Covid-19 detection on x-ray images using a deep learning architecture

Abstract

Recently, coronavirus disease (Covid-19) has become a serious public health threat, spreading worldwide in a very short time and threatening the lives of millions. Furthermore, many being infected with coronavirus have the potential to transmit the disease without showing any symptoms. Covid-19 causes upper respiratory and lung infections in many patients. With the increasing number of cases and mutations, medical resources are being drained day by day due to the rapid transmission of the disease, and the health systems of many countries are negatively affected. For this reason, it is very important to use available resources appropriately and timely for the detection and treatment of the disease. In this study, VGG16 and ResNet50 deep learning models were used to quickly evaluate x-ray images and to make the pre-diagnosis of Covid-19, and an alternative model was proposed. The proposed model was developed using the convolutional neural network deep learning architecture. VGG16, ResNet50 and the proposed model were trained and tested using a total of 12,739 x-ray images belonging to 6,157 patients (9,121 images with Covid-19 findings and 3,618 with normal findings reported by specialist physicians). As a result of the training of the models, success accuracy of 99.92% in the VGG16 model, 99.65% in the ResNet50 model and 99.76% in the proposed model were obtained.