Investigation of Deep Learning Algorithms for COVID-19 Detection Based on Chest X-ray Images


Artificial Intelligence Theory and Applications, vol.2, no.1, pp.280-287, 2021 (Peer-Reviewed Journal)


In the last report of World Health Organization 2021, it has stated that 290,000 to 650,000 people may die each year with respiratory diseases caused by seasonal influenza. Additionally, it predicts that influenza will result in more deaths than other illnesses such as flu-related cardiovascular diseases. Sars-Cov-2 virus is also the deadly respiratory disease virus and has different mutations. In 2020, about 1.8 million people died with the Covid-19 virus. The fastest detection and treatment of the disease should be initiated in the fight against COVID-19. The most important indicator that can be used in this regard is radiological data. From this point of view, artificial intelligence (AI) systems based on deep learning will make it easier for radiologists to diagnose the disease in this process. Therefore, in this study, deep convolutional neural networks were used to detect COVID-19 cases from up to date chest radiography images with open source access. DenseNet and SqueezeNet algorithms were used in data set for classification and feature extraction. X-ray images of normal and COVID-19 cases are scaled to 224x224 and the data set is divided into 80% training and 20% testing. The data augmentation process was applied to the images by making angular change, brightness change, horizontal and vertical shifts. In classification with DenseNet and SqueezeNet algorithms, high accuracy values of 99.09% and 97.7%, respectively, were obtained by applying 30 epochs. The results obtained have shown that artificial intelligence algorithms give very high accuracy results in detecting COVID-19 patients using chest X-ray radiography images