PARAGraph: Pathology-Anatomy-Aware Hierarchical Graph for Diabetic Retinopathy Grading

2026-08-08

AI Summary

This work addresses the limitations of existing diabetic retinopathy (DR) grading methods, which predominantly rely on image-level classification and fail to explicitly model clinically critical evidence such as lesion types and their spatial relationships. To overcome this, the authors propose a three-tier hierarchical graph architecture that integrates pathological-anatomical priors, incorporating a normalized coordinate system centered on the optic disc and macula. A dual-fusion strategy is further introduced to enhance robustness against segmentation noise in lesion detection. Notably, this approach is the first to jointly model anatomical semantics and lesion characteristics within a hierarchical graph framework, enabling clinically interpretable DR grading. Extensive experiments demonstrate superior performance over state-of-the-art methods on the Messidor-2, APTOS, and DDR datasets, achieving both high accuracy and strong robustness.

Abstract

Diabetic retinopathy (DR) remains a leading cause of vision loss among working-age adults worldwide, making reliable severity grading clinically important. Despite strong performance, most deep models formulate DR grading as image-level classification and do not explicitly model clinically grounded evidence, such as lesion types and spatial relations. In this paper, we propose PARAGraph, a Pathology-Anatomy-Aware Hierarchical Graph framework for DR grading. PARAGraph represents each image as a three-level hierarchical graph with lesion-level nodes, intermediate category and region nodes, and global anatomical and semantic nodes. To incorporate medical priors into nodes, we construct an optic disc-fovea-anchored coordinate frame that provides a scale- and rotation-normalized retinal reference system. Within this frame, lesion nodes are encoded with category, normalized area, and anatomical coordinates. To mitigate noisy lesion segmentation, PARAGraph uses a dual-fusion strategy that introduces global visual context into a graph semantic node and a decision-level prediction branch, improving robustness when lesion evidence is unreliable. Extensive experiments on Messidor-2, APTOS, and DDR show that PARAGraph achieves consistent DR grading performance over state-of-the-art methods. Interpretability and robustness analyses further demonstrate that its predictions are clinically grounded, closely associated with lesion evidence and robust to lesion segmentation noise.

Problem

• Diabetic Retinopathy
• Severity Grading
• Lesion Modeling
• Spatial Relations
• Clinical Interpretability

Innovation

• hierarchical graph
• pathology-anatomy-aware
• optic disc-fovea coordinate frame
• dual-fusion strategy
• lesion-aware DR grading