RoentGen is a domain-adapted latent diffusion model capable of generating high-quality, text-conditioned chest X-rays (CXRs). This repository contains the code and resources for our paper "A vision–language foundation model for the generation of realistic chest X-ray images" published in Nature Biomedical Engineering.
RoentGen addresses the challenge of generating realistic medical images by fine-tuning a large, general domain-pre-trained vision-language latent diffusion model (Stable Diffusion) to the medical domain. Our model demonstrates the ability to generate high-fidelity CXRs controllable with on free-form medical text prompts.
For more details, visit our project page.
Access to the RoentGen model weights is provided to researchers credentialed for MIMIC-CXR access. To request access, please fill out this form.
Please refer to the supplementary materials of the paper for a model card for RoentGen v1.0.
If you use RoentGen in your research, please cite our paper:
@Article{Bluethgen2024,
author={Bluethgen, Christian
and Chambon, Pierre
and Delbrouck, Jean-Benoit
and van der Sluijs, Rogier
and Po{\l}acin, Ma{\l}gorzata
and Zambrano Chaves, Juan Manuel
and Abraham, Tanishq Mathew
and Purohit, Shivanshu
and Langlotz, Curtis P.
and Chaudhari, Akshay S.},
title={A vision--language foundation model for the generation of realistic chest X-ray images},
journal={Nature Biomedical Engineering},
year={2024},
month={Aug},
day={26},
issn={2157-846X},
doi={10.1038/s41551-024-01246-y},
url={https://doi.org/10.1038/s41551-024-01246-y}
}