Retinal Vessel Segmentation Using a Novel U-Net Architecture with Data Augmentation
摘要
In this paper, the retinal vessel segmentation problem is highlighted and a novel U-Net architecture with data augmentation is proposed to segment the retinal vessel. The proposed architecture is applied to a benchmark dataset like Digital Retinal Images for Vessel Extraction (DRIVE). After performance evaluation, it has been observed that the novel U-Net architecture with augmentation generates a 79.67% F1-Score, 78.48% Recall Rate, 81.38% Precision Rate, and 96.52% accuracy. The result of the proposed architecture proves to be superior with respect to U-Net architecture without data augmentation and other architectures proposes in recent times.