A comparison based analysis on the performance of deep neural network models in terms of classifying pneumonia from chest x-ray images

Abstract

Pneumonia is one of those alarming diseases which causes a huge mortality rate among children and older people with 2 million deaths each year. People from the poor regions of Africa and Asia are mostly affected by pneumonia because of low medical monitoring in those regions. In recent times, a lot of computer aid based diagnostic systems have been developed in order to provide assistance in terms of detecting pneumonia. In this research work, we have proposed a convolutional neural network (CNN) based model comparison system for chest X-ray images to classify and detect pneumonia. A dataset containing 2,861 chest X-ray images of normal and pneumonia affected patients have been used to classify pneumonia from analyzing the lung images. We used 3 different neural network architectures: VGG16, Inception v3, ResNet50 in order to classify Pneumonia. After classification, we compared the result and we achieved a maximum of 95.0% accuracy, 94% precision, 96.40% sensitivity, 92.80% specificity from VGG16.