Convolutional neural network based SARS-CoV-2 patients detection model using CT images. (30th July 2021)

• Record Type: Journal Article
• Authors: Khan, Shahnawaz; Thirunavukkarasu, K.; Hammad, Rawad; Bali, Vikram; Qader, Mohammed Redha
• Is Part Of: International journal of intelligent engineering informatics. Volume 9:Number 2(2021)
• Page Start: 211
• Page End: 228
• Publication Date: 2021-07-30
• Subjects: deep learning -- CNN -- convolutional neural network -- classification -- covid19 -- SARS-CoV-2 -- image classification; Artificial intelligence -- Engineering applications -- Periodicals; Engineering -- Computer programs -- Periodicals; Knowledge management -- Periodicals; 620.0028563
• Languages: English
• ISSNs: 1758-8715

Abstract

The COVID-19 disease caused by the SARS-CoV-2 infection has widely spread round the globe. Due to the large number of infected cases and rapid spread, it has been declared a global pandemic by World Health Organization on March 2020. There are several methods that identify and detect the COVID patient. However, detection using these methods can be confirmed after up to 10 days of the infection. This research presents a convolutional neural network (CNN) based classification model for detecting a COVID patient using CT image of patient. The dataset, used for the study, consists of CT images of variable sizes. It is a challenge for building a CNN model for variable sizes of the input image. This research uses a hybrid technique to overcome this challenge. It employs and analyses three different methods (such as Adam optimiser, Stochastic gradient descent with momentum optimiser, and RMSProp optimiser) for building the CNN model. Among the three CNN models, for CT image-based classification for infected or non-infected patient, adam performs better than RMSprop and sgdm. The classification accuracy achieved using adam is 94.9%, while RMSprop achieved an accuracy of 91.8% and sgdm reached 93.1%.