical changes observed in subsequent treatment fractions.

---

📂 Repository Structure

.
├── code
│     ├── ddf/                          # Dense Deformation Field approach
│     ├── stable_diffusion/             # Stable Diffusion approach  
│     ├── train_test_data_split.py      # Data preparation script
│     ├── data_dict.json                # Train/test/fold assignments
│     └── README.md                     # readme
├── data
│     ├── CT/                           # Folder with patient subfolders, each subfolder containing fraction CTs in nifty format 
│     ├── DOSE/                         # Folder with patient subfolders, each subfolder containing planned dose in nifty format
│     ├── STRUCTURES/                   # Folder with patient subfolders, each subfolder containing segmented OARs and target
│     ├── prepareFractionCT.py          # Script for converting DICOMS to nifty
│     ├── prostaty.txt                  # TXT file with target names used in our database
│     └── README.md                     # readme
├── LICENSE                            # License statement
└── README.md                          # This file

🚀 Getting Started

Prerequisites

• Python 3.x
• SimpleITK
• PyTorch

---

🔑 Applications

This framework enables:

1. Robust treatment planning: Account for anatomical variations in dose optimization
2. Adaptive radiotherapy: Predict anatomical changes for plan adaptation
3. Uncertainty quantification: Quantify dose uncertainty due to anatomical variations
4. Quality assurance: Validate delivered dose against predicted variations
5. Research: Study inter-fraction anatomical changes and their dosimetric impact

---

📖 Citation

If you use this code or data in your research, please cite:

@article{your_paper,
  title={Generative Models for Anatomical Variation in Radiation Therapy},
  author={},
  journal={},
  year={2026},
  doi={}
}

---

📝 License

MIT

---

⚠️ Data Privacy

This repository structure is designed for research purposes. When working with patient data:

• Ensure all data is de-identified and compliant with HIPAA/GDPR
• Obtain appropriate IRB approval for your institution
• Follow institutional data sharing and privacy policies
• Do not include identifiable patient information in public repositories
• DICOM files must be anonymized before conversion to NIfTI

---

🤝 Contributing

Contributions are welcome! Areas for improvement:

• Additional generative model architectures
• Enhanced uncertainty quantification methods
• Integration with treatment planning systems
• Validation on additional datasets
• Improved DICOM parsing for different vendors

Please open an issue or submit a Pull Request.

---

📧 Contact

For questions or collaborations, please contact ztabor@agh.edu.pl