Classification of COVID-19 on Chest X-Ray Images Through the Fusion of HOG and LPQ Feature Sets

Hamaamin, Rebin Abdulkareem; Wady, Shakhawan Hares; Kareem, Ali Wahab

doi:10.24271/psr.2022.337896.1131

Classification of COVID-19 on Chest X-Ray Images Through the Fusion of HOG and LPQ Feature Sets

Document Type : Original Article

Authors

Rebin Abdulkareem Hamaamin ¹

Shakhawan Hares Wady ²

Ali Wahab Kareem ³

¹ Applied Computer / Medicals and Applied Sciences / Charmo University

² Applied Computer / Medicals and Applied Sciences / Charmo university

³ General Science / Education and Language / Charmo University

10.24271/psr.2022.337896.1131

Abstract

Covid-19 is a contagious disease that affects people's everyday life, personal health, as well as a nation's economy. COVID-19 infected individuals, according to a clinical study, are most usually contaminated with a severe condition after coming into a primary infection. The chest radiograph (also known as the chest X-ray or CXR) or a chest CT scan is a more reliable imaging method for diagnosing COVID-19 infected individuals. This article proposed a novel technique for classifying CXR scan images as healthy or affected COVID-19 by fusing the features extracted using Histogram of Oriented Gradient (HOG) and Local Phase Quantization (LPQ). This research is an experimental study that employed 7232 CXR images from a COVID-19 Radiography dataset as training and testing data. As a result, by using both individual and fused feature extraction methodologies, a developed model was created and fed into the machine learning techniques. The testing results reveal that the improved architecture outperforms current methods for identifying COVID-19 patients in terms of accuracy rate, which reached 97.15 %.

Keywords

COVID-19

CXR images

feature extraction

machine learning

Image classification