Deep Learning Models for Classification of Pediatric Chest X-ray Images using VGG-16 and ResNet-50

Abstract

Medical imaging plays a crucial role in diagnosing various diseases, including pneumonia, in pediatric patients. In this research, we investigate the performance of two convolutional neural networks (CNNs), VGG-16 and ResNet-50, for the task of pneumonia detection using chest X-ray images of pediatric patients aged one to five years old. The dataset consists of anterior-posterior chest X-ray images, categorized into two classes: Pneumonia and Normal. For preprocessing, we standardized the dataset by setting each sample mean to zero and dividing the inputs by the standard deviation of the dataset. Additionally, we applied ZCA whitening to further enhance the data. The two CNN models, VGG-16 and ResNet-50, were trained and evaluated on the dataset. VGG-16, with 16 layers, can classify images into 1000 object categories, while ResNet-50 is a deeper CNN with 50 layers. The experimental results demonstrate that the ResNet-50 model outperformed the VGG-16 model in terms of accuracy and loss during testing. The VGG-16 model achieved a testing accuracy of 74.9% with a testing loss of 48.8%, whereas the ResNet-50 model achieved a significantly higher testing accuracy of 88.9% with a lower testing loss of 28.9%. This study highlights the efficacy of deep learning models in pediatric pneumonia detection and underscores the superior performance of ResNet-50 over VGG-16. These findings have significant implications for developing more accurate and efficient diagnostic tools to aid medical professionals in diagnosing pneumonia in pediatric patients.