Detection and Classification of Covid-19 on X-Ray Images using Convolutional Neural Network

Abstract

The enduring COVID-19 outbreak has placed the world's healthcare systems on edge. To stop the spread of an epidemic or pandemic, prompt and accurate detection is always preferred. Medical imaging techniques have demonstrated excellent potential for more rapid and effective disease transmission control and containment. It is generally known that X-ray imaging and chest computed tomography (CT) are two efficient methods for diagnosing clinical COVID-19 disease. For effective diagnosis, assessment, and therapy, identifying COVID-19 in chest X-Ray (CXR) pictures is favored due to the quicker imaging time and significantly lower cost compared to CT. However, because COVID-19 and pneumonia share many similarities, CXR samples with deep features distributed close to category boundaries are easily misclassified by hyperplanes learnt from sparse training data. In the planned study, patients with COVID-19 and pneumonia will be reliably identified using X-rays, one of the medical imaging modalities used to assess the patient's inflammation. For the detected dataset, the appropriate deep convolutional neural network model is chosen. On the realworld dataset of chest X-ray images, the algorithm can identify COVID-19 patients and patients with pneumonia. For several categories including Normal, COVID-19, and Pneumonia, images are preprocessed and trained. By choosing the right features from photos in each dataset after preprocessing, the disease is detected. The outcome indicates that COVID-19 vs. Normal and COVID-19 vs. Pneumonia were accurately detected. Since ResNet-50 has performed exceptionally well in a number of medical imaging applications, it is used in this situation.