xabimoreno/3D-LDMs-for-CT-MAR

Large-Volume Conditioned 3D Latent Diffusion Models for CT Metal Artifact Suppression

Official model weights for the paper: Large-Volume Conditioned 3D Latent Diffusion Models for CT Metal Artifact Suppression
Deep Generative Models for Medical Imaging (DGM4MICCAI), MICCAI 2026 Workshop (In press).
Authors: Xabier Moreno Casado, Jef Vandemeulebroucke, Jakub Ceranka
Organization: ETRO - Vrije Universiteit Brussel (VUB) & imec

GitHub Code: https://github.com/ETRO-MIT/3D-LDMs-for-CT-MAR

Checkpoints Included

Filename	Architecture	Description
vqvae_checkpoint.pth	Stage 1 3D VQ-VAE	4x spatial compression autoencoder
anatomy_ldm_checkpoint.pth	Model 2.1	3D LDM conditioned on artifacted CT
anatomy_metadata_ldm_checkpoint.pth	Model 2.2	3D LDM conditioned on artifacted CT + implant metadata

Usage

Clone the GitHub repository and download automatically:

git clone https://github.com/ETRO-MIT/3D-LDMs-for-CT-MAR.git
cd 3D-LDMs-for-CT-MAR
pip install -e ".[inference]"
python DownloadWeights.py

Citation

@inproceedings{casado2026largevolume,
  title={Large-Volume Conditioned 3D Latent Diffusion Models for CT Metal Artifact Suppression},
  author={Moreno Casado, Xabier and Vandemeulebroucke, Jef and Ceranka, Jakub},
  booktitle={Deep Generative Models for Medical Imaging (DGM4MICCAI), MICCAI Workshop},
  year={2026},
  note={In press}
}