OmniMRI

• Introduction
• Getting Started
• Publication
• Contacts

Introduction

OmniMRI is a vision–language foundation model that unifies the full MRI workflow, including reconstruction, segmentation, detection, diagnosis, and report generation, within a single multimodal architecture. Trained on 60 public datasets with 220,000+ MRI volumes and 19M slices, OmniMRI integrates imaging and clinical language through a multi-stage training paradigm, enabling inference across anatomies, contrasts, and tasks.

📄 Read the paper on arXiv

Getting Started

Coming soon...

Publication

@misc{he2025omnimriunifiedvisionlanguagefoundation,
      title={OmniMRI: A Unified Vision--Language Foundation Model for Generalist MRI Interpretation}, 
      author={Xingxin He and Aurora Rofena and Ruimin Feng and Haozhe Liao and Zhaoye Zhou and Albert Jang and Fang Liu},
      year={2025},
      eprint={2508.17524},
      archivePrefix={arXiv},
      primaryClass={cs.CV},
      url={https://arxiv.org/abs/2508.17524}, 
}

Contacts

Intelligent Imaging Innovation and Translation Lab [github] at the Athinoula A. Martinos Center of Massachusetts General Hospital and Harvard Medical School

• Xingxin He (xihe2@mgh.harvard.edu)
• Fang Liu (fliu12@mgh.harvard.edu)

149 13th Street, Suite 2301 Charlestown, Massachusetts 02129, USA

For specific code requests, please contact the authors.