Application of Deep Learning Techniques for COVID-19 Management

Gupta, R.; Gupta, A.; Bedi, M.; Pal, S. K.

The sudden penetration of coronavirus has taken several countries in a devastating state worldwide. The novelty of the COVID-19 virus and its wide range of symptoms have posed significant difficulties for the governments, authorities, and the medical community in dealing with this global outbreak. Various researchers have been focusing their studies on finding methods for this pandemic’s forecasting and diagnosis effectively. Evolutionary advancements in Artificial Intelligence, specifically in Deep Learning, have enabled it to become one of the most promising technologies for solving problems related to the current pandemic and other epidemics in the past. This study presents a Concept Mapping based Research Framework, abbreviated as DL-CoTAUM, to review the usage of Deep Learning Techniques for the management of the COVID-19 pandemic using multiple steps. Through the first step, different concepts around Deep Learning are presented. Multiple deep learning techniques using the basic concepts are studied through the second step. In the next step, various epidemics from the past are learned from the Deep Learning perspective. In the fourth step, multiple use-cases for COVID-19 management are covered up, like trend forecasting, virus and drug discovery, patient diagnosis, clinical analysis, and monitoring for maintaining social distancing & face masks. The last step presents the various evaluation parameters available for analyzing the modeling techniques’ performance under Deep Learning. In this review study, it was found that Deep Learning is gaining importance for various applications under COVID-19 management. Techniques like LSTM, CNN, RNN, and GANs were found to be prominently used modeling techniques for growth curve fitting and image analysis. Accuracies of more than 90% have been reported for Deep Learning techniques in COVID-19 use-cases, which is higher than the other baseline methods. This study will be useful for researchers, academicians, and practitioners in Epidemiological Modeling and Image Processing. The proposed research framework can be used as a reference point by other researchers for studying the application of a new concept in a different domain. And this research study will also be useful for naïve users in the field of Deep Learning. © 2022, The Author(s), under exclusive license to Springer Nature Switzerland AG.