Application of deep learning techniques in supportting the diagnosis of pneumonia through X-ray images

Abstract

In the context of the outbreak of the COVID-19 epidemic in Vietnam and around the world, an artificial intelligence application that accurately diagnoses pneumonia will help reduce time and human resources for medical examination and treatment. This helps patients receive timely treatment, reducing the risk of aggravation and death. This paper presents the characteristics of modern deep learning network architectures based on convolutional neural networks such as ResNet50, VGG16, Inception, DenseNet. Thereby performing a test to evaluate these models in the diagnosis of pneumonia using the Chest-Xray dataset. The test results show that the deep learning model using VGG16 deep learning network architecture gives the highest accuracy rate. This is the basis for proposing to build an application to support effective pneumonia diagnosis based on X-ray images.