Pneumonia Detection from Chest X-Ray Using Binary Classification

Abstract

In this era of science, pneumonia kills thousands of people every year. People get pneumonia from bacteria or sometimes virus-like covid-19. To ensure proper treatment, it is necessary to detect pneumonia at the early stage. For this reason, scientists, researchers, specialists are looking for deep learning and image processing-based systems where they want to use X-Ray images or Computed Tomography (CT) images for detecting pneumonia within a short period and high accuracy as the conventional method of detecting pneumonia is time-consuming and sometimes are subject to disagreement between the specialists. In this paper, the authors have presented an image processing-based system to detect the normal and abnormal (having pneumonia) images from their chest X-rays. This binary classification-based pneumonia detection system takes a total of 5,856 images where 1,583 were normal and 4,273 were the images having pneumonia. For testing purposes, three methods have been used i.e.; ResNet50V2, ResNet152V2, and InceptionResNetV2. Among these three methods, ResNet50V2 shows a maximum accuracy of 95%.