Deep Features for COVID-19 Detection: Performance Evaluation on Multiple Classifiers

ABSTRACT

Humanity has faced the greatest difficulties in recent years in COVID-19. These diseases are caused by significant alveolar damage and progressive respiratory failure. To address this issue, healthcare facilities needed rapid testing methods to identify COVID-19 patients and treat them immediately. In this paper, we developed a rapid testing strategy using machine and deep learning architecture with three different categories of chest x-ray images, such as COVID-19, normal, and pneumonia, were considered to identify the COVID-19 affected images. It is very difficult to diagnose COVID-19 from the pool of chest x-ray images, as pneumonia and COVID-19 affected x-ray images closely resemble each other. For this issue, feature extraction plays an important role. Here we considered deep features which were extracted from deep learning models such as VGG19 and InceptionResnetV2. These deep features were classified using different machine learning algorithms. It was observed that 96.81% accuracy was obtained after classification using MLP. © 2022, The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd.