AUTOMATED OPTIC DISC SEGMENTATION FOR FUNDUS IMAGES BASED ON ARTIFICIAL NEURAL NETWORKS: U-NET
No Thumbnail Available
Date
2024-08-27
Authors
Alhendi, Nour
Journal Title
Journal ISSN
Volume Title
Publisher
An-Najah National University
Abstract
Optic disc (OD), located at the back of the eye, is a significant part of the retina. It represents the entry point for the optic nerve and blood vessels. Accurate OD segmentation provides critical information about the anatomy and health state of the retina, aiding in diagnosing and managing various eye conditions such as glaucoma, diabetic retinopathy (DR), and optic nerve abnormalities. With automatic OD segmentation, computer-based systems can efficiently analyze large numbers of retinal images, enabling early detection and monitoring of eye diseases. This automation not only enhances the speed and accuracy of diagnosis but also facilitates cost-effective and accessible healthcare, especially in areas with limited ophthalmic expertise.
In this study, an automatic method for OD segmentation in retinal images using a convolutional neural network (CNN) architecture, known as U-Net, was introduced. First, a region of interest (ROI) was extracted from the fundus images using the bounding box technique. For faster calculations, the cropped images were resized to 128 × 128 pixels. Then, these images were enhanced using the contrast limited adaptive histogram equalization (CLAHE) to eliminate the noise and improve their qualities. After that, a U- Net model was constructed and trained to obtain segmented images.
The proposed model was trained and evaluated using the public dataset ORIGA, and the predicted results were compared with the ground truth (GT) images. This method competed with other studies and achieved average accuracy of 98.42%, average precision of 97.46%, and average sensitivity of 95.33%. As the execution time is short, this enables the proposed method to be an online implemented method.