Automated Detection of Lung Diseases (covid-19) Based on X-ray Images Using A Deep Learning Approach

Abstract

COVID-19 pandemic has presented an unprecedented threat to the global public health system. The respiratory tract's epithelial cells which line the airways are the target of the virus's primary attack. Humans are susceptible to respiratory infections caused by it, which can have mild to severe symptoms like coughing, fever, and weakness. However, it has the potential to develop into other lethal diseases and has already been the cause of millions of fatalities. Thus, a precise diagnosis of these disorders is essential in the current healthcare system. Additionally, quicker chain breaks during early identification result in less disease burden on society. Because of a lack of Reverse Transcription Polymerase Chain Reaction (RT-PCR), early illness diagnosis is challenging. Deep learning (DL) models based on radiographic images could be used to tackle COVID-19. The study offers a Convolutional Neural Network (CNN) model built on a Visual Geometry Group (VGG-16) for classifying and identifying individuals with COVID-19 infection in X-ray chest images. The datasets included 7,245 X-ray images and included 2,898 images for the binary classes and 2,898 images for multi-classes for the model's training and testing. In order to maximize the accuracy of classification, the image enhancement approach was used to highlight important information in the image and minimize some secondary information. Gamma Correction, Histogram Equalization (HE), and Contrast Limited Adaptive Histogram Equalization (CLAHE) were used as three filters for the preprocessing approaches. The suggested model performed most accurately in three classes (97.3% with the use of CLAHE) and the binary class (99.7% with Gamma correction). The system's classification accuracy for lung disorders is higher than that of other DL systems.