COVID-19 Detection From Chest X-Ray Using Deep Learning and Contrast Enhancement

Abstract

In the current COVID-19 pandemic, it has become extremely important to detect the affected patients as soon as possible and isolate them in order to break the chain of the spreading virus. Testing in large numbers at laboratories has overwhelmed their resources. Furthermore, the diagnosis report often takes more than a day to be returned. All this adds up to the incapability of our healthcare infrastructure to test all the possibly infected patients. Radiologists across the world have used chest X-rays to detect chest diseases. X-rays being readily available in far less time than RT-PCR reports make them an easy and quick alternative in comparison to current testing methods. However, examining a vast number of X-rays in an already overwhelmed healthcare facility may still lead to delays in determining the presence of the disease. In addition, it would require expertise and profound knowledge about the much recently explored COVID-19 virus in order to make an accurate assessment of the X-rays. In this study, to find solutions to these problems, we have made use of deep learning for the detection of coronavirus. The proposed system uses three different Convolutional Neural Network (CNN) models to detect COVID-19 from pre-processed chest X-ray images with reliable accuracy and hence provide an alternative for people to be aware of being infected rather than wait days for results.