Joint Optic Disc and Cup Segmentation for the Identification of Glaucoma Using Fully Convolutional Neural Network

  • E. Sudheer Kumar, C. Shoba Bindu

Abstract

Glaucoma is an eye condition that results in vision loss and, in severe cases, blindness.Fundus photography's Optic Disc (OD) segmentation is the first and most important step in detecting retinal diseases. For ophthalmology diagnoses like eye glaucoma and other disorders, accurate segmentation of the Optic Cup (OC) from fundus pictures is particularly critical. The cup to disc ratio (CDR) is a crucial glaucoma screening and diagnostic indication. CDR may be obtained with accurate segmentation of the optic disc and cup. Although numerous deep learning-based techniques for segmenting the disc and cup for fundus images have been suggested, owing of the substantial overlap between the optic disc and cup, obtaining highly accurate segmentation performance remains a significant problem. This paper aims to perform joint segmentation with the help of modified U-Net architecture in two steps: as a first step the optic disc boundary will be segmented from the fundus image and in the next step the fundus image will be cropped to the size of disc boundary to identify the cup region. After segmenting both the regions of disc and cup the CDR will be estimated to identify whether the eye is affected with glaucoma or not. The method proposed in this paper was validated on RIM-ONE r3 dataset. The suggested approach can better predict the CDR for a large-scale glaucoma screening, according to experimental findings.

Published
2021-11-22
How to Cite
E. Sudheer Kumar, C. Shoba Bindu. (2021). Joint Optic Disc and Cup Segmentation for the Identification of Glaucoma Using Fully Convolutional Neural Network. Design Engineering, 16034 - 16046. Retrieved from http://www.thedesignengineering.com/index.php/DE/article/view/6567
Section
Articles