Detection of COVID-19 on Chest X-Ray Images using Inverted Residuals Structure-Based Convolutional Neural Networks

Abstract

China officially reported the COVID-19 coronavirus's existence to the World Health Organization (WHO) on December 31, 2019. Since then, it has spread and has infected millions of people around the world. COVID-19 is a highly contagious disease and it can cause severe respiratory distress. Insevere cases it can result in failure of the function of organs simultaneously. Recent studies haveshown that chest X-rays of patients suffering from COVID-19 show the specific characteristics of those infected with the virus. This paper presents a method to detect the presence of COVID-19 on chest X-ray images based on inverted residuals structure implemented in MobileNetV2 as a base model. We also explore the performance of using a Fully connected layer with dropout and using the Global Average Pooling layer as top layers of the base model to classify each image into COVID-19 or NonCOVID-19. Our proposed method was able to achieve COVID-19 detection with the best accuracy of 0.81, with precision, recall, and F1-score of 0.81, 0.75, and 0.77, respectively, using the Global AveragePooling layer with data augmentation version.