Deep learning classification of Covid-19, pneumonia, and lung cancer on chest radiographs

Abstract

Lung illnesses like lung cancer, Covid-19 and pneumonia have in most cases deadly effects on humans if not immediately treated. In recent times, deep learning with medical imaging, like chest X-rays, has been used for diagnoses and to assist radiographers in several medical applications. In this paper, we investigate using deep learning architecture AlexNet the problem of classifying Covid-19, lung cancer and pneumonia medical images due to the similarities in medical chest X-rays imaging of the three diseases. The comparative results show that the classifier distinguishes Covid-19 from lung cancer with 94 percent accuracy, distinguishes Covid-19 from pneumonia with 96 percent accuracy, and also distinguishes lung cancer from pneumonia with 93 percent accuracy. Overall, AlexNet was able to distinguish Covid-19 from pneumonia with an excellent accuracy that is slightly better than the other two classifications.