Deep Networks Based Classification of COVID-19 Chest X-Ray Images

• Published in: 2020 14th International Conference on Open Source Systems and Technologies (ICOSST) pp. 1 - 6
• Main Authors: Baseer, Amina, Bhatti, Naeem
• Format: Conference Proceeding
• Language: English
• Published: IEEE 16.12.2020
• Subjects: COVID-19; Deep feature extraction; Deep transfer learning; Diagnostic radiography; Feature extraction; Radiographic Images; Solid modeling; Support vector machines; Tuning; X-ray imaging
• DOI: 10.1109/ICOSST51357.2020.9333076

The worldwide spread of COVID-19 (Corona virus) has been declared a global health crisis by WHO. In order to perform early diagnosis of COVID-19 infection a computer-aided diagnosis (CAD) system is of great importance. Chest X-ray radiographic imaging is a preferred diagnostic mean used by radiologists. Machine and deep learning algorithms have been found useful in many CAD equipped health care systems. In this paper, we present an approach to classify chest X-ray radiographic images into normal, COVID-19, and viral-pneumonia categories. The proposed approach performs feature extraction using pre-trained and fine tuned deep networks using transfer learning. A support vector machine (SVM) is trained using the extracted features and used for classification. We collect a dataset comprised of publicly available chest X-ray images from normal, COVID-19 and viral-pneumonia patients. Experimental results demonstrate that the proposed approach achieves 99.29% classification accuracy in case of COVID-19 images and an overall classification accuracy of 97.36%, which is better than most of the existing approaches.