Directed Ordinal Diffusion Regularization for Progression-Aware Diabetic Retinopathy Grading

2026-02-25

AI Summary

This work addresses a critical limitation in existing diabetic retinopathy (DR) grading methods, which model disease stages as static, symmetric ordinal categories, thereby ignoring the irreversible and unidirectional nature of DR progression and often yielding clinically implausible feature representations. To overcome this, the authors propose a novel directed diffusion mechanism that constructs a progression-constrained directed graph in feature space, explicitly modeling the unidirectional worsening of DR. A multi-scale diffusion regularization is introduced to penalize score inversions along invalid transition paths, thereby enforcing ordinal representations that align with clinical disease trajectories and preventing unrealistic proximity or reverse transitions between non-consecutive stages. Experiments demonstrate that the proposed method outperforms current state-of-the-art approaches in both DR grading and ordinal regression across multiple metrics, significantly improving diagnostic accuracy and clinical plausibility.

Technology Category

• Computer Vision: Diffusion Models for Vision
• Knowledge Representation and Reasoning: Diagnosis and Abductive Reasoning
• Machine Learning: Graph-based Machine Learning

Application Category

• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Graph Algorithms and Modeling for the Web: Graph embeddings and representation learning for Web-related graphs
• User Modeling, Personalization and Recommendation: Fairness-aware retrieval and ranking

Abstract

Diabetic Retinopathy (DR) progresses as a continuous and irreversible deterioration of the retina, following a well-defined clinical trajectory from mild to severe stages. However, most existing ordinal regression approaches model DR severity as a set of static, symmetric ranks, capturing relative order while ignoring the inherent unidirectional nature of disease progression. As a result, the learned feature representations may violate biological plausibility, allowing implausible proximity between non-consecutive stages or even reverse transitions. To bridge this gap, we propose Directed Ordinal Diffusion Regularization (D-ODR), which explicitly models the feature space as a directed flow by constructing a progression-constrained directed graph that strictly enforces forward disease evolution. By performing multi-scale diffusion on this directed structure, D-ODR imposes penalties on score inversions along valid progression paths, thereby effectively preventing the model from learning biologically inconsistent reverse transitions. This mechanism aligns the feature representation with the natural trajectory of DR worsening. Extensive experiments demonstrate that D-ODR yields superior grading performance compared to state-of-the-art ordinal regression and DR-specific grading methods, offering a more clinically reliable assessment of disease severity. Our code is available on https://github.com/HovChen/D-ODR.

Problem

• Diabetic Retinopathy
• ordinal regression
• disease progression
• feature representation
• biological plausibility

Innovation

• Directed Ordinal Diffusion Regularization
• Diabetic Retinopathy Grading
• Ordinal Regression
• Disease Progression Modeling
• Directed Graph Diffusion