Computer-Aided Design System for Early Detection and Screening of Glaucoma Using Pretrained Convolutional Neural Networks

Abstract

Glaucoma, the second-leading cause of blindness worldwide, is a neuro-degenerative eye condition primarily caused by an increase in intraocular pressure within the retina. This paper presents a novel approach to developing a CAD system aimed at assisting in the early detection, screening, and treatment of glaucoma. The proposed system utilizes three pre-trained CNNs for classifying glaucoma. To train and evaluate the CNN models, several datasets were employed, including the DRIONS-DB, HRF dataset, and DRISHTI-GS dataset. Our system achieved impressive accuracy rates by leveraging the DRIONS-DB, HRF, DRISHTI-GS, and combined datasets. Specifically, accuracy rates of 95.63%, 98.67%, 95.64%, and 88.96% were obtained for each dataset. The results demonstrate the effectiveness of utilizing pre-trained CNNs in the proposed CAD system for glaucoma detection. The high accuracy rates obtained indicate the potential of this system in aiding medical professionals with early diagnosis and screening of glaucoma, thereby facilitating timely intervention and treatment.