Classification of Thoracic Pathologies Including COVID-19 in Chest X-ray Images by Using Convolutional Neural Networks

The COVID-19 pandemic has made an enormous impact on countries around the globe. One of the best approaches for tackling complications associated with COVID-19 and other thoracic pathologies is early detection of the diseases. In this work, seven convolutional neural network models for classification of chest X-ray images from four classes (COVID-19 vs lung opacity vs normal vs viral pneumonia) are developed and evaluated. The models were built without utilizing transfer learning. The best performing model is a 5-layer convolutional neural network that achieves a classification accuracy of 90.69% with a recall and precision of 91.25% and 92.03% across the four classes.