101. A novel method for detection of COVID-19 cases using deep residual neural network
- Author
-
Noshad, Ali, Arjomand, Parham, Khonaksar, Ahmadreza, and Iranpour, Pooya
- Abstract
ABSTRACTThe COVID-19 pandemic continues to exert destructive effects on the health and well-being of the global population. Previous studies show that screening helps us identify COVID-19 and provides opportunity to isolate infected patients and put them under treatment. Patients infected with prevalent types of chest infections are frequently misdiagnosed with COVID-19. Motivated by this, in the current study, a novel deep residual COVID-Net architecture for identification of COVID-19 using raw chest X-ray images is presented. According to the results, the proposed model was able to identify COVID-19, pneumonia, and normal cases with a classification accuracy of 99.65%. The proposed model has also demonstrated a superiority in terms of sensitivity, specificity, precision, and f1-score and attained the results of 99%, 99.37%, 99%, and 99%, respectively. Additionally, DRCOVID-Net was validated on the chest X-ray images obtained from 19 patients at Namazi hospital in Iran. It was able to identify 12/13 COVID-19 cases correctly with an accuracy of 84%. The ResNet50 was used as a based architecture in our study. From these experiments and evaluations, we can conclude that the proposed model can reliably be employed to assist radiologists in validating their initial screening and also utilize via cloud as a CAD system.
- Published
- 2021
- Full Text
- View/download PDF