Computational Disease Modeling

We build imaging-based models that probe disease mechanisms at the individual patient level.

Representative directions include:

• Tumor connectomics – modeling tumors as networks to understand spatial organization, invasion patterns, and risk.
• Longitudinal evolution – using multi-timepoint imaging to predict how tumors or organs change under treatment.
• Counterfactual modeling – exploring alternative trajectories in latent space to reason about “what if” scenarios for therapy.

The goal is to move beyond static endpoint prediction toward dynamic models of disease behavior.