A Novel Computer-Aided Approach for Predicting COVID-19 Severity Using Hyperparameters in ResNet50v2 from X-ray Images

Abstract

This research has been globally impacted by COVID-19 virus, which was a very uncommon, highly contagious & dangerous respiratory illness demanding early detection for effective containment and further spread. In this research, we proposed an innovative methodology that utilizes images of X-rays for COVID-19 detection at an early stage. By employing a convolution neural network, we enhance the accuracy performance via using ResNet50v2 using a hyperparameter. The methodology achieves a remarkable accuracy with an average accuracy of 99.12%. This accuracy surpasses other available models based on different deep learning models like VGG, Xception and DenseNet for COVID identification & detection with the help of X-ray images. X-ray scans are now preferably used modality for the identification & detection of COVID-19, given its widespread utilization and effectiveness. However, manual treatment & examination using X-ray images is very challenging, specifically in the field which is facing a limitation of skilled medical staff. Utilization of deep learning models has demonstrated significant potential and effective results in automating the diagnosis for timely identification of COVID with the help of X-ray films. The suggested architecture is specifically developed for timely prediction and analysis of COVID cases employing X-ray films. It firmly believes that this study holds significant potential in alleviating the workload of frontline radiologists, expediting patient diagnosis and treatment, and facilitating pandemic control efforts.