Abstract

 Retina blood vessel segmentation is a critical step in ophthalmological image analysis, providing essential diagnostic insights for diseases such as diabetic retinopathy, glaucoma, and age-related macular degeneration. Automated segmentation of retinal blood vessels enables early diagnosis and effective treatment planning. In this study, a U-Net-based Convolutional Neural Network (CNN) is implemented for precise segmentation of retinal blood vessels from fundus images. The U-Net model consists of a contracting path for feature extraction and an expansive path for accurate localization, with skip connections to preserve finegrained spatial information. This architecture is particularly effective in addressing the challenges posed by the thin and complex structures of retinal vessels.The proposed system is trained using publicly available datasets, including DRIVE, STARE, and CHASE_DB1, using data augmentation techniques to improve robustness against image variations. Additionally, the use of Dice loss and cross-entropy loss functions ensures better handling of class imbalances between vessels and background pixels. Experimental results demonstrate that the U-Net model outperforms traditional methods and other CNN architectures, achieving high accuracy, sensitivity, and specificity. The model achieves an average segmentation accuracy of over 95% across multiple datasets.Furthermore, the implementation of preprocessing techniques such as contrast enhancement using CLAHE (Contrast Limited Adaptive Histogram Equalization) and Gaussian filtering contributes to improved vessel visibility. Post-processing steps, including morphological operations and region filtering, further refine the segmented results. The system is computationally efficient, offering real-time performance with low latency, making it suitable for deployment in clinical settings.Future work will explore the integration of attention mechanisms and hybrid models to further enhance segmentation accuracy, particularly in cases of low-quality images. Additionally, expanding the model’s capabilities to support multi-modal retinal imaging and real-world application scenarios will enhance its diagnostic potential.