Hidden Neuron Analysis for Detection Cataract Disease Based on Gray Level Co-occurrence Matrix and Back Propagation Neural Network

Abstract

A cataract is a disease when the lens of the eye becomes cloudy. The most common cause of cataracts is the aging process. Several other conditions can cause cataracts in the lens of the eye, such as diabetes and smoking. It will be caused the decreased vision until blindness. Cataracts are the number one cause of blindness in Indonesia and the world. Blindness due to cataracts is relatively high because many sufferers do not know it. Because of that, a system for detecting cataracts is needed for taking further action quickly. The availability of medical officers and equipment is deficient in rural areas. The proposed method is expected to function as a doctor in detecting eye diseases, especially cataracts, to treat patients quickly. The combination of Gray Level Co-occurrence Matrix (GLCM) as feature extraction and Back-propagation Neural Network (BPNN) classification has been proposed. The proposed method uses four features GLCM, which are contrast, homogeneity, correlation, and energy. The angular orientation of GLCM is formed based on four angular directions, namely, 0°, 45°, 90°, and 135°, and distance between pixel uses 1, 2, 3, and 4. The highest accuracy is on 9 hidden neurons, 4 input layers, and 2 output layers with an accuracy of 0.824.