GC-NET for classification of glaucoma in the retinal fundus image

• DOI: 10.1007/s00138-020-01091-4
• Publication Date: 2020-06-12T10:03:36Z
• Authors: Mamta Juneja; Niharika Thakur; Sarthak Thakur; Archit Uniyal; Anuj Wani; Prashant Jindal
• Keywords: 0202 electrical engineering, electronic engineering, information engineering; 02 engineering and technology

Glaucoma is the second-most dominant cause for irreversible blindness, resulting in damage to the optic nerve. Ophthalmologist diagnoses this disease using a retinal examination of the dilated pupil. Since the diagnosis is a manual and laborious procedure, an automated approach for faster diagnosis is desirable. Convolutional neural networks (CNN) could allow automation of the diagnosis procedure due to their self-learning capabilities. This paper presents a deep learning-based glaucoma classification network (GC-NET) for classifying a retinal image as glaucomatous or non-glaucomatous. The proposed GC-NET has been tested on RIM-One and Drishti datasets. Our experimental results showed that GC-NET achieves accuracy of 97.51%, sensitivity of 98.78% and specificity of 96.20% with 0.81 true positive (Tp), 0.03 false positive (Fp), 0.76 true negative (Tn) and 0.01 false negative (Fn) which outperforms state of the art. Thus, the proposed approach could be very useful for initial screening of glaucoma patients.