HenryJaQiu / Diff-UNet

Diff-UNet: A Diffusion Embedded Network for Volumetric Segmentation. (using diffusion for 3D medical image segmentation)

Diff-UNet

Diff-UNet: A Diffusion Embedded Network for Volumetric Segmentation. Submitted to MICCAI2023.

https://arxiv.org/pdf/2303.10326.pdf

We design the Diff-UNet applying diffusion model to solve the 3D medical image segmentation problem.

Diff-UNet achieves more accuracy in multiple segmentation tasks compared with other 3D segmentation methods.

dataset

We release the codes which support the training and testing process of two datasets, BraTS2020 and BTCV.

BraTS2020(4 modalities and 3 segmentation targets): https://www.med.upenn.edu/cbica/brats2020/data.html

BTCV(1 modalities and 13 segmentation targets): https://www.synapse.org/#!Synapse:syn3193805/wiki/217789

Once the data is downloaded, you can begin the training process. Please see the dir of BraTS2020 and BTCV.