LFRA-Net: A Lightweight Focal and Region-Aware Attention Network for Retinal Vessel Segmentatio

2025-09-15

AI Summary

Retinal vessel segmentation is critical for early diagnosis of vision- and systemic diseases, yet existing deep learning methods face bottlenecks in fine-vessel detection and computational efficiency, limiting deployment in resource-constrained clinical settings. To address this, we propose LFRA-Net—a lightweight encoder-decoder network that integrates focal modulation attention at the bottleneck to enhance global context modeling, and embeds region-aware attention into selective skip connections to preserve local structural details. The model achieves high parameter efficiency with only 0.17M parameters, 0.66 MB memory footprint, and 10.50 GFLOPs. Evaluated on DRIVE, STARE, and CHASE_DB datasets, LFRA-Net attains Dice scores of 84.28%, 88.44%, and 85.50%, respectively—demonstrating a superior balance between segmentation accuracy and real-world deployability.

Technology Category

• Computer Vision: Segmentation
• Machine Learning: Hardware-aware ML
• Search and Optimization: Learning to Search

Application Category

• Search and Retrieval-Augmented AI: Efficiency and scalability of Web search engines
• Responsible Web: Machine-in-the-loop, human agency and autonomy
• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs

Abstract

Retinal vessel segmentation is critical for the early diagnosis of vision-threatening and systemic diseases, especially in real-world clinical settings with limited computational resources. Although significant improvements have been made in deep learning-based segmentation methods, current models still face challenges in extracting tiny vessels and suffer from high computational costs. In this study, we present LFRA-Net by incorporating focal modulation attention at the encoder-decoder bottleneck and region-aware attention in the selective skip connections. LFRA-Net is a lightweight network optimized for precise and effective retinal vascular segmentation. It enhances feature representation and regional focus by efficiently capturing local and global dependencies. LFRA-Net outperformed many state-of-the-art models while maintaining lightweight characteristics with only 0.17 million parameters, 0.66 MB memory size, and 10.50 GFLOPs. We validated it on three publicly available datasets: DRIVE, STARE, and CHASE_DB. It performed better in terms of Dice score (84.28%, 88.44%, and 85.50%) and Jaccard index (72.86%, 79.31%, and 74.70%) on the DRIVE, STARE, and CHASE_DB datasets, respectively. LFRA-Net provides an ideal ratio between segmentation accuracy and computational cost compared to existing deep learning methods, which makes it suitable for real-time clinical applications in areas with limited resources. The code can be found at https://github.com/Mehwish4593/LFRA-Net.

Problem

• Lightweight network for retinal vessel segmentation
• Improves tiny vessel extraction accuracy
• Reduces computational costs for clinical use

Innovation

• Lightweight network with focal modulation attention
• Region-aware attention in skip connections
• Efficient local and global dependencies capture