Programmable Detection of COVID-19 Infection Using Chest X-Ray Images Through Transfer Learning

Abstract

The novel Coronavirus (COVID-19), which has been designated a pandemic by the World Health Organization, has infected over 1 million individuals and killed many. COVID-19 infection may progress to pneumonia, which can be diagnosed via a chest X-ray. This research work proposes a novel technique for automatically detecting COVID-19 infection using chest X-rays. This research used 500 X-rays of patients diagnosed with coronavirus and 500 X-rays of healthy individuals to generate a data set. Due to the scarcity of publicly accessible pictures of COVID-19 patients, this research study has been attempted via the lens of knowledge transmission. Also, this research work integrates different convolutional neural network (CNN) architectures trained on Image Net to function as X-ray image feature extractors. After that, integrate CNN with well-established machine learning methods such as k Nearest Neighbor, Bayes, Random Forest, Multilayer Perceptron (MLP). The findings indicate that the most successful extractor-classifier combination for one of the data sets is the InceptionV3 architecture, which has an SVM classifier with a linear kernel that achieves an accuracy of 99.421 percent. Another benchmark, the best combination, is ResNet50 with MLP, which has 97.461%accuracy. As a result, the suggested technique demonstrates the efficacy of detecting COVID-19 using X-rays.