A Novel approach for Glaucoma Disease Identification through Optic Nerve Head Feature Extraction and Random Tree Classification

Abstract

Glaucoma is one of the second leading causes of blindness behind cataracts and this sight-stealing disease affected 4.5 million people worldwide was estimated by World Health Organization. Glaucoma is a group of related eye disorders that cause extra fluid builds up in the front part of an eye leads to ocular hypertension can damage the optic nerves. The optic Nerve Head is a bundle of one million nerve fibers that carries visual signals from the eye to the brain. A novel approach is proposed for glaucoma identification by optic nerve head feature extraction from multi color channel using image processing followed by disease classification using data mining techniques. The proposed system uses combination of optic Disc localization, optic nerve head region segmentation, color space conversion, color channel selection, extracts gray level co-occurence matrix, histogram and statistical features of 29 color channels, feature relevance analysis and disease classification process. This system was tested on three publically available databases Drishti-GS1, RIM-ONE r1 and RIM-ONE r2 and also evaluated on ground truth given by experts achieves the overall positive predictive value of 97.96% shows that proposed approach is more robust and outperforms the state-of-the-art techniques.