Search papers, labs, and topics across Lattice.
This study introduces LEADS, a novel framework that leverages large language models (LLMs) to autonomously discover hybrid cardiac electrophysiology (EP) models tailored to individual patients. By structuring cardiac domain knowledge into an action space, LEADS enables an iterative reasoning-and-action process that combines and refines model architectures, ensuring they remain physically grounded and numerically stable. Validation on both synthetic and real cardiac EP data reveals that LEADS significantly outperforms traditional human-designed models and existing LLM-based approaches in terms of accuracy and stability.
LEADS enables the automated discovery of personalized cardiac models, outperforming expert-designed approaches and setting a new standard for stability in cardiac simulations.
Building personalized cardiac electrophysiology (EP) digital twins requires identifying the appropriate model structure for each patient, not merely fitting parameters. Traditional methods rely on experts to manually prescribe hybrid physics-neural architectures, which requires deep domain expertise and does not transfer across patients. Recent works have applied large language models (LLMs) to generate or act as hybrid models. However, despite their promising generalization capacity, these LLM-based methods lack the structural priors needed for stable cardiac simulations. Hence, we propose LEADS, a framework that formulates cardiac EP domain knowledge as a structured action space and utilizes an LLM agent to discover hybrid models. The agent follows an iterative reasoning-and-action loop to select, combine, and refine hybrid models, whilst gradient descent handles parameter fitting. The proposed LEADS designs every candidate model towards physically grounded, interpretable, and numerically stable, while allowing open-ended architectural discovery. We validate LEADS on synthetic data with three ground-truth reaction models and on real cardiac EP data, demonstrating that it outperforms both human-designed hybrid models and other LLM-based hybrid modeling.
