Pathology-Aware MRI to PET Cross-modal Translation with Diffusion Models

Official Pytorch Implementation of Paper - 🍝 PASTA: Pathology-Aware MRI to PET Cross-modal Translation with Diffusion Models

🎉 PASTA has been early-accepted at MICCAI 2024 (top 11%)!

Installation

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

Data

We used data from Alzheimer's Disease Neuroimaging Initiative (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 6, containing a list of non-image clinical data, including age, gender, education, MMSE, ADAS-Cog-13, ApoE4.
   4. A string attribute DX containing the diagnosis labels: CN, Dementia or MCI, if available.
   5. A scalar attribute RID with the patient ID, if available.
   6. A string attribute VISCODE with ADNI's visit code.

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

/stats/tabular           Group
/stats/tabular/columns   Dataset {6}
/stats/tabular/mean      Dataset {6}
/stats/tabular/stddev    Dataset {6}

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

Usage

The package uses PyTorch. To train and test PASTA, execute the train_mri2pet.py script. The configuration file of the command arguments is stored in src/config/pasta_mri2pet.yaml. The essential command line arguments are:

• --data_dir: Path prefix to HDF5 files containing either train, validation, or test data.
• --results_folder: Path to save all the training/testing output.
• --model_cycling: True to conduct cycle exchange consistency.
• --eval_mode: False for training mode and True for evaluation mode. The model for evaluation is specified in results_folder/model.pt.
• --synthesis: True to save all generated images during evaluation.

After specifying the config file, simply start training/evaluation by:

python train_mri2pet.py

Contacts

For any questions, please contact: Yitong Li (yi_tong.li@tum.de)

Acknowlegements

The codebase is developed based on lucidrains/denoising-diffusion-pytorch and openai/guided-diffusion.

If you find this repository useful, please consider giving a star 🌟 and citing the paper:

@InProceedings{Li2024pasta,
    author="Li, Yitong
    and Yakushev, Igor
    and Hedderich, Dennis M.
    and Wachinger, Christian",
    title="PASTA: Pathology-Aware MRI to PET Cross-Modal Translation with Diffusion Models",
    booktitle="Medical Image Computing and Computer Assisted Intervention -- MICCAI 2024",
    year="2024",
    publisher="Springer Nature Switzerland",
    address="Cham",
    pages="529--540",
    isbn="978-3-031-72104-5"
}