Transfer learning-based method for detection of COVID-19 using X-Ray Images

Rehman, Abdul; Tariq, Zain; Jan, Sana Ullah; Aziz, Sumair; Khan, Muhammad Umar; Chaudry, Hassan Nazeer

Abstract

In this paper, we have performed transfer learning using different pre-trained convolutional neural networks for binary classification of X-ray images into COVID-19 disease and normal. The dataset is gathered from two open sources. Our dataset is consisting of 254 COVID-19 and 310 Normal X-ray images. The pandemic situation all around the world demands an efficient solution so that the disturbance of global health, daily life, and economy can be controlled. In this regard, we introduced the deep feature fusion-based technique which could help to design an embedded system. We fine-tuned and trained the thirteen independent pre-trained models and we found that the Resnet50V2 model performed efficiently for binary classification scenarios. Our proposed technique using transfer learning gives a detection rate of 99.5% for binary classification (Normal and COVID).

Citation

Rehman, A., Tariq, Z., Jan, S. U., Aziz, S., Khan, M. U., & Chaudry, H. N. (2021, October). Transfer learning-based method for detection of COVID-19 using X-Ray Images. Presented at 2021 International Conference on Robotics and Automation in Industry (ICRAI), Rawalpindi, Pakistan

• Presentation Conference Type: Conference Paper (published)
• Conference Name: 2021 International Conference on Robotics and Automation in Industry (ICRAI)
• Start Date: Oct 26, 2021
• End Date: Oct 27, 2021
• Online Publication Date: Dec 28, 2021
• Publication Date: 2021
• Deposit Date: Feb 14, 2022
• Publisher: Institute of Electrical and Electronics Engineers
• Book Title: 2021 International Conference on Robotics and Automation in Industry (ICRAI)
• DOI: https://doi.org/10.1109/icrai54018.2021.9651463
• Keywords: transfer-learning, COVID-19, deep feature extraction, feature selection, k-nearest neighbor
• Public URL: http://researchrepository.napier.ac.uk/Output/2845390
• Publisher URL: https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=9651463