Article Preview
Top1 Introduction
Pneumonia caused by the 2019 novel coronavirus infection was named COVID-19 by the World Health Organization (WHO), which has brought a substantial negative impact on the global economy and life. COVID-19 is highly contagious, and its pathogen, SARS-CoV-2, continues to evolve and mutate, producing mutant strains with changes in transmissibility and virulence (C. Wang & Han, 2022). WHO names new coronavirus strains with Greek letters from May 2021, such as Alpha (B.1.1.7), Beta (B.1.351), Gamma (P.1), Delta (B.1.617.2) and Omicron (B.1.1.529) (WHO, 2022). These variants have evolved to spread faster than the original strain (Darby & Hiscox, 2021; Satyanarayana, 2021), making it more difficult to control the virus. Therefore, a rapid and accurate diagnosis of COVID-19 plays a crucial role in the treatment of patients and prevention of the spread of the disease.
In recent years, artificial intelligence (AI) (Y. Li, 2017; Liu, 2016; Y. Zhang, 2016) has been used to identify, classify, and diagnose medical images. In the screening, diagnosis and prediction of COVID-19 (Y.-D. Zhang, 2021), deep learning-based techniques were introduced (Khan et al., 2021; Santoni, Sensuse, Arymurthy, & Fanany, 2015). Some researchers have adapted DL-based techniques to the COVID-19 pandemic, such as CNNs, RNNs, and LSTMs (Heidari, Jafari Navimipour, Unal, & Toumaj, 2022).
Serte et al. (Serte & Demirel, 2021) used a ResNet-50 deep learning model to predict COVID-19 on each CT image of 3D CT scans, providing 96% AUC for detecting COVID-19 in CT scans. Cosimo Ieracitano et al. (Ieracitano et al., 2022) proposed a fuzzy-enhanced deep learning approach to distinguish CXR images of patients with Covid-19 pneumonia and patients with interstitial pneumonia unrelated to Covid-19. The experimental results showed that the accuracy of the classification performance was as high as 81%. Khabir Uddin Ahamed et al. (Ahamed et al., 2021) developed a deep learning-based COVID-19 case detection model, and an improved ResNet50V2 architecture was used as the deep learning architecture. The experimental results show that the combined accuracy of detecting three classes of cases (COVID-19/Normal /Community-acquired pneumonia) and two classes of cases (Normal/COVID-19) using chest CT scan images was 99.012% and 99.99%, respectively. In Ref. (Hammad et al., 2022), the DLM consists of convolutional and pooling layers for COVID-19 detection based on chest X-ray images, and batch normalization was used to reduce overfitting. Robert Hertel et al. (Hertel & Benlamri, 2022) designed a deep learning pipeline with a segmentation module and ensemble classifier to more accurately diagnose COVID-19 patients with an accuracy of 91 percent and sensitivity of 92 percent. In Ref. (R et al., 2022), an efficient hardware architecture based on integrated deep learning models was built for chest X-ray (CXR)-based recognition of COVID-19, which integrates five deep learning models, namely ResNet, FitNet, IRCNN, EffectiveNet, and Fitnet. Computation time, energy, and latency are minimized.