Effective Diagnosing of Covid-19 from CXR Images Using Deep Learning Approaches and Optimized XG Boost Model

ABSTRACT

Since the first confirmed incidence of the novel coronavirus Covid-19 in China, it has spread fast around the world, reaching a population of 442,602,593(at the start of 2022), according to World Health Organization figures.Therefore, the diagnosis of the virus is crucial to prevent its separation. However, the tools available for Covid-19 diagnosis are limited compared to the pressure at the increasing number of infected people. Therefore, to prevent the virus thread, it is necessary to find a quick automated system that can handle a bulk amount of data with high accuracy and a lower amount of false positive or false negative. This research presents a hybrid machine learning-based system that uses a pre-trained MobilNet model for feature extraction from chest X-ray images, followed by a dimensionality reduction technique to speed up the classification process and an XGBoost classifier to complete the task. Furthermore, the Bayesian algorithm is used to choose the optimum hyperparameters for the XGBoost classifier. The suggested approach was evaluated on two datasets of X-ray images and produced both high and near results.The results for the first dataset were 97.65% accuracy, 97.63% F1-score, 97.65% recall, and 97.69% precision,and for the second dataset were 96.35% accuracy, 95.82% F1-score, 98.35% recall, and 96.38 precision.