<p>Retinal vascular image segmentation is essential for diagnosing retinal diseases, yet existing algorithms often encounter challenges. A primary issue is the misclassification of background pixels as vessels, leading to segmentation inaccuracies. Additionally, segmenting fine blood vessels is particularly challenging due to their intricate structures and low-contrast features. To address these issues, this study proposes the Edge-Enhanced Attention Guided Network (EEA-Net) for automated retinal vessel segmentation. EEA-Net employs a dual-branch receptive field encoder to capture both local and global features, enhancing segmentation accuracy and reducing misclassification. An edge enhancement module, leveraging the Laplacian operator, is incorporated to improve edge feature extraction, facilitating the identification of fine vessels with ambiguous boundaries. Furthermore, a directional cyclic convolution module is introduced to mitigate noise while preserving the integrity of small vessels, ensuring detailed segmentation. Experiments on three public datasets demonstrate that EEA-Net outperforms existing methods, achieving superior segmentation performance, particularly in fine-vessel delineation,underscoring its effectiveness and practical applicability.</p>