DCNET-DCGAN: A Novel Deep Convolutional Neural Network for COVID-19 Classification

Abstract

Coronavirus referred to as COVID-19 affects innumerable lives, causing havoc on public health and global economy. The limitation of clinical expertise, medical tools, and testing kits increases the widespread of COVID -19 across the globe. An accurate identification process is necessary for early detection of COVID-19. Recent studies state that the images obtained from the chest X- Rays are highly consistent in diagnosing COVID-19 rather from RT-PCR (Reverse-Transcription-Polymerase-Chain Reaction). Developing an automated CXR image diagnosing method for the accurate prediction is the objective of the proposed model. This objective is achieved by developing a proposed model composed of Deep Convolutional Generative Adversarial Networks (DCGANs) and a Deep Convolutional Neural Network (DCNET) using four distinct datasets (COVID -19 X-ray, COVID-chest X-ray, COVID-19 Radiography, and Corona Hack-chest X-ray). The proposed model exploits the deep learning features of DCNET with four layers of convolution, three layers of max pooling and fully connected layers, thereby achieving a classification accuracy of 98.8% which is better than the pre-existing method. It classifies the result as Normal, COVID-19 and Pneumonia. This model will be an apt solution for facilitating faster screening process for affected patients.