Multi-Agent Reasoning for Cardiovascular Imaging Phenotype Analysis

2025-07-04

AI Summary

Traditional cardiovascular imaging phenotyping relies on manual hypothesis generation and linear modeling, limiting its ability to capture complex nonlinear associations between imaging phenotypes and multifactorial risk determinants. To address this, we propose MESHAgents—a novel large language model–based multi-agent self-organizing reasoning framework—the first of its kind designed for cardiology. It emulates collaborative reasoning among domain experts in cardiology, biomechanics, statistics, and clinical research to enable assumption-free, holistic phenotype–factor association discovery. The framework dynamically integrates statistical correlation analysis with multi-expert consensus, ensuring interpretability and scalability. In population-scale imaging studies, MESHAgents autonomously identified novel confounding factors beyond standard demographic variables. Phenotypes derived by MESHAgents achieved near-equivalent disease classification performance to expert-curated features: AUC decreased by only 0.004, while recall improved significantly for 6 out of 9 disease classes—demonstrating clinical-grade efficacy without human curation.

Technology Category

• Multiagent Systems: Modeling other Agents
• Cognitive Modeling & Cognitive Systems: Agent Architectures
• Machine Learning: Mixture of Experts (MoE)

Application Category

• Semantics and Knowledge: Data modeling to support human-machine intelligence, including LLMs agents, intelligent system behavior, explanations, and user-friendly interactions
• Responsible Web: Machine-in-the-loop, human agency and autonomy
• Search and Retrieval-Augmented AI: Agentic search

Abstract

Identifying the associations between imaging phenotypes and disease risk factors and outcomes is essential for understanding disease mechanisms and improving diagnosis and prognosis models. However, traditional approaches rely on human-driven hypothesis testing and selection of association factors, often overlooking complex, non-linear dependencies among imaging phenotypes and other multi-modal data. To address this, we introduce a Multi-agent Exploratory Synergy for the Heart (MESHAgents) framework that leverages large language models as agents to dynamically elicit, surface, and decide confounders and phenotypes in association studies, using cardiovascular imaging as a proof of concept. Specifically, we orchestrate a multi-disciplinary team of AI agents -- spanning cardiology, biomechanics, statistics, and clinical research -- which spontaneously generate and converge on insights through iterative, self-organizing reasoning. The framework dynamically synthesizes statistical correlations with multi-expert consensus, providing an automated pipeline for phenome-wide association studies (PheWAS). We demonstrate the system's capabilities through a population-based study of imaging phenotypes of the heart and aorta. MESHAgents autonomously uncovered correlations between imaging phenotypes and a wide range of non-imaging factors, identifying additional confounder variables beyond standard demographic factors. Validation on diagnosis tasks reveals that MESHAgents-discovered phenotypes achieve performance comparable to expert-selected phenotypes, with mean AUC differences as small as -0.004 on disease classification tasks. Notably, the recall score improves for 6 out of 9 disease types. Our framework provides clinically relevant imaging phenotypes with transparent reasoning, offering a scalable alternative to expert-driven methods.

Problem

• Identifying associations between imaging phenotypes and disease risks
• Overcoming human-driven limitations in complex data analysis
• Automating phenome-wide association studies with multi-agent reasoning

Innovation

• Multi-agent framework using large language models
• Dynamic confounder and phenotype identification
• Automated pipeline for phenome-wide association studies