Gaze into the Details: Locality-Sensitive Enhancement for OCTA Retinal Vessel Segmentation

2026-05-19

AI Summary

This study addresses the challenge of vessel fragmentation and detail loss in OCTA retinal vasculature segmentation, commonly caused by low local contrast and prevalent in existing U-Net–based methods. To overcome these limitations, the authors propose three novel modules integrated into a U-Net framework: a Patch-level Interactive Enhancement (PIE) module that replaces conventional skip connections with a block-wise attention mechanism to mitigate vessel discontinuities; a Multi-scale Feature Fusion (MFF) module that integrates interpretable multi-scale features to preserve fine structural details; and a Contextual Receptive-field Design (CRD) module employing large convolutional kernels to refine feature representations across all network levels and reduce segmentation fragmentation. This work is the first to systematically incorporate a locality-sensitive enhancement mechanism into OCTA vessel segmentation, achieving state-of-the-art performance on three public benchmarks—OCTA-500, ROSE-1, and ROSSA—with a lower parameter count.

Abstract

Existing deep learning frameworks for Optical Coherence Tomography Angiography (OCTA) vessel segmentation are largely derived from the U-Net architecture, which serves as the foundation for most current designs. However, most of these methods focus only on holistic representation, struggling to address the problem of low local contrast unique to OCTA, which leads to vessel discontinuities and loss of detail. To address these problems, we propose LSENet, which builds upon the U-Net architecture by introducing three core innovative modules: To address vessel discontinuities, we introduce the Patch Information Enhance module (PIE), which replaces standard skip connections to execute patch-wise attention. To mitigate detail loss, the Multiscale Feature Fusion module (MFF) is proposed to feed the PIE module rich, multi-scale information by extracting visually interpretable features from both the original input and preceding layers. Finally, the Connectivity Refinement Decoder (CRD) is designed to refine features from all levels and utilize a large kernel in the final convolutional layer to reduce fragmentation. Experiments on three public datasets (OCTA-500, ROSE-1, and ROSSA) demonstrate that our proposed LSENet achieves state-of-the-art performance while requiring fewer parameters.

Problem

• OCTA
• retinal vessel segmentation
• low local contrast
• vessel discontinuities
• detail loss

Innovation

• LSENet
• Patch-wise Attention
• Multiscale Feature Fusion
• Connectivity Refinement
• OCTA Vessel Segmentation