COVID-19 prediction through X-ray images using transfer learning-based hybrid deep learning approach. (2022)

• Record Type: Journal Article
• Authors: Kumar, Mohit; Shakya, Dhairyata; Kurup, Vinod; Suksatan, Wanich
• Is Part Of: Materials today. Volume 51:Part 8(2022)
• Page Start: 2520
• Page End: 2524
• Publication Date: 2022
• Subjects: Convolutional Neural Network -- Recurrent Neural Network -- Grad-CAMs -- DarkCovidNet -- CovidGAN; Materials science -- Congresses -- Periodicals; 620.1
• DOI: 10.1016/j.matpr.2021.12.123
• Languages: English
• ISSNs: 2214-7853

Abstract

Abstract: Over the past few months, the campaign against COVID-19 has developed into one of the world's most sought anti-toxin treatment scheme. It is fundamental to distinguish cases of COVID-19 precisely and quickly to help avoid this pandemic from taking a wrong turn with a proper medical reasoning and solution. While Reverse-Transcription Polymerase Chain Reaction (RT-PCR) has been useful in detection of corona virus, chest X-Ray techniques has proven to be more successful and beneficial at detection of the effects of virus. With the increase in COVID patients and the X-Rays done, it is currently possible to classify the X-Ray reports with transfer learning. This paper presents a novel approach, i.e., Hybrid Convolutional Neural Network (HDCNN), which integrates Convolutional Neural Network (CNN) and Recurrent Neural Network (RNN) architecture for the finding of COVID-19 using the chest X-Ray. The transfer learning approach, namely slope weighted activation class planning (Grad-CAMs), is used with HDCNN to display images responsible for taking decisions. In this study, HDCNN is compared with other CNNs such as Inception-v3, ShuffleNet, SqueezeNet, VGG-19 and DenseNet. As a result, HDCNN has achieved an accuracy of 98.20%, precision of 97.31%, recall of 97.1% and F1 score of 0.97. Compared to other current deep learning models, the HDCNN has achieved better results, and this can be used for diagnosis purpose after proper approvals.