A Deep CNN Biomedical Imaging Technique for Detecting Infected Covid Patients

No Thumbnail Available



Journal Title

Journal ISSN

Volume Title


Department of Electronic and Telecommunication Engineering


The newly discovered coronavirus (COVID-19) sickness has caused havoc for people all over the globe and placed the whole planet on high alert. A second wave of the deadly illness has occurred because coronavirus infections have returned. This has been confirmed in most countries. The contagious virus may cause symptoms such as an itchy throat all the way up to pneumonia, and it has been responsible for the deaths of thousands of people while infecting millions more throughout the world. Getting a COVID-19 infection diagnosed as soon as possible is very important because it helps stop the illness from spreading and makes it easier to keep sick people apart and treat them. Recent studies in radiological imaging show that X-rays and CT scans of the lungs can show how COVID-19 infection will progress in people with severe symptoms. The main goal of this experiment is to quickly diagnose COVID-19 progression and other lung diseases by looking at X-rays of people with many symptoms. Using modern applications of artificial intelligence, a cutting-edge Deep CNN model that is both new and very good has been made to predict COVID-19 infections quickly and accurately. The model that has been made (healthy) divides X-ray pictures of the lungs into two groups: COVID and non-COVID. Accuracy, sensitivity, precision, and the f1 score are some of the assessment measures used to determine how well the suggested systems work. In the present study, a dataset consisting of X-ray samples was used. It was found that the CNN model had a high recognition rate of 99.38% and was in line with the current state of the art. The suggested model is very effective and accurate. It could help radiologists, and other medical professionals diagnose COVID-19 infection early in people who are showing signs of the disease. In addition, investigate various transfer learning approaches to contrast them with our model.


Submitted by Soumen Barua T181056