ABSTRACT
Coronavirus of 2019 is an ongoing pandemic that has infected millions of people and costed the life of more than three million people. It is a highly transmitting disease that has exhausted all the healthcare facilities trying to contain its spread. It has exposed the need for more health facilities and experts to cope with this pandemic without impacting on the safety of healthcare workers. The hardworking and struggling healthcare sector is in need of automated diagnostic devices that could lift the burden off the limited practitioners and also ensure their safety from coming in direct contact with the infection. This pandemic has made the world realize the need of automation for an infectious disease like COVID-19. Deep learning in radiology is an extensively researched topic over the last decade and has the potential to provide the much-needed automation required for COVID-19 diagnosis. In this paper we have fine-tuned three deep learning models—ResNet50, DenseNet121 and InceptionV3—for classification of COVID-19 CXR from regular pneumonia cases. Our models achieved an accuracy of 99.45, 99.50 and 98.55 respectively. © 2022, Springer Nature Switzerland AG.