PREDICTING COVID-19 FROM CHEST X-RAY IMAGES USING OPTIMIZED CONVOLUTION NEURAL NETWORK

J. P. Oguntoye; O. O. Awodoye; J. A. Oladunjoye; B. I. Faluyi; S. A. Ajagbe; E. O. Omidiora

J. P. Oguntoye Department of Computer Engineering, Faculty of Engineering & Technology, Ladoke Akintola University of Technology (LAUTECH), Nigeria
O. O. Awodoye Department of Computer Engineering, Faculty of Engineering & Technology, Ladoke Akintola University of Technology (LAUTECH), Nigeria
J. A. Oladunjoye Department of Computer Science. Faculty of Computing and Information System, Federal University Wukari. Wukari, Taraba state.
B. I. Faluyi Department of Computer Science, School of Science and Computer, The Federal Polytechnic Ado-Ekiti, Ekiti State, Nigeria
S. A. Ajagbe Department of Computer & Industrial Production Engineering, First Technical University Ibadan, Nigeria.
E. O. Omidiora Department of Computer Engineering, Faculty of Engineering & Technology, Ladoke Akintola University of Technology (LAUTECH), Nigeria.

Keywords: Convolution, Neural Network, Particle Swarm Optimization, Deep Learning, COVID-19, Chest X-Ray

Abstract

Machine learning is emerging as a unique powerful method to improve the diagnosis and prognosis of several multifactorial diseases, including COVID-19. The COVID-19 pandemic is a major threat, and it has severe impact on the health and life of many people worldwide. The recent advances in computer vision made possible by various computational method has paved the way for computer assisted diagnosis in fighting COVID-19. Early detection of the COVID-19 through accurate diagnosis, may decrease the patient’s mortality rate. Chest X-ray images are crucial and mostly used for the diagnosis of this disease. Thus, this study used optimized Convolution Neural Network (OCNN) to support the diagnosis of COVID-19 using chest x-ray. Particle Swarm Optimization (PSO) was applied to optimize the network of CNN for improved performance. The dataset used in this study was acquired from Kaggle repository. The dataset contains the Chest X-Ray images of COVID-19 patients and normal patients. The model is created, and the results have been evaluated by using the various evaluation metrics, i.e., sensitivity, false positive rate, precision, accuracy, and prediction time. The approach adopted in this study enhances CNN by making it free from iterative adjustment of weights which increases the computational speed to a higher extent. The experimental results reveal that the proposed technique achieved an improved performance which indicates the very high accuracy of the proposed model.