Evo-RAD: Navigating Rare Retinal Disease Diagnosis via Self-Evolving Agentic Retrieval

2026-06-22

AI Summary

This study addresses the limited performance of existing large models in diagnosing rare retinal diseases, primarily due to scarce training data and the susceptibility of conventional retrieval-augmented methods to bias toward common conditions. To overcome these challenges, this work proposes the first self-evolving intelligent retrieval framework tailored for rare disease diagnosis, formulating evidence retrieval as a Markov decision process. A graph-structured agent dynamically performs deletion, insertion, or termination actions to optimize the pathological consistency of the support set. Integrating graph neural networks with Group Relative Policy Optimization (GRPO), the approach introduces a homophily-aware reward mechanism that transcends the limitations of static retrieval. Evaluated on a retinal disease benchmark, the method improves diagnostic accuracy by 21.04% over the base model and outperforms current retrieval-augmented and parameter-efficient fine-tuning approaches by 3.56%.

Abstract

Large-scale pretrained foundation models have revolutionized general medical screening, but often falter on rare diseases because such conditions are underrepresented in real-world clinical datasets. While retrieval-augmented diagnosis attempts to mitigate this, conventional static methods frequently succumb to the hubness problem, retrieving visually similar but semantically incorrect common diseases. To address this, we propose Evo-RAD, a self-evolving agentic framework that transforms evidence acquisition into a dynamic decision-making task. We formulate retrieval as a Markov Decision Process (MDP) where a graphbased agent observes the reference set state and executes actions to purge discordant evidence (DELETE), acquire pathologically consistent samples (INSERT), or conclude the evolution (TERMINATE). Optimized via Group Relative Policy Optimization (GRPO) with a homogeneityaware reward, the agent learns to maximize the diagnostic homogeneity of the support reference set. Experiments on retinal disease benchmarks show that Evo-RAD substantially improves rare-disease diagnosis, outperforming retinal foundation models by +21.04%, while also surpassing retrieval-based and parameter-efficient fine-tuning methods by +3.56%. Code is available at https://github.com/SDH-Lab/Evo-RAD.

Problem

• rare retinal disease
• retrieval-augmented diagnosis
• hubness problem
• diagnostic accuracy
• foundation models

Innovation

• self-evolving agentic retrieval
• Markov Decision Process
• retrieval-augmented diagnosis
• rare retinal disease
• Group Relative Policy Optimization