AUTOMATIC DETECTION OF COVID-19 AND VIRAL PNEUMONIA IN X-RAY IMAGES USING DEEP LEARNING APPROACH

Abstract

The early detection and treatment of COVID-19 infection are necessary to save human life. The study aims to propose a time-efficient and accurate method to classify lung infected images by COVID-19 and viral pneumonia using chest X-ray. The proposed classifier applies end-to-end training approach to classify the images of the set of normal, viral pneumonia and COVID-19-infected images. The features of the two infected classes were precisely captured by the extractor path and transferred to the constructor path for precise classification. The classifier accurately reconstructed the classes using the indices and the feature maps. For firm confirmation of the classification results, we used the Matthews correlation coefficient (MCC) along with accuracy and F1 scores (1 and 0.5). The classification accuracy of the COVID-19 class achieved was about ([Formula: see text])% with MCC score ([Formula: see text]). The classifier is distinguished with great precision between the two nearly correlated infectious classes (COVID-19 and viral pneumonia). The statistical test suggests that the obtained results are statistically significant as [Formula: see text]. The proposed method can save time in the diagnosis of lung infections and can help in reducing the burden on the medical system in the time of the pandemic.