Prediction of COVID-19 Through Chest X-Ray Images Employing Various Machine Learning Techniques

ABSTRACT

The COVID-19 is the most infectious disease which is recently discovered. The COVID-19 pandemic has led to excruciating loss to human life and it also caused mild to severe respiratory illness, including death. Detecting the infected patients and taking special care is crucial in fighting covid-19. Radiography and Radiology images are used to diagnose the patients. These are the fastest ways to identify COVID-19 disease. It is observed that patients affected with COVID-19 have specific abnormalities in their chest radiograms. Initially there were few limited set of CT images are available publicly in performing research. Board-certified radiologist role is to perform identification of images exhibiting COVID-19 disease. Chest CT scans are helpful to diagnose COVID-19 disease in individuals. COVID-19 directly shows impact on lungs and it damages and the tiny air sacs. In this paper we have used deep transfer learning models Residual Network (ResNet50) and VGG19 (Visual Geometry Group) to predict the disease at earlier stages. These models obtained a specificity rate of 90% and achieved a highest sensitivity rate of 98 %. In addition to sensitivity and specificity rates ROC curve, average prediction and confusion matrix of each model are presented in the papers. While this achieved performance is very encouraging if we have large set of COVID-19 images then it may give more reliable estimation of accuracy rates. © 2022 IEEE.