Abstract
By the end of the year 2019, a global pandemic novel coronavirus, known as COVID-19, hits the world. The most widely used test for COVID-19 is the Real-Time Polymerase Chain Reaction. However, Real-Time Polymerase Chain Reaction test is time-consuming. Moreover, it suffers from a high false-negative diagnosis rate (low sensitivity). Computed Tomography scans, compared to the Real-Time Polymerase Chain Reaction test, can produce a result in a short amount of time. In this paper, we propose a novel model that hybridizes deep learning and machine learning together. Deep learning is utilized to extract the important features from Computed Tomography images, then the selected features are passed to an ensemble model for the classification. We used RseNet50 for features selection, and the classification is performed by an ensemble model that combines Support-vector Machine, Logistic Regression, and Multilayer Perceptron. The proposed model is compared with eleven state-of-the-art techniques and surpassed them using accuracy, precision, recall, and F1-score. The contribution of this paper is introducing a novel model with high performance for the diagnosis of COVID-19. With the aid of this model, we could identify positive cases rapidly for early isolation. At the same time, we can use it in combination with Real-Time Polymerase Chain Reaction test to increase its sensitivity.