COVID-19 Detection from Chest X-ray Images using CNNs Models: Further Evidence from Deep Transfer Learning

Abstract

In this study we revisit the identification of COVID-19 from chest x-ray images using Deep Learning. We collect a relatively large COVID-19 dataset comparing with previous studies that contains 309 real COVID-19 chest x-ray images. We prepare also 2000 chest x-ray images of pneumonia cases and 1000 images of healthy chest cases. Deep Transfer Learning is used to detect abnormalities in our images dataset. We fine-tune three pre-trained deep convolutional neural networks (CNNs) models on a training dataset; DenseNet 121, NASNetLarge and NASNetMobile. The evaluation of our models on a test dataset shows that these models achieve a sensitivity rate of around 99.45 % on average, and a specificity rate of around 99.5 % on average. These results could be helpful for an automatic diagnosis of COVID-19 infections, but the clinical diagnosis stills always necessary.