Official Implementation of Diffusion Bridge Networks Simulate Clinical-grade PET from MRI for Dementia Diagnostics.

1. Create environment: conda env create -n sim2p --file requirements.yaml
2. Activate environment: conda activate sim2p

We used public datasets from Alzheimer's Disease Neuroimaging Initiative (ADNI) and Japanese Alzheimer's Disease Neuroimaging Initiative (J-ADNI). Since we are not allowed to share our data, you would need to process the data yourself. Data for training, validation, and testing should be stored in separate HDF5 files, using the following hierarchical format:

1. First level: A unique identifier, e.g. image ID.
2. The second level always has the following entries:
   1. A group named MRI/T1, containing the T1-weighted 3D MRI data.
   2. A group named PET/FDG, containing the 3D FDG PET data.
   3. A dataset named tabular of size 13, containing a list of non-image clinical data, including age, gender, education level, MRI brain segmentation volumes obtained by Freesurfer including cerebrospinal fluid volume, the total grey matter volume, cortical white matter volume, left hippocampus volume, right hippocampus volume, left entorhinal thickness, right entorhinal thickness, cognitive examination scores MMSE, ADAS-Cog-13, and genetic risk factor ApoE4.
   4. A string attribute DX containing the diagnosis labels: CN, Dementia or MCI, if available.

Finally, the HDF5 file should also contain the following meta-information in a separate group named stats:

/stats/tabular           Group
/stats/tabular/columns   Dataset {13}
/stats/tabular/mean      Dataset {13}
/stats/tabular/stddev    Dataset {13}

They are the names of the features in the clinical data, their mean, and standard deviation.

We provide bash files train_sim2p.sh and test_sim2p.sh for model training and evaluation. Important model variables can be set in the bash file args.sh. For example, the length of the clincial data input can be accustomized and assigned to TAB_DIM.

To train, run

bash train_sim2p.sh

# to resume, set CKPT to your checkpoint, or it will automatically resume from your last checkpoint based on your experiment name.

bash train_sim2p.sh $CKPT

For evaluation, you need to be set MODEL_PATH for your checkpoint to be evaluated. Setting --save_syn_scans True will save the generated synthetic scans into your experiment folder. To evaluate, run

bash test_sim2p.sh $MODEL_PATH --save_syn_scans True

This script will print and also save the evaluation scores into .txt and .csv files into your experiment folder.

The codebase is inspired by alexzhou907/DDBM and DiT-3D/DiT-3D. Thanks for their wonderful works.

If you find this method and/or code useful, please consider giving a star 🌟 and citing the paper:

@article{li2025diffusion,
  title={Diffusion Bridge Networks Simulate Clinical-grade PET from MRI for Dementia Diagnostics},
  author={Li, Yitong and Buchert, Ralph and Schmitz-Koep, Benita and Grimmer, Timo and Ommer, Bj{\"o}rn and Hedderich, Dennis M and Yakushev, Igor and Wachinger, Christian},
  journal={arXiv preprint arXiv:2510.15556},
  year={2025}
}