Application of liver disease detection using iridology with back-propagation neural network

Iridology is the study of iris structure as a reflection of the organ condition and system in the human’s body. In this study, the organ which is detected is liver. To determine the condition of the liver through iris, texture analysis and classification process are needed to distinguish iris of eye that contains the condition of normal and abnormal liver. The purpose of this study is to detect the condition of the liver through iris using back-propagation neural network with the Gray Level Co-occurrence Matrix (GLCM) for feature extraction. Application to detect liver conditions was made using Matlab version 8.1.0.604 (R2013a). Inputs for this study which is used is the eye image with both normal and abnormal conditions of the liver, based on Bernard Jensen’s iridology chart. The image is then carried out with iris localization process, ROI-making organ of the liver, and GLCM feature extraction. Results of feature extraction is used as input data (training data and test data) for the back-propagation neural network method, then used to diagnose liver organ conditions. On the obtained test results, a number of hidden layer units showed a growing number of units in the hidden layer that makes Mean Square Error (MSE) value will decrease. It makes network performance is getting better. Based on the test results, 35 test data with four variations of the number of units in the hidden layer, namely, the variation of the number of hidden layer units [ 40 ( layer 1 ), 20 ( layer 2 ) ], [ 50 ( layer 1 ), 20 ( layer 2 ) ], [ 70 ( layer 1 ), 30 ( layer 2 ) ], and [ 80 ( layer 1 ), 30 ( layer 2 ) ]. Sequentially, the data show the success rate percentage of 77.14 %, 80 %, 88.57 %, and 91.42 % . Thus in this test, the best success percentage is 91.42 % .