Enhancing Retinal Vessel Segmentation Generalization via Layout-Aware Generative Modelling

2025-03-03

AI Summary

To address weak model generalization in retinal vessel segmentation—caused by scarce annotations and imaging variability—this paper proposes RLAD, a layout-aware diffusion generative framework. RLAD enables the first decoupled modeling and conditional controllable synthesis of key anatomical structures (vessels, lesions, optic disc), enforcing strong structural constraints on vessels while allowing flexible variation in other components. We introduce REYIA, the first open-source, high-quality benchmark dataset comprising 586 expert-annotated retinal images. Furthermore, we design a real-image-driven, structure-guided sampling strategy to generate high-fidelity, diverse paired image–mask data. In cross-domain vessel segmentation, RLAD-based data augmentation improves Dice score by up to 8.1%. Both the codebase and the REYIA dataset are fully open-sourced, facilitating reproducible research in medical image generation and segmentation.

Technology Category

• Computer Vision: Diffusion Models for Vision
• Machine Learning: Deep Generative Models & Autoencoders
• Natural Language Processing: Generation

Application Category

• Search and Retrieval-Augmented AI: Retrieval-Augmented Generation (RAG) and multi-modal RAG
• User Modeling, Personalization and Recommendation: Fairness-aware retrieval and ranking
• Responsible Web: Data and user privacy-enhancing technologies for the Web

Abstract

Generalization in medical segmentation models is challenging due to limited annotated datasets and imaging variability. To address this, we propose Retinal Layout-Aware Diffusion (RLAD), a novel diffusion-based framework for generating controllable layout-aware images. RLAD conditions image generation on multiple key layout components extracted from real images, ensuring high structural fidelity while enabling diversity in other components. Applied to retinal fundus imaging, we augmented the training datasets by synthesizing paired retinal images and vessel segmentations conditioned on extracted blood vessels from real images, while varying other layout components such as lesions and the optic disc. Experiments demonstrated that RLAD-generated data improved generalization in retinal vessel segmentation by up to 8.1%. Furthermore, we present REYIA, a comprehensive dataset comprising 586 manually segmented retinal images. To foster reproducibility and drive innovation, both our code and dataset will be made publicly accessible.

Problem

• Improves retinal vessel segmentation generalization using generative modeling.
• Addresses limited annotated datasets and imaging variability in medical segmentation.
• Enhances model performance by generating diverse, layout-aware retinal images.

Innovation

• RLAD framework generates layout-aware retinal images.
• Augments datasets with synthesized paired images.
• Improves segmentation generalization by 8.1%.