Pathology-Aware Prototype Evolution via LLM-Driven Semantic Disambiguation for Multicenter Diabetic Retinopathy Diagnosis

TL;DR

This paper introduces a pathology-aware framework for diabetic retinopathy grading that leverages foundation models and semantic prompts to improve prototype-based diagnosis, outperforming existing methods.

Contribution

The paper proposes a novel Hierarchical Anchor Prototype Modulation framework integrating semantic prompts and clinical knowledge for improved DR grading.

Findings

Outperforms state-of-the-art methods on eight public datasets.

Effectively incorporates pathological descriptions into prototype evolution.

Demonstrates robustness across diverse datasets.

Abstract

Diabetic retinopathy (DR) grading plays a critical role in early clinical intervention and vision preservation. Recent explorations predominantly focus on visual lesion feature extraction through data processing and domain decoupling strategies. However, they generally overlook domain-invariant pathological patterns and underutilize the rich contextual knowledge of foundation models, relying solely on visual information, which is insufficient for distinguishing subtle pathological variations. Therefore, we propose integrating fine-grained pathological descriptions to complement prototypes with additional context, thereby resolving ambiguities in borderline cases. Specifically, we propose a Hierarchical Anchor Prototype Modulation (HAPM) framework to facilitate DR grading. First, we introduce a variance spectrum-driven anchor prototype library that preserves domain-invariant pathological patterns. We further employ a hierarchical differential prompt gating mechanism, dynamically selecting discriminative semantic prompts from both LVLM and LLM sources to address semantic confusion between adjacent DR grades. Finally, we utilize a two-stage prototype modulation strategy that progressively integrates clinical knowledge into visual prototypes through a Pathological Semantic Injector (PSI) and a Discriminative Prototype Enhancer (DPE). Extensive experiments across eight public datasets demonstrate that our approach achieves pathology-guided prototype evolution while outperforming state-of-the-art methods. The code is available at https://github.com/zhcz328/HAPM.