Diffusion Deformable Model (DDM)

The official source code of "Diffusion deformable model for 4D temporal medical image generation" which is presented in MICCAI 2022.

[MICCAI paper]

Implementation

A PyTorch implementation of deep-learning-based model.

• Requirements
  • OS : Ubuntu / Windows
  • Python 3.6
  • PyTorch 1.4.0

Dataset

• In our experiment, we used 4D cardiac MR scans provided by the Automated Cardiac Diagnosis Challenge (ACDC).
• You can download the data at ACDC.
• In our experiment, we resampled all MRI scans with a voxel spacing of 1.5 x 1.5 x 3.15 mm, and saved the data with .mat.
• Examples of the data we used can be downloaded at here.
  • Copy downloaded directory of "ACDC_dataset" to './data'.
  • The data in the directory "data_ED_ES" have 3D MR scans at the end diastolic and at the end systolic phases, and their corresponding segmentation labels.
  • The data in the directory "data_ED2ES" have 4D MR scans from the end diastolic to the end systolic phases, which is used in the test stage.

Training

• DDM_train.py is the implementation code for training the proposed diffusion deformable model.
• You can run the code through "sh train.sh" in terminal.
• All parameter settings we used are written in ./config/DDM_train.json file.

Evaluation

• DDM_test.py is the implementation code of inference for 4D temporal image generation.
• You can run the code by running "sh test.sh" in terminal.