Performance Assessment of DL Model in Iraq for Covid-19 Patient Severity Prediction from X-Ray Scan Images

Abstract

The livelihoods of many people are greatly affected by the covid-19 virus, and the high death rate has led to a global pandemic. With early detection, the possibility of spreading coronavirus (covid-19) can be reduced. The way people live their lives and the global economic and social systems have undergone a major transformation. It is difficult to treat almost all cases of coronavirus due to limited medical infrastructure, causing the death toll to rise rapidly. Therefore, thousands of lives could be saved if their occurrence and severity could be predicted in advance, allowing the rapid adoption of appropriate treatments. Deep learning (DL) is crucial for determining the severity of the lungs in patients with Covid-19. The severity of lung disease among Covid-19 patients is determined using a variety of techniques, including X-rays, CT scans and MRI scans. The prediction result depends highly on how well each stage of lung disease detection performs. The low prediction accuracy leads to a major reason: the large size of the storage model. To address this problem, in order to increase predicting accuracy, it is suggested that the new deep transfer learning model be enhanced by the incorporation of a novel attention mechanism. VGG16 is used as the foundation model for a brand-new deep transfer learning model. We suggest adding a convolutional block attention module (GhostNet) to the conventional suggested network model and upgrading a new model for this purpose in order to improve the accuracy of forecasting the severity of lung illness among Covid-19 patients.