Medical World Model

Paper Code

Yijun Yang¹ Zhao-Yang Wang² Qiuping Liu³ Shuwen Sun³ Kang Wang⁴
Rama Chellappa² Zongwei Zhou² Alan Yuille² Lei Zhu^1,5,† Yu-Dong Zhang³ Jieneng Chen^2,†

¹ The Hong Kong University of Science and Technology (Guangzhou) ² Johns Hopkins University
³ The First Affiliated Hospital of Nanjing Medical University ⁴ University of California, San Francisco
⁵ The Hong Kong University of Science and Technology
^† Corresponding Authors

Abstract

Medical World Model (MeWM) is a generative model that simulates future disease states conditioned on clinical decisions. It integrates vision-language policy models to generate treatment plans and tumor dynamics models to predict visual disease progression or regression. An inverse dynamics module enables survival-guided evaluation of treatment efficacy and selection of optimal interventions. MeWM synthesizes post-treatment tumors with high realism validated by radiologists and outperforms medical-specialized GPTs in personalized treatment planning. It enhances interventional decision-making, improving the F1-score for optimal TACE protocol selection by 13%, demonstrating its potential as a second reader in clinical workflows.

Tumor Generation

Survival Analysis

Intervention Planning

BibTeX Citation

    @article{yang2025mewm,
      title={Medical World Model: Generative Simulation of Tumor Evolution for Treatment Planning },
      author={Yijun Yang and Zhao-Yang Wang and Qiuping Liu and Shuwen Sun and Kang Wang and 
          Rama Chellappa and Zongwei Zhou and Alan Yuille and Lei Zhu and Yu-Dong Zhang and Jieneng Chen},
      journal={arXiv preprint arXiv:2506.02327},
      year={2025}
    }