Classification of Pneumonia in Thoracic X-Ray images based on texture characteristics using the MLP (Multi-Layer Perceptron) method

Abstract

One of the diseases that attack the lungs is pneumonia. This disease can attack someone with a weak immune system. Pneumonia is inflammation of the lungs that can be caused by pathogens, such as bacteria, viruses, and fungi. The purpose of this study was to classify fungal pneumonia, bacterial pneumonia, and lipoid pneumonia based on texture characteristics and the MLP method using machine learning WEKA. The method in this study has three stages including pre-processing, extraction of texture features consisting of Histogram and GLCM, and classification using the MLP (Multi Layer Perceptron) method. The results of the texture feature extraction showed that the three types of pneumonia were: lipoid pneumonia with brightness, sharp contrast random distribution on correlation characteristics, bacterial pneumonia with high brightness, high contrast random distribution on energy characteristics, and fungal pneumonia with brightness, sharp contrast, random distribution of homogeneity attributes. The third similarity of pneumonia is in the gray level that collects each other in the middle. The method used in this study resulted in the same accuracy, sensitivity, and specificity, namely 100%. The image classification in this study shows the success of the texture features and the MLP method in classifying pneumonia images accurately so that they can be used as additional tools that make it easier for medical experts.   ©2020 JNSMR UIN Walisongo. All rights reserved.