Fundus Image-based Glaucoma Screening via Retinal Knowledge-Oriented Dynamic Multi-Level Feature Integration

TL;DR

This paper presents a retinal knowledge-oriented framework for glaucoma screening from fundus images, integrating multi-scale features, domain-specific priors, and adaptive lesion localization to improve robustness and accuracy.

Contribution

It introduces a novel tri-branch deep learning model with dynamic region identification and knowledge-guided attention, enhancing glaucoma diagnosis across diverse datasets.

Findings

Achieved 98.5% AUC on AIROGS dataset.

Outperformed baseline models in cross-domain tests.

Demonstrated robustness and generalization in multiple datasets.

Abstract

Automated diagnosis based on color fundus photography is essential for large-scale glaucoma screening. However, existing deep learning models are typically data-driven and lack explicit integration of retinal anatomical knowledge, which limits their robustness across heterogeneous clinical datasets. Moreover, pathological cues in fundus images may appear beyond predefined anatomical regions, making fixed-region feature extraction insufficient for reliable diagnosis. To address these challenges, we propose a retinal knowledge-oriented glaucoma screening framework that integrates dynamic multi-scale feature learning with domain-specific retinal priors. The framework adopts a tri-branch structure to capture complementary retinal representations, including global retinal context, structural features of the optic disc/cup, and dynamically localized pathological regions. A Dynamic Window Mechanism is devised to adaptively identify diagnostically informative regions, while a Knowledge-Enhanced Convolutional Attention Module incorporates retinal priors extracted from a pre-trained foundation model to guide attention learning. Extensive experiments on the large-scale AIROGS dataset demonstrate that the proposed method outperforms diverse baselines, achieving an AUC of 98.5% and an accuracy of 94.6%. Additional evaluations on multiple datasets from the SMDG-19 benchmark further confirm its strong cross-domain generalization capability, indicating that knowledge-guided attention combined with adaptive lesion localization can significantly improve the robustness of automated glaucoma screening systems.