COVID-19 Chest X-rays Classification Through the Fusion of Deep Transfer Learning and Machine Learning Methods

One of the most challenging issues that humans face in the last decade is in the health sector, and it is threatening his existence. The COVID-19 is one of those health threats as declared by the World Health Organization (WHO). This spread of COVID-19 forced WHO to declare this virus as a pandemic in 2019. In this paper, COVID-19 chest X-rays classification through the fusion of deep transfer learning and machine learning methods will be presented. The dataset “DLAI3 Hackathon Phase3 COVID-19 CXR Challenge” is used in this research for investigation. The dataset consists of three classes of X-rays images. The classes are COVID-19, Thorax Disease, and No Finding. The proposed model is made up of two main parts. The first part for feature extraction, which is accomplished using three deep transfer learning algorithms: AlexNet, VGG19, and InceptionV3. The second part is the classification using three machine learning methods: K-nearest neighbor, support vector machine, and decision trees. The results of the experiments show that the proposed model using VGG19 as a feature extractor and support vector machine. It reached the highest conceivable testing accuracy with 97.4%. Moreover, the proposed model achieves a superior testing accuracy than VGG19, InceptionV3, and other related works. The obtained results are supported by performance criteria such as precision, recall, and F1 score.