Covid-19 Detection From X-Rays Images Using Deep Learning Methods

Georgios Sapountzakis, Paraskevi-Antonia Theofilou, Paraskevi K. Tzouveli

Abstract: COVID-19 is a serious and highly contagious disease that causes infection of the respiratory system. Its severity varies from person to person. Many times patients’ health deteriorates rapidly and they need specialized medical care. For this reason, its early and reliable detection is important. Chest X-ray (CXR) imaging is widely used as an easy, accessible, economical and valid mean of COVID-19 diagnosis. In this work, using the COVID-QU-Ex dataset, which is the largest dataset of such images, we aim to build a deep learning system to produce efficient and reliable predictions in the detection problem of the disease COVID-19. Therefore, we focus on the study and comparison of state-of-the-art Deep Convolutional Neural Networks (DCNNs) for classification of CXR images and we use Grad-CAM to test their explainability.