A deep feature fusion model using transfer learning for effective detection of COVID-19 infected chest x-ray images

A globally affected pandemic SARS-CoV-2 created health emergencies the world all over. After originating in China in 2019, COVID-19 viruses obtained various mutations in the past two years and created severe impacts on the quality of human life with a high mortality rate. Early diagnosis and rapid isolation is the best solution to prevent the spread of the viruses. Among various diagnosis tests, a chest X-ray scan shows better results. Since manual analysis of X-ray images of the chest is burdensome, many deep learning-based methods have evolved in the past two years. However, the methods developed so far still need improvements for successful implementation in clinical settings. This piece of writing proposes an avant-garde deep feature fusion model for the effective detection of COVID-19. The experiments in this work are arranged in two stages. In the first phase, the images of chest X-ray are tried to classify using three robust deep learning networks such as ResNet-50,Inception-v3, and DenseNet. We have combined the features from the highly performed deep learning networks and designed a deep feature fusion model in the second stage. The promising results in a combined dataset (combination of two publically available data sets) of X-ray images of the chest show the competence of the schemed model. We have also used four various data augmentation methods to generate more COVID-19-infected samples in this work.