Non-iterative fast and efficient HOG-KELM hybrid model for COVID-19 diagnosis using Chest X-ray images

Abstract

COVID-19 is one of the most widely spread infectious disease, which was declared as pandemic in year 2019 by WHO. The viral infection can be diagnosed using RT-PCR test and through radiological scans, which are time consuming and not 100% accurate. Researchers are trying to develop more reliable, accurate and faster diagnostic system with the help of advance machine learning techniques. Some of the advance models developed using deep learning techniques are efficient but dependent on larger dataset and high-end processing system. In this paper, unique features extracted from COVID-19 chest X-ray images are used for diagnosis with the help of Histogram of Oriented Gradients-Kernel based Extreme Learning Machine (HOG-KELM) model. The proposed model is fast, efficient and economic, which does not rely on heavy machinery or processing unit. The experimental analysis is performed on a dataset of 15,153 CX-ray images, in which 3616 are COVID-19 infected, 10,192 normal lung, and 1345 pneumonia infected images. This model produces 99.78% accuracy for binary classification which evoke the efficacy of the proposed model compared to the existing state-of-art models.