Hybrid approach for COVID-19 detection from chest radiography

Abstract

Automatic and rapid screening of COVID-19 from the chest X-ray and Computerized Tomography (CT) images has become an urgent need in this pandemic situation of SARS-CoV-2 worldwide. However, accurate and reliable screening of patients is a massive challenge due to the discrepancy between COVID-19 and other viral pneumonia in both X-ray and CT images. Several models were introduced, but always there was a glitch that might be due to the use of a single classifier, and this reduces their accuracy. In this paper, we study the use of multi-classifiers and show their effect on different models working on X-ray and CT images. We perform a comparison study to show the high impact of ensemble stacking approach on top performer CNN models that recorded the highest detection accuracy in image detection and classification: COVID-Net, VGG16, ResNet, Bayesian, DenseNet, and DarkNet. We presented multi-classifiers instead of a single classifier stacked in an ensemble stacking approach for the diagnosis of the COVID19 from the Chest CT and Xray images. We provide a quantitative evaluation of the proposed ensemble stacking approach on two types of datasets: X-ray images and CT images datasets, with percentages reaching 99%. Keywords— COVID-19, stacked algorithm, ensemble technique, deep learning, chest X-ray images, Computerized Tomography (CT) images.