Analysis of Deep Learning algorithms on COVID-19 Radiography Database

  • Aman Jaiswal, Ankur Singh Bist

Abstract

The Coronavirus disease, caused by severe acute respiratory syndrome coronavirus 2(SARS-CoV-2) virus is obstreperous in large parts of the world. Various techniques like genome sequencing, tomography imaging, and electron microscopy were used initially to detect the presence of COVID19 but these techniques take more a day time to produce the result. This paper aims to analyze the various deep learning algorithms on the radiography database that could be useful in diagnosing the COVID19 existence. CNN has achieved a lot in various fields of computer vision problems in recent years. CNN architectures used in this experimental analysis include basic models like simple CNN, LeNet5 in continuation with large models like VGG, DenseNet, ResNet, Inception, NasNet, and MobileNet. Further, a majority rule has been proposed where decision produced by all networks will be in consideration to derive final results. One can take the top five models based on an accuracy metric and involve them in voting. COVID-19 detection will be termed as positive when more than 50% nodes are in favour otherwise it will be termed as negative. As the number of cases is rising at a rapid rate, so the medical community isn’t able to tackle the problem effectively. Hence there is a need for fast and reliable means for quick diagnostic of the virus. This novel approach presented will be crucial in large population screening, the prognosis of the inflection, prioritizing the use and allocation of limited resources available specifically in developing countries. It will also help design targeted responses within the limited time frame (of a few minutes) and lower cost as compared to typical existing testing procedures available.

Published
2020-06-04
How to Cite
Aman Jaiswal, Ankur Singh Bist. (2020). Analysis of Deep Learning algorithms on COVID-19 Radiography Database. International Journal of Advanced Science and Technology, 29(11s), 1268 - 1275. Retrieved from https://sersc.org/journals/index.php/IJAST/article/view/20825
Section
Articles