Analysis of Fifteen Approaches to Automated COVID-19 Detection Using Radiography Images

ABSTRACT

The COVID-19 pandemic has caused economic, physiological, and psychological harm to the world. A crucial step, hence, in the fight against covid is the highly efficient screening of patient cases. Conventional RT-PCR testing, even though more reliable, cannot be done on every patient as the virus has spread way faster than the world’s resources could afford. One very important screening approach that is being used across the globe is chest X-ray imaging. Since X-ray facilities are readily obtainable in healthcare systems of most countries across the globe, and with more and more X-ray systems being digitized, the cost and time of transportation are cut as well. Hence, if the detection of the virus in a CXR image can be automated using AI techniques, it will save a lot of time and effort of radiologists to have to go through hundreds of such images, and in some cases will also spare the need of doing RT-PCR testing, and since saving resources in this time is vital, automated detection can be very effective. In this work, we will explore, analytically discuss, and do a comparative study of many ML and deep learning techniques that have been taken for automated COVID-19 detection through chest X-rays (CXR). We carefully analyze the papers and derive a set of key factors for discriminating the methodologies, classification techniques, approaches, and the results that yielded. © 2022, The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd.