SciGroveBeta
Medicine

Learning Cardiac Electrophysiology Digital Twins Through Agentic Discovery of Hybrid Structure

Ziqi Zhou, Yubo Ye, Sumeet Atul Vadhavka, Linwei Wang, Zhiqiang Tao

Featured June 17, 2026

Original Paper

This analysis was generated by SciGrove. Upload your own PDFs or enter a DOI — and get the same AI breakdown on any paper.

Get started

AI-generated analysis — This is SciGrove's AI interpretation of the paper, not peer-reviewed content. Always refer to the original paper.

Simply

A smart computer agent uses its reasoning to pick and combine building blocks from a special catalog of heart models, automatically creating a personalized digital twin that simulates a patient's heart better than old methods.

In depth
The paper introduces LEADS, a framework that uses a Large Language Model (LLM) agent to automatically discover personalized hybrid cardiac electrophysiology (EP) digital twin models. Instead of manually designing complex model structures, the agent iteratively searches a structured action space of physics-based and neural components, guided by domain knowledge, to find the optimal architecture for each patient. This approach ensures physical plausibility while enabling open-ended architectural discovery.

Key Takeaways

  • 1
    LEADS automates the discovery of personalized cardiac EP digital twin model structures using an LLM agent, moving beyond manual expert design.
  • 2
    The framework leverages a structured action space that encodes cardiac EP domain knowledge, combining physics-based reaction models and neural diffusion architectures.
  • 3
    The LLM agent employs an Observe-Think-Act loop to iteratively refine hybrid models, outperforming both human-designed and unconstrained LLM-based methods.

Conceptual Flow

HIGH LEVEL
1
Methodology (The Logic)

A smart computer agent looks at different ways to build a heart model, tries them out, and learns to pick the best parts to make a super accurate digital twin.

Heart Data
Model Building Blocks
Agent Tries Combinations
Best Heart Model
2
Results (The Impact)

The new method creates heart models that are much better at predicting how a patient's heart will behave compared to models designed by people or other computer programs.

Old Heart Models
New Heart Models
Compare Accuracy
New Models Are Better