DETECTION ABNORMALITIES OF THE X-RAY IMAGE OF THE LUNG BASED ON THE GRAY LEVEL COOCCURANCE MATRIX USING NEURAL NETWORK

Abstract

In this study a software system has been built that can detect abnormalities of x-ray images using backpropagation neural network. This software is expected to be a tool for lung doctors to diagnose lung cancer quickly and accurately. the results of this software are normal x-ray images as well as those diagnosed with lung cancer. this process begins with the preprocessing stage, the second stage is feature extraction using Gray level co-occurance matrix (GLCM) to look for feature texture characteristics of the pulmonary x-ray image. the third stage is classification using backpropagation neural network. Classification using backpropagation consists of the training and testing stages. The parameters used for the training and testing stages are learning rate = 0.1 and target error = 0.001. at the training stage, there are various hidden layer and epoch variations to find the best network architecture. The training results show hidden layer = 15 and the number of epoch = 200 is the best neural network architecture with an MSE value of 0.165. the testing phase using the same parameters shows the accuracy of 87.5%