Detection of Pediatric Pneumonia from X-Ray Images using ResNet50 and GAL Networks

Abstract

Known as inflammation of one or two lung tissues, pneumonia affects approximately 150 million people worldwide each year; it is also the leading cause of death in children aged 0-5. For these reasons, early diagnosis of pneumonia will enable patients to start treatment without delay and thus increase the chance of survival. In this study, a deep learning-based system is proposed for automatic detection of pneumonia from chest Xray images. ResNet50, one of the proven deep transfer learning models, is used to extract the features in the images, and an artificial neural network model, Grow and Learn (GAL) network, is used for classification. The generated model was trained and tested on a total of 5856 X-ray images labeled as normal and pneumonia; and an accuracy rate of 92.3% was obtained in the detection of pneumonia. The proposed model performs better than most studies in the literature in detecting pneumonia.