Robust Whole-body PET Image Denoising Using 3D Diffusion Models: Evaluation Across Various Scanners, Tracers, and Dose Levels

Boxiao Yu¹, Savas Ozdemir², Yafei Dong³, Wei Shao⁴, Tinsu Pan⁵, Kuangyu Shi⁶, Kuang Gong¹

¹J. Crayton Pruitt Family Department of Biomedical Engineering, University of Florida; ²Department of Raiology, University of Florida; ³Yale School of Medicine, Yale University; ⁴Department of Medicine, University of Florida; ⁵Department of Imaging Physics, University of Texas MD Anderson Cancer Center; ⁶Department of Nuclear Medicine, University of Bern

Purpose

Whole-body PET imaging plays an essential role in cancer diagnosis and treatment but suffers from low image quality. Traditional deep learning-based denoising methods work well for a specific acquisition but are less effective in handling diverse PET protocols. In this study, we proposed and validated a 3D Denoising Diffusion Probabilistic Model (3D DDPM) as a robust and universal solution for whole-body PET image denoising.

Method

The proposed 3D DDPM gradually injected noise into the images during the forward diffusion phase, allowing the model to learn to reconstruct the clean data during the reverse diffusion process. A 3D convolutional network was trained using high-quality data from the Biograph Vision Quadra PET/CT scanner to generate the score function, enabling the model to capture accurate PET distribution information extracted from the total-body datasets. The trained 3D DDPM was evaluated on datasets from four scanners, four tracer types, and six dose levels representing a broad spectrum of clinical scenarios.

Result

The proposed 3D DDPM consistently outperformed 2D DDPM, 3D UNet, and 3D GAN, demonstrating its superior denoising performance across all tested conditions. Additionally, the model’s uncertainty maps exhibited lower variance, reflecting its higher confidence in its outputs.

Installation

Step 1: Clone the Repository

git clone https://github.com/Miche11eU/PET-Image-Denoising-Using-3D-Diffusion-Model.git
cd PET-Image-Denoising-Using-3D-Diffusion-Model

Step 2: Create and activate the conda environment from the environment.yml file:

conda env create -f environment.yml
conda activate PET-3D-DDPM

Step 3: Download Pre-trained Models

Download the pre-trained model files from this link and place them into the ./checkpoint/ folder.

Note: This model is licensed under CC BY-NC-SA 4.0. Commercial use is prohibited.

Testing

Data Preparation

Before running the denoising script, modify the load_data_for_worker function in ./scripts/test.py to align with your data format and dimensions. This function is responsible for loading your low-dose PET data into the model.

Running the Denoising Script

We provide a shell script test_DDPM_3d_mpi.sh to facilitate the testing process.

Usage

• --base_samples: Path to the .npz files containing your low-dose PET images.
• --num_samples: Total number of samples you wish to process.
• -n: Number of GPUs to utilize for parallel processing.
• --save_dir: Path to the directory where you want to save the denoised images.

License

• The code in this repository is licensed under the MIT License.
• The model weights are licensed under CC BY-NC-SA 4.0, meaning:
  • You can share and modify the model weights, but must use the same license.
  • You cannot use it for commercial purposes.

For details, check:

• LICENSE (MIT for code)
• MODEL_LICENSE (CC BY-NC-SA 4.0 for model weights)

Citation

If you find our work is useful in your research or applications, please consider giving us a star 🌟 and citing it by the following BibTeX entry.

@article{yu2025robust,
  title={Robust whole-body PET image denoising using 3D diffusion models: evaluation across various scanners, tracers, and dose levels},
  author={Yu, Boxiao and Ozdemir, Savas and Dong, Yafei and Shao, Wei and Pan, Tinsu and Shi, Kuangyu and Gong, Kuang},
  journal={European Journal of Nuclear Medicine and Molecular Imaging},
  pages={1--14},
  year={2025},
  publisher={Springer}
}

Contact

For any questions or inquiries, please contact boxiao.yu@ufl.edu.