Disentangling Co-Occurring Retinal Pathologies with Saliency-Guided Sparse Expert Routing

2026-08-10

AI Summary

Retinal fundus images often exhibit multiple co-occurring pathologies, yet conventional deep learning models rely on static computation and struggle to adaptively address such complex comorbidities. This work proposes a sparse conditional computation architecture that integrates a Guided Contextual Gating (GCG) spatial attention mechanism with an interpretable sparse-routing Mixture-of-Experts (MoE) module. To the best of our knowledge, this is the first application of MoE to disentangle multiple retinal diseases, enabling dynamic allocation of specialized experts according to specific pathology types. Evaluated on a benchmark comprising five lesion categories, the model achieves a macro-AUC of 0.912 ± 0.008 and a macro-F1 score of 0.653 ± 0.014. Visualization via Grad-CAM++ and t-SNE confirms that expert routing aligns with localized lesions and effectively captures the underlying comorbidity structure.

Abstract

Retinal fundus images frequently exhibit multiple co-occurring pathologies, yet standard deep learning classifiers apply static, identical computation to every image regardless of the underlying disease distribution. We propose a novel architecture that resolves this via sparse conditional computation, pairing a Guided Context Gating (GCG) spatial attention front-end with a sparsely-routed Mixture-of-Experts (MoE) block operating over feature tokens. Crucially, this routing yields an interpretable, data-driven decomposition. Expert allocation is significantly disease-dependent (p < 0.001), with the healthy Normal state and morphologically distinct pathologies (e.g., ERM, AMD) isolating to dedicated experts. On a five-class, patient-disjoint 5-fold cross-validation benchmark, our model achieves 0.912 +/- 0.008 macro AUC and 0.653 +/- 0.014 macro F1. Furthermore, Grad-CAM++ and post-MoE t-SNE visualizations confirm that expert routing aligns with localized lesions and geometrically maps co-occurring cases between their constituent clusters, positioning sparse MoE as an interpretable approach to multi-disease retinal screening.

Problem

• retinal pathologies
• co-occurring diseases
• sparse expert routing
• multi-disease classification
• fundus images

Innovation

• Sparse Mixture-of-Experts
• Saliency-Guided Routing
• Disentangled Retinal Pathologies
• Interpretable Deep Learning
• Conditional Computation