Master Thesis Automated Glaucoma Detection based on Combination of Feature Extraction Methods

Abstract

Machine learning techniques have been using increasingly in medical image analysis field to perform features recognition and decision-making tasks that demonstrate major advances in medical care field. Automated analysis of medical images contributes to increase the classification performance. Increasing the number of glaucoma patients in our country motivated us to establish an automated system for detecting the disease. Glaucoma is chronic and degenerative disease causing irreversible damage in nerve system of an eye and is led to blindness. This research aimed to present an automated glaucoma detection approach by identifying non-morphological attributes using combination of feature extraction methods. The proposed methodology was divided into: image acquisition through RIM-ONEs datasets, 2D-Discrete Wavelet Transform was applied to de-noising databases images. Local Binary Pattern represented the separated images and Gray Level Run-Length Matrix used to describe the texture patterns. Finally, two classifiers were applied to learn the models using extracted features, the proposed methodology was evaluate by measuring the sensitivity, specificity and accuracy of the models and the results were promising and more accurate compared to the results of related literatures.