Pneumonia Detection Using Convolutional Neural Networks

Abstract

Pneumonia is an infectious and deadly illness in respiratory that is caused by bacteria, fungi, or a virus that infects the human lung air sacs with the load full of fluid or pus. Chest X-rays are the common method used to diagnose pneumonia and it needs a medical expert to evaluate the result of X-ray. The troublesome method of detecting the pneumonia cause a life loss due to improper diagnosis and treatment. With the emerging computer technology, development on an automatic system to detect pneumonia and treating the disease is now possible especially if the patient is in a distant area and medical services is limited. This study intends to incorporate deep learning methods to alleviate the problem. Convolutional Neural Network is optimized to perform the complicated task of detecting diseases like pneumonia to assist medical experts in diagnosis and possible treatment of the disease. The authors developed several models to determine the best possible model in detecting pneumonia with the most accurate results. This study has trained five different models of CNN, namely AlexNet, LeNet, GoogleNet, ResNet and VGGNet using 1024 by 1024 resolution of 26,684 dataset images. The result achieved a 97 percent accuracy rate for VGGNet and the lowest rate is 74 percent achieved by the ResNet model. The result of statistics shows that the trained model was able to detect Pneumonia through examined images of chest X-ray.