Research paper: https://arxiv.org/abs/2508.16783
🧨Inference with diffusers
import torch
from diffusers import DiffusionPipeline
device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
pipe = DiffusionPipeline.from_pretrained("stanfordmimi/RoentGen-v2")
pipe = pipe.to(device)
prompt = "50 year old female. Normal chest radiograph."
image = pipe(prompt).images[0]
More info and instructions for use on GitHub.
🩻 Synthetic CXR Dataset
565k synthetic chest radiographs and associated text prompts: stanfordmimi/RoentGen-v2-synthetic-dataset
Important: The generated images are for research and educational purposes only and cannot replace real chest x-rays for medical diagnosis.
@misc{moroianu2025improvingperformancerobustnessfairness,
title={Improving Performance, Robustness, and Fairness of Radiographic AI Models with Finely-Controllable Synthetic Data},
author={Stefania L. Moroianu and Christian Bluethgen and Pierre Chambon and Mehdi Cherti and Jean-Benoit Delbrouck and Magdalini Paschali and Brandon Price and Judy Gichoya and Jenia Jitsev and Curtis P. Langlotz and Akshay S. Chaudhari},
year={2025},
eprint={2508.16783},
archivePrefix={arXiv},
primaryClass={cs.CV},
url={https://arxiv.org/abs/2508.16783},
}