COVID-19 Lesion Segmentation and Classification of Lung CTs Using GMM-Based Hidden Markov Random Field and ResNet 18

Abstract

COVID-19 has been circulating around the world for over a year, causing a severe pandemic in every country, affecting billions of people. One of the most extensively utilized diagnostic methodologies for diagnosing and detecting the presence of the COVID-19 virus is reverse transcription-polymerase chain reaction (RT-PCR). Various ideas have been proposed for the detection of COVID-19 using medical imaging. CT or computed tomography is one of the beneficial technologies for diagnosing COVID-19 patients, the need for screening of positive patients is an essential task to prevent the spread of the disease. Segmentation of Lung CT is the initial step to segment the infection caused by the virus in the lungs and to analyze the lungs CT. This article introduces a novel Hidden Markov Random Field based on Gaussian Mix Model (GMM-HMRF) method ensembled with the modified ResNet18 deep architecture for binary classification. The proposed architecture performed well in terms of accuracy, sensitivity, and specificity and achieved 86.1%, 86.77%, and 85.45%, respectively.