Detection of pneumonia from chest X-ray images using machine learning

Abstract

A bacterial infection is the cause of the lung condition known as pneumonia. An essential component of a successful treatment procedure is early diagnosis. Without early diagnosis, pneumonia can be severe or even can cause death. Viewing X-ray images is one of the ways to detect pneumonia. For accurate viewing or reading of X-ray images, a computer-based algorithm is preferable over reading X-ray images manually. In this study, a pneumonia detection system is created using grounded feature extraction from convolutional neural networks (CNN). To predict the occurrence of pneumonia, different classification algorithm models are used. For classi-fication, customized CNN models and various pre-trained models such as VGG-16, Inceptionv3, ResNet50, and VGG-19 are applied to the x-ray image dataset. After implementing all these models we obtained our best accuracy from the Customized CNN model which is 90.43% and the best f1-score from Customized CNN, ResNet50, and VGG-19, the score is 0.87.