Detecting Covid-19 in Chest X-Ray Images With Convolutional Neural Network

The World Health Organization (WHO) had declared Coronavirus disease (COVID-19), a deadly respiratory illness, as a pandemic in March 2020. The number of infected and deaths in COVID-19 positive patients has been rising at a distressing rate. A massive number of chest X-rays to analyse and the exhaustion from monotonous routine work may become prone to the escalation of error rate. For this reason, the use of the artificial intelligence-based tool will be beneficial in detecting subtle COVID-19, as well as to prevent false diagnoses on patients. In this study, several multi-class image classifications were developed using pre-trained models, namely VGG16, ResNet50, MobbileNetV2, and Xception. A publicly available dataset of 14,220 chest X-ray images comprised of four different classes was used. Extensive evaluation and analysis of the Convolutional Neural Network model in predicting COVID-19 positives have been conducted. The MobileNetV2 model has been demonstrated as the best performing model out of four classifiers with an overall accuracy score of 92.48%. The model also yields 92.67% precision and 92.48% recall score, thus exhibiting a high success in the detection of COVID-19.