COVID-19 Detection: An Approach Using X-Ray Images and Deep Learning Techniques

ABSTRACT

In the recent history of human civilization, a pandemic affecting such an enormous population like COVID-19 was about 140 years ago-The Smallpox Worldwide Epidemic (1877–1977, Deaths-500 M). It can be easily inferred that the health management system over the globe in the nineteenth century was too underdeveloped than that of today, which also refers to the fact that the present epidemic must not be allowed to last much longer as the number of deaths is increasing nonlinearly (506 K, with 10.3 M affected). While the medical community around the globe is striving to find a permanent cure, it becomes evident responsibility of all professionals who can contribute in stabilizing the medical management systems of countries particularly underdeveloped/developing countries or those with highest rate of increase in COVID-19 cases like USA, Brazil. In this regard, this study introduces a fast, robust and practically effective method for detection of COVID-19 from chest x-ray images utilizing enhanced deep learning techniques. An object detection network is proposed to be trained with publicly existing datasets. In this model, SSD is used with ResNet101 as a base layer and some pre-processing, achieving a sensitivity of 0.9495 and a specificity of 0.9247. If practically implemented, this can prove very beneficial in aiding economies and health systems of the above-mentioned countries. © 2021, The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd.