Interpretable Retinal Disease Prediction Using Biology-Informed Heterogeneous Graph Representations

2025-02-23

AI Summary

To address the dual challenges of limited predictive performance and poor clinical interpretability in diabetic retinopathy (DR) staging, this paper proposes a biologically grounded heterogeneous graph modeling framework. Leveraging optical coherence tomography angiography (OCTA) images, the method automatically constructs a heterogeneous graph integrating retinal vascular segments, non-perfusion areas, and the foveal avascular zone (FAZ), casting DR grading as a graph-level classification task. The framework synergistically combines biology-informed graph construction, graph neural network (GNN) representation learning, and explainable AI (XAI) attribution techniques—enabling, for the first time, lesion-level localization of pathological vessels/regions and feature-level attribution explanations. Evaluated on two independent OCTA datasets, the model significantly outperforms CNNs, Vision Transformers (ViTs), and conventional biomarker-based models. It delivers fine-grained, clinically intelligible decision rationales, thereby enhancing clinician trust and diagnostic support capability.

Technology Category

• Computer Vision: Interpretability, Explainability, and Transparency
• Machine Learning: Graph-based Machine Learning
• Knowledge Representation and Reasoning: Diagnosis and Abductive Reasoning

Application Category

• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs
• Search and Retrieval-Augmented AI: Web query analysis, representation and understanding
• User Modeling, Personalization and Recommendation: Fairness-aware retrieval and ranking

Abstract

Interpretability is crucial to enhance trust in machine learning models for medical diagnostics. However, most state-of-the-art image classifiers based on neural networks are not interpretable. As a result, clinicians often resort to known biomarkers for diagnosis, although biomarker-based classification typically performs worse than large neural networks. This work proposes a method that surpasses the performance of established machine learning models while simultaneously improving prediction interpretability for diabetic retinopathy staging from optical coherence tomography angiography (OCTA) images. Our method is based on a novel biology-informed heterogeneous graph representation that models retinal vessel segments, intercapillary areas, and the foveal avascular zone (FAZ) in a human-interpretable way. This graph representation allows us to frame diabetic retinopathy staging as a graph-level classification task, which we solve using an efficient graph neural network. We benchmark our method against well-established baselines, including classical biomarker-based classifiers, convolutional neural networks (CNNs), and vision transformers. Our model outperforms all baselines on two datasets. Crucially, we use our biology-informed graph to provide explanations of unprecedented detail. Our approach surpasses existing methods in precisely localizing and identifying critical vessels or intercapillary areas. In addition, we give informative and human-interpretable attributions to critical characteristics. Our work contributes to the development of clinical decision-support tools in ophthalmology.

Problem

• Interpretable diabetic retinopathy prediction
• Biology-informed graph representation
• Enhanced clinical decision-support tools

Innovation

• Biology-informed heterogeneous graph representation
• Graph-level classification using GNN
• Detailed interpretable vessel localization