ABSTRACT
Since it was firstly reported in China, COVID-19 becomes the cause of death for a large number of individuals and the spreading speed of the disease throughout the world is very high. Due to this, the World Health Organization declares it as a pandemic. Early detection of the COVID-19 has a high probability to cure it. Chest X-ray is the common and a lower cost modality of image for detecting COVID-19, this is because the disease affects the human lung. For processing and analyzing images for the purpose of detecting human disease, deep learning is the most promising technique. Inspired by the previous works, in this work we have taken radiography images to detect COVID-19 by using four pre-trained deep learning techniques, namely VGG19, DenseNet-201, DenseNet-169, and MobileNetV2, which have better performance in the earlier works. We have used a COVIDx dataset, which consists of 15,156 chest X-ray images divided into three classes. After preparing the dataset we have trained and tested the models separately and then we compare the performance of each model. As a result, DenseNet-169 performs better and scores an accuracy, AUC, and loss of 0.94, 0.99, and0.27 respectively. © 2022 IEEE.