TOF-MRA-Synth

MICCAI 2025 early accpet

Official code for Diffusion-based Multi-modal MR Fusion for TOF-MRA Image Synthesis

Data preprocessing

All the data was registered first by SimpleITK

We use hdf5 as the data storage method

• data: 3D data array (shape: z, x, y)
• max: max intensity
• max_len: num of slice (Axial)

Usage

Requirements

pip install -r requirements.txt

Data storage

/hdf5_root/case_names/modality.dhf5  
/dcm_root/case_name/...  # TOF only, for dicom templete 

'modality' contains '3D-T1W', '3D-T2W', '3D-FLAIR', '3D-TOF-MRA'

Train

python train.py

options.yaml: Adjust parameters

3data.json: data split

Inference

python test_all.py
options.data_root: hdf5_root
options.dcm_root: dicom_root

Acknowledgements

Thanks to Shanghai Sixth People's Hospital and Subtle Medical Inc.

Our code is based on BBDM