Classification of Chest X-Ray Images to Diagnose COVID-19 Disease Through Transfer Learning

Abstract

The world has been trapped in a pandemic caused by the deadly SARS-COV2 virus, which is very contagious and is rapidly spreading across the world. It affects humans’ respiratory organs and later causing breathing problems, cough, loss of smell, fever, and other symptoms. With the rapidly growing Coronavirus cases, the demand for its testing has increased drastically; however, the labs capable of performing the diagnosis are minimal. Thus, many patients cannot get them diagnosed due to a lack of testing facilities available nearby. The purpose of this paper is to aid the Coronavirus scanning process with Artificial Intelligence for a faster and cost-efficient scanning mechanism. A severe respiratory ailment triggers the sudden occurrence of the Coronavirus disease or Coronavirus. Due to the multiplicative rate of its spread, large population has been infected and is increasing every day at exponential speed. Performing tests to diagnose the Corona positive patients in this large population has become the most significant challenge due to limited resources. In this project, a tool has been proposed based on deep learning algorithms that will be capable of diagnosing the Coronavirus using chest x-rays. The training data contains images of Coronavirus and Pneumonia cases used to train CNN based models. Six training approaches will be performed, including VGG16, VGG19, InceptionV3, ResNet50, and DenseNet 201, and further, the best performing model will be considered for prediction. Having such a model will help us reduce the diagnosis time and the cost; it will also increase the capability to test a more significant number of people daily.