ABSTRACT
Covid-19 disease that directly affecting lungs is an acute disease caused death of many people around word, so the early detecting of it and asses the relative ratio of the lung infection is a vital need. In this work, Histogram based contrast adjustment was implemented to enhance four lung abnormal CT scan images to highlight the abnormal regions within the experimental images. Fuzzy c-mean algorithm then was applied to segment the images in order to detect and isolate the infected regions. Besides, several morphological operations were employed to extract the refined infected Covid-19 areas effectively with accuracy of 96%.
ABSTRACT
The World Health Organization (WHO) compiled this medical imaging reference guide in response to the emergence of the COVID-19 virus. The Beijing Country Office of the World Health Organization learned on December 31 that there was an epidemic of pneumonia patients in Wuhan, China. The causative agent of the pandemic was quickly identified as a novel coronavirus. In 2019, we should anticipate seeing an increase in the prevalence of coronavirus sickness, also known as the SARS-CoV-2 virus, and the SARS-CoV-1 virus. In order to determine the presence of this virus (COVID-19), we have created two models. Finally, the distorted part of the image was located. Some of the processes that we go through regularly have been the subject of our efforts to automate them. Using Resnet-18 models in combination with Deep Convolutional Neural Network (DenseNet-121 & Resnet-18) models, we were able to successfully detect COVID-19. The Densenet-121 model did well in its training and evaluation on a dataset of 1600 chest X-ray images. Over 2700 CXR pictures may be used for model training and evaluation with Resnet18. We have separated the data into groups according to the suggested models and found widely varying degrees of precision across the board. Data from both sources showed that Densenet-121 was the most reliable model.