Integrating Grey Wolf Optimizer with U-Net for Precision Retinal Vessel Segmentation in Fundus Images

Retinal Vessel Segmentation (RVS) is pivotal in diagnosing and monitoring ocular diseases. While Deep Learning (DL) methodologies, particularly Convolutional Neural Networks (CNN), have revolutionized medical image analysis, achieving consistent and superior segmentation accuracy remains challenging. The conventional U-Net architecture, despite its merits, presents limitations in terms of adaptability and precision. This research article introduces a novel, “Hybrid U-Net Deep Neural Network” tailored for RVS in Fundus Images (FI). This approach integrates the Grey Wolf Optimizer (GWO) to fine-tune U-Net’s hyperparameters and elevate its segmentation accuracy. The GWO’s intrinsic ability to search for optimal solutions optimizes the U-Net’s performance. Comparative analyses on benchmark datasets, HRF and STARE, underscore this model’s superior segmentation results. Through rigorous quantitative and qualitative evaluations, the hybrid model significantly enhances RVS, positioning it as a promising tool for clinical and research applications in ophthalmology. This research demonstrates the potential of integrating optimization algorithms with Deep Learning (DL) and paves the system for future advancements in Medical Image Segmentation (MIS).