JoWUA

Abstract

Fighting the outbreak of COVID-19 is now one of humanity's most critical matters. Rapid detection and isolation of infected people are crucial for decelerating the disease's spread. Due to the pandemic, the conventional technique for COVID-19 detection, reverse transcription-polymerase chain reaction, is time-consuming and in small abundance. Therefore, studies have been searching for alternate methods for detecting COVID-19, and thus applying deep learning methods to patients' chest images has been rendering impressive performance. The primary objective of this study is to suggest a technique for COVID-19 detection in chest images that is both efficient and reliable. We propose a deep learning method for COVID-19 classification based on a modified UNet called (Covid-MUNet). The Covid-MUNet model is trained using publicly available datasets, including chest X-ray images for multi-class classification (3-class and 4-classes) and CT scans images for binary/multi-class classification (2-classes and 3-classes). Using chest images, the Covid-MUNet is a successful methodology that helps physicians rapidly identify patients with COVID-19, thereby delaying the fast spread of COVID-19. The proposed model achieved an overall accuracy of 97.44% in classifying three categories (COVID-19, Normal, and Pneumonia) and an accuracy of 96.57% in classifying two categories (COVID-19 and Normal).