Title: Feature selection and classification for automatic detection of retinal nerve fibre layer thinning in retinal fundus images

Authors: Medha V. Wyawahare; Pradeep M. Patil

Addresses: Vishwakarma Institute of Technology, 666, Upper Indira Nagar, Bibwewadi, Pune 411037, Maharashtra, India ' KJ's Educational Institutes, Kondhwa Saswad Road, Pune 411048, Maharashtra, India

Abstract: Glaucoma which is a leading cause of blindness in the world is not a single disease but a group of disorders with diverse clinical manifestations. If not treated at an early stage it leads to loss of vision. Careful evaluation of optic nerve head structure and Retinal Nerve Fibre Layer (RNFL) is extremely important for diagnosis of the disease and subsequent medication. This work focuses on automatic detection of RNFL thinning, an early indication of glaucoma. A texture-based method is proposed in this work. Texture features were extracted from gray-level co-occurrence matrix and the feature set was optimised using Analysis of Variance (ANOVA) method. Radial basis function classifier was used to discriminate between normal and abnormal regions. The classification results were promising with accuracy of 96.41%.

Keywords: optic nerve head; optic disc; retinal nerve fibre layer; glaucoma; ANOVA; feature selection; feature classification; automatic detection; nerve fibre layer thinning; retinal fundus images; blindness; texture features; radial basis function; RBF; retina.

DOI: 10.1504/IJBET.2015.072991

International Journal of Biomedical Engineering and Technology, 2015 Vol.19 No.3, pp.205 - 219

Received: 31 Jan 2015
Accepted: 06 May 2015

Published online: 11 Nov 2015 *

Full-text access for editors Full-text access for subscribers Purchase this article Comment on this article