Detection of Pediatric Pneumonia using Convolutional Neural Networks and Vision Transformer

Pneumonia is one of the primary causes of death worldwide and the leading cause of death among children. Therefore, early detection of pneumonia in children is necessary to ensure that the patient receives the appropriate treatment and that the disease is prevented from spreading. Using Chest X-Ray (CXR), the study employed Convolutional Neural Network (CNN) and Vision Transformer (ViT) models to detect pediatric pneumonia. Resnet50, VGG19, and AlexNet were the CNN models applied to this investigation. In this study, CXR was used to classify child pneumonia into two categories: pneumonia and normal. Compared to other models, the Resnet50 model has the highest efficacy. The accuracy, precision, recall, and F1-measure for the Resnet50 model are respectively 98.21%, 97.68%, 97.77%, and 97.73%. The following models with the greatest performance are VGG19, Vision Transformer, and AlexNet. This demonstrates that the CNN, specifically the Resnet50 model, outperforms Vision Transformer in detecting pediatric pneumonia based on CXR.