Automatic Detection of Covid-19 from Chest X-ray Images using Corona Net

Abstract

The most devastating pandemic to ever infiltrate humans is COVID-19. An automatic detection system is an instantaneous diagnosis option to prevent COVID-19 transmission. The objective of this research work is to propose a novel CNN (Convolutional Neural Network) based Covid-19 detection system to classify the radiological (chest X-ray) images into binary classes (Covid-19 and Non-Covid-19) and three (multi) different classes ( Normal Lungs, Lungs infected by Covid-19 and Lungs infected by Pneumonia). The efficiency of the proposed CNN(CoronaNet) model is compared with six existing pre-trained models (AlexNet, GoogleNet, VGG-16, SqueezeNet, Inception-V3 and ResNet-50) for identifying Covid-19 from radiological images. The computer experimental results demonstrate that the proposed CoronaNet model has achieved an overall accuracy of 96.4% for binary-class classification (Covid-19 and Non-Covid-19) and 94.4 % for multi- class classification (Normal, Covid-19 and Pneumonia). The proposed technique could be a useful tool for radiologists to diagnose and treat Covid-19 patients promptly.