This repository contains PyTorch code for the preprint "Joint Reconstruction of Activity and Attenuation in PET by Diffusion Posterior Sampling in Wavelet Coefficient Space" (2025) [arXiv link]

If you use this code in your research, please consider citing:

@article{phung2025joint,
  title={Joint Reconstruction of Activity and Attenuation in PET by Diffusion Posterior Sampling in Wavelet Coefficient Space},
  author={Phung-Ngoc, Cl{\'e}mentine and Bousse, Alexandre and De Paepe, Antoine and Merlin, Thibaut and Laurent, Baptiste and Dang, Hong-Phuong and Saut, Olivier and Visvikis, Dimitris},
  journal={arXiv preprint arXiv:2505.18782},
  year={2025}
}

conda env create -f environment.yml
conda activate jraadpsenv

Data is preprocessed from DICOM patient files into 128x256x256 volumes stored in .npy format for easy dataloading with ActivityAttenuationDataset.

python train.py \
    --data configs/data.yml \
    --model configs/model.yml \
    --train configs/train.yml

python reconstruct.py \
    --data configs/data.yml \
    --model configs/model.yml \
    --proj configs/projector.yml \
    --infer configs/inference.yml

Code is base on the following repositories:

• https://github.com/pfriedri/wdm-3d.