Diagnose COVID-19 by using hybrid CNN-RNN for chest X-ray

Abstract

Combating the COVID-19 epidemic has emerged as one of the most promising healthcare the world's challenges have ever seen. COVID-19 cases must be accurately and quickly diagnosed to receive proper medical treatment and limit the pandemic. Imaging approaches for chest radiography have been proven in order to be more successful in detecting coronavirus than the (RT-PCR) approach. Transfer knowledge is more suited to categorize patterns in medical pictures since the number of available medical images is limited. This paper illustrates a convolutional neural network (CNN) and recurrent neural network (RNN) hybrid architecture for the diagnosis of COVID-19 from chest X-rays. The deep transfer methods used were VGG19, DenseNet121, InceptionV3, and Inception-ResNetV2. RNN was used to classify data after extracting complicated characteristics from them using CNN. The VGG19-RNN design had the greatest accuracy of all of the networks with 97.8% accuracy. Gradient-weighted the class activation mapping (Grad-CAM) method was then used to show the decision-making areas of pictures that are distinctive to each class. In comparison to other current systems, the system produced promising findings, and it may be confirmed as additional samples become available in the future. For medical personnel, the examination revealed an excellent alternative way of diagnosing COVID-19.