A Deep Learning-Based Approach for Detection of Viral and Bacterial Pneumonia from Chest X-Rays

Abstract

Pneumonia is a disease which has been prevalent for a long period of time, despite active attempts to limit the harm caused due to it. The effect it can have on an individual can vary diversely from person to person, with age being a critical factor. Infants and people of an older age tend to be at maximum risk due to this disease, which affects the lungs. Two categories of pneumonia exist, viral pneumonia and bacterial pneumonia. Over the past few years, the deep learning community has made significant contributions, especially in order to assist medical staff in the diagnosis of pneumonia in suspected patients. Several pretrained models, whose weights have been trained on different datasets, have also been applied for this application by making use of transfer learning. In this paper, various such methods have been analyzed. Pretrained models, customized to suit the current purpose of detecting pneumonia, have been implemented to observe the performance. Along with this, we have proposed our own convolutional neural network architecture in order to detect pneumonia. We compare the results of these models with our proposed system, by performing experiments on a dataset containing three categories: normal, bacterial pneumonia and viral pneumonia.