An Effective Deep Learning Model to Discriminate Coronavirus Disease From Typical Pneumonia

Introduction

In recent few years, human civilizations over the world have been threatened by emerging dangerous species of Coronaviridae termed Coronavirus disease (COVID-19) that has already given rise to millions of infected people and rapidly increased the death rate. Consequently, the early recognition infected cases is extremely significant to perform the medical treatment processes and the procedures of preventive containment (Abdel-Basset et al., 2021). Though numerous criteria provide successful COVID-19 diagnosis for people, the tools of clinical laboratory based on RT-PCR and sequencing of virus nucleic acid suffer from various deficiencies, such as the test results become prepared at the earliest twenty-four hours concerning crucial cases and usually require many days to dedicate a decision (Kumar et al., 2022). Recently, the world health organization announced that RT-PCR may provide incorrect results in COVID-19 cases owing to low-quality specimens obtained from patients, unsuitable specimens processing, and taking specimens at the late or early stages of the disease. The alternative common technique of COVID-19 diagnosing utilized today is Computed Tomography (CT). However, this technique is not easily accessible, and it is very costly. The majority common technique that medical specialists utilize for monitoring, triaging, and diagnosing varieties of pneumonia and COVID-19 disease courses is chest X-rays radiography. In contrast to RT-PCR and CT techniques, having a chest X-ray image is inexpensive and requires only a few seconds. Therefore, X-ray imaging holds a considerable possibility to be an alternate technique to other tools. The COVID-19 manual diagnosis may be prone to human error which leads to consuming time, and accordingly, this requires the help of adequate radiologists to realize high accuracy of diagnosis (Yamaç et al., 2021).

At present, various medical health issues and complications such as breast cancer diagnosis, brain tumor diagnosis, and many more are utilizing machine and deep learning models-based solutions (Fati et al., 2022; Mathiyazhagan et al., 2022; Mohanty et al., 2021; Flayyih et al., 2020; Hasan et al., 2020; Hussien et al., 2020; Inbarani et al., 2020; Kumar et al., 2019, 2015; Hassanien et al., 2014; Emary et al., 2014a,b; Aziz et al., 2013; Jothi et al., 2020, 2019a,b, 2013; Anter et al., 2013; Azar et al., 2013, 2012). The technology of deep learning can disclose the features of the image that are non-obvious in the original image (Boulmaiz et al., 2022, Zaidi et al. 2022; Azar et al., 2021; Koubaa et al., 2020; Elkholy et al., 2020; Ibrahim et al., 2020). Specifically, Convolutional Neural Networks (CNNs) have been mostly adopted via the community of researches (Al-Dulaimi et al., 2022; Ramadan et al., 2022; Aslam et al., 2021). The recent developments of CNN models have appeared with a successful invention in the field of natural images analysis, and in other computer vision areas. These models are capable of extracting and learning enhanced representations of visual features. Such developments supply additional proof that optimal performance can be obtained via deep architecture (Deepalakshmi et al., 2021; Waleed et al., 2021). The CNN-based image recognition models have been capable of distinguishing a considerable number of targets, and the obtained recognition accuracies exceed the standard individual’s level (Jiang, 2022). Therefore, it can be concluded that, deep CNN can be exploited for realizing such developments also in the COVID-19 discrimination too.

The contribution of this work is as follows: