COVID-19 classification from chest x-ray images via convolutional neural networks, transfer learning, and support vector machines

Coronavirus disease (COVID-19) is an infectious disease that causes severe acute respiratory syndrome caused by SARS-CoV-2. Fast, accurate and early diagnosis of the disease causing a worldwide pandemic is important due to its high rate of transmission and deathrates. In the literature, many deep learning-based classification studies have been carried out to assist for the diagnosis of COVID-19. In this thesis, a deep transfer learning method has been proposed to classify Covid-19 disease from chest X-ray radiographs. In addition to other studies, COVID-19 chest X-ray appearances (typical, atypical, indeterminate) have also been used in convolutional neural network training. Besides, in order to demonstrate the classification ability of the convolutional neural network models, the same classification operations were performed with the support vector classifiers. According to the test findings, high performance in classification of typical appearance has been determined in spite of lower performance in atypical and indeterminate appearances.