Attention based Covid-19 Detection using Generative Adversarial Network

Abstract

The novel Coronavirus Disease 2019 (nCOVID-19) pandemic is a global health challenge, that requires collaborative efforts from multiple research communities. Effective screening of infected patients is a significant step in the fight against COVID-19, as radiological examination being an important screening methods. Early findings reveal that anomalies in chest X-rays of COVID-19 patients exist. As a result, a number of deep learning methods have been developed, and studies have shown that the accuracy of COVID-19 patient recognition using chest X-rays is very high. In this paper, we propose an attention based deep neural network for classifying the COVID-19 images, and extracting useful clinical information. Generative adversarial network is used to generate the synthetic COVID-19 images, as well as a good latent representation of both COVID-19 and normal images. Experiment results on public datasets shows the effectiveness of the proposed approach.