COVID-19 and Associated Lung Disease Classification Using Deep Learning

Coronavirus 2019, well familiar as COVID-19, is a virus that causes significant pneumonia and has varying degrees of severity based on the patient's capability. The coronavirus infection was initially discovered in the Chinese town of Wuhan in Dec. 2019 and quickly spread around the world as a worldwide pandemic. Early detection of positive cases and prompt treatment of infected individuals is required to prevent viral transmission. The necessity of testing kits for COVID-19 has grown, and most of the growing nations are encountering a scarcity of testing kits as new cases emerge daily. In this case, the current study is with the help of radiology imaging techniques, including X-ray, to help in detecting COVID-19. In several disease diagnoses and decision-making circumstances, the information provided in a chest X-ray sample is sufficient to assist medical experts. With the help of a Deep Convolutional Neural Network (CNN), the research proposes an intelligent method to classify various nine diseases, including coronavirus disease 2019, with the help of X-ray instances applying pre-trained DenseNet169 architecture. The fundamental goal of this paper is to classify lung diseases with COVID-19. The used datasets are collected from online repositories, i.e., Kaggle and NIH contained X-ray images of all nine classes. This dataset consists of 1200 images for each class. Various rotations and scaling operations have been applied to the dataset, and the data in the dataset are divided into the test, train, and validation sets. In comparison to other studies in the literature, our models performed well. The highest accuracy attained by DenseNet169 is for COVID-19 with an accuracy of 99.4%, F1-score of 97.5%, precision of 97%, recall of 98%, and specificity of 99.6%. The highest True Positive rate we got in this is 99% for COVID-19, followed by 97% for Cardiomegaly. The minimal rate we got is 88% in Atelectasis. DesnseNet169 proved to be more robust and reliable in classifying nine classes, including COVID-19, after adopting a testing strategy proposed in the literature, making them suitable methods for classification using chest X-ray samples. Which in the future will be helpful for radiologists and physicians during the pandemic of Coronavirus 2019.