Competitive Deep Learning Methods for COVID-19 Detection using X-ray Images

Abstract

After the World War II, every country throughout the world is experiencing the biggest crisis induced by the devastating Coronavirus disease (COVID-19), which initially arose in the city of Wuhan in December 2019. This global pandemic has severely affected not only the health of billions of people but also the economy of countries all over the world. It has been evident that novel virus has infected a total of 20,674,903 lives as on 12 August 2020. The dissemination of the virus can be regulated by detecting the positive COVID cases as soon as possible. The reverse-transcriptase polymerase chain reaction (RT-PCR) is the basic approach used in the identification of the COVID-19. As RT-PCR is less sensitive to determine the novel virus at the beginning stage, it is worthwhile to develop more robust and other diagnosis approaches for the detection of the novel coronavirus. Due to the accessibility of medical datasets comprising of radiography images publicly, more robust diagnosis approaches are contributed by the researchers and technocrats for the identification of COVID-19 images using the techniques of deep leaning. In this paper, we proposed VGG16 and MobileNet-V2, which makes use of ADAM and RMSprop optimizers for the automatic identification of the COVID-19 images from other pneumonia chest X-ray images. Then, the efficiency of the proposed methodology has been enhanced by the application of data augmentation and transfer learning approach which is used to overcome the overfitting problem. From the experimental outcomes, it can be deduced that the proposed MobileNet-V2 model using ADAM and RMSprop optimizer achieves better accomplishment in terms of accuracy, sensitivity and specificity when contrasted with the VGG 16 using ADAM and RMSprop optimizers.

CO-ResNet: Optimized ResNet model for COVID-19 diagnosis from X-ray images

“…In the literature, we discussed existing studies on COVID-19 for X-ray images. The work of [34] and [60] illustrated binary classification where the work of [34] achieved 80% accuracy for Xception and Incep-tionResNetV2 individually. Moreover, the authors [60] showed the result of VGG 16 model for Adam and RM-Sprop optimizer where Adam optimizer provided the accuracy of 90.55% and RMSprop optimizer also provided the accuracy of 90.55%.…”

“…The work of [34] and [60] illustrated binary classification where the work of [34] achieved 80% accuracy for Xception and Incep-tionResNetV2 individually. Moreover, the authors [60] showed the result of VGG 16 model for Adam and RM-Sprop optimizer where Adam optimizer provided the accuracy of 90.55% and RMSprop optimizer also provided the accuracy of 90.55%. The number of images of the work of [34] and [60] were limited where our proposed work used 5,935 X-ray images.…”

“…Moreover, the authors [60] showed the result of VGG 16 model for Adam and RM-Sprop optimizer where Adam optimizer provided the accuracy of 90.55% and RMSprop optimizer also provided the accuracy of 90.55%. The number of images of the work of [34] and [60] were limited where our proposed work used 5,935 X-ray images. Furthermore, the authors of [58] and [59] depicted three level classification including COVID-19, pneumonia and normal lungs.…”

The Performance Evaluation of Transfer Learning VGG16 Algorithm on Various Chest X-ray Imaging Datasets for COVID-19 Classification

“…We reviewed several studies that made use of the CNN architecture to diagnose COVID-19 on chest X-rays, as shown in Table I. Our findings indicate that the CNN model developed by the visual geometry group with 16 depth layers (VGG16) has been applied in about 50% of the COVID-19 studies [6], [8], [9], [13]- [15]. The VGG16 also performed very well when compared with other established models.…”

“…CNN algorithms are easy to model and reliable. As a result, they are currently the most widely used artificial intelligence (AI) model for the detection of COVID-19 on X-ray images [6]- [13]. We reviewed several studies that made use of the CNN architecture to diagnose COVID-19 on chest X-rays, as shown in Table I.…”