Feature extraction with capsule network for the COVID-19 disease prediction though X-ray images. (2022)

• Record Type: Journal Article
• Authors: Darji, Pinesh Arvindbhai; Nayak, Nihar Ranjan; Ganavdiya, Sunny; Batra, Neera; Guhathakurta, Rajib
• Is Part Of: Materials today. Volume 56:Part 6(2022)
• Page Start: 3556
• Page End: 3560
• Publication Date: 2022
• Subjects: Batch normalization -- Max pooling -- L2 regularization -- Sensitivity -- U net model; Materials science -- Congresses -- Periodicals; 620.1
• DOI: 10.1016/j.matpr.2021.11.512
• Languages: English
• ISSNs: 2214-7853

Abstract

Abstract: Past couple of years, the world is going through one of the biggest pandemic named COVID-19. In the mid of year 2019, it is a very difficult process to predict the COVID-19 just by viewing the images. Later on AI based technology has done a significant role in the prediction of COVID-19 through biomedical images such as CT scan, X ray etc. This study also implemented the deep learning model for the prediction of COVID-19 through X-ray images. The implemented model is termed as XR-CAPS which consist of two models such as U-Net model and the capsule network. The U Net model is used for performing the segmentation of the images and the capsule networks are applied for performing the feature extraction. The XR-CAPS model is applied on the X-ray images for the prediction of COVID-19 and the evaluation of the model is done by three parameters that are accuracy, sensitivity and specificity. The model is compared with other existing models like ResNet50, DenseNet121 and DenseCapsNet, this has achieved an accuracy of 93.2%, sensitivity of 94% and specificity of 97.1% which is better than other states of the art algorithms.