A high-performance, CUDA-accelerated library for circular orbits CT reconstruction with end-to-end differentiable operators, enabling advanced optimization and deep learning integration.

⭐ Please star this project if you find it is useful!

This is the stable version supporting circular trajectory CT reconstruction.

The dev branch includes experimental features:

• Random trajectory projection and backprojection operators
• New examples with non-circular trajectories

⚠️ Note: The dev branch is under active development. If you find any bugs, please raise an issue.

• Fast: CUDA-accelerated projection and backprojection operations
• Differentiable: End-to-end gradient propagation for deep learning workflows

• Parallel Beam: 2D parallel-beam geometry
• Fan Beam: 2D fan-beam geometry
• Cone Beam: 3D cone-beam geometry

diffct/
├── diffct/
│   ├── __init__.py            # Package initialization
│   ├── differentiable.py      # Differentiable CT operators
├── examples/                  # Example usages
│   ├── fbp_parallel.py
│   ├── fbp_fan.py
│   ├── fdk_cone.py
│   ├── iterative_reco_cone.py
│   ├── iterative_reco_fan.py
│   ├── iterative_reco_parallel.py
├── pyproject.toml             # Project metadata
├── README.md                  # README
├── LICENSE                    # License
├── requirements.txt           # Dependencies

• CUDA-capable GPU
• Python 3.10+
• PyTorch, NumPy, Numba, CUDA

CUDA 12 (Recommended):

# Create and activate conda environment
conda create -n diffct python=3.12
conda activate diffct

# Install CUDA (here 12.8.1 as example) PyTorch, and Numba
conda install nvidia/label/cuda-12.8.1::cuda-toolkit

# Install Pytorch, you can find the commend here: https://pytorch.org/get-started/locally/

# Install Numba with CUDA 12
pip install numba-cuda[cu12]

# Install diffct
pip install diffct

# Create and activate conda environment
conda create -n diffct python=3.12
conda activate diffct

# Install CUDA (here 13.0.2 as example) PyTorch, and Numba
conda install nvidia/label/cuda-13.0.2::cuda-toolkit

# Install Pytorch, you can find the commend here: https://pytorch.org/get-started/locally/

# Install Numba with CUDA 13
pip install numba-cuda[cu13]

# Install diffct
pip install diffct

# Create and activate conda environment
conda create -n diffct python=3.12
conda activate diffct

# Install CUDA (here 11.8.0 as example) PyTorch, and Numba
conda install nvidia/label/cuda-11.8.0::cuda-toolkit

# Install Pytorch, you can find the commend here: https://pytorch.org/get-started/locally/

# Install Numba with CUDA 11
pip install numba-cuda[cu11]

# Install diffct
pip install diffct

If you use this library in your research, please cite:

@software{diffct2025,
  author       = {Yipeng Sun},
  title        = {diffct: Differentiable Computed Tomography
                 Reconstruction with CUDA},
  year         = 2025,
  publisher    = {Zenodo},
  doi          = {10.5281/zenodo.14999333},
  url          = {https://doi.org/10.5281/zenodo.14999333}
}

This project is licensed under the Apache 2.0 - see the LICENSE file for details.

This project was highly inspired by:

• PYRO-NN
• geometry_gradients_CT

Issues and contributions are welcome!