Deep Dual Attention Network for Precise Diagnosis of COVID-19 From Chest CT Images

Automatic diagnosis of Coronavirus disease 2019 (COVID-19) using chest computed tomography (CT) images is of great significance for preventing its spread. However, it is difficult to precisely identify COVID-19 due to the following problems: 1) the location and size of lesions can vary greatly in CT images; 2) its unique characteristics are often imperceptible in imaging findings. To solve these problems, a Deep Dual Attention Network ($\text {D}^{\text {2}}\text {ANet}$) is proposed for accurate diagnosis of COVID-19 by integrating dual attention modules (DAMs) with different scales of the feature extractor, where DAM can adaptively detect relevant lesion regions to extract discriminative imaging features of COVID-19. Specifically, DAM is implemented by two parallel blocks: 1) global attention block (GAB) and 2) local attention block (LAB), in which GAB is designed to roughly locate the infected regions from the entire image by modeling global contexts while LAB is developed to explicitly highlight subtle differences of COVID-19 from other viral pneumonia in the infected regions by learning detailed lesion information. Experimental results on several public datasets show that $\text {D}^{\text {2}}\text {ANet}$ outperforms the state-of-the-art approaches in various performance metrics.

I. Introduction

Coronavirus disease 2019 (COVID-19) is an ongoing global pandemic affecting 221 countries and territories. According to World Health Organization reports [1], as of August 11, 2021, there have been 203.9 million confirmed COVID-19 cases with more than 4.31 million deaths around the world. To date, the number of COVID-19 cases worldwide is dramatically increasing every day. One of the key steps to fight against COVID-19 is to timely detect the infected patients. Clinically, there are mainly two ways for this task: one is reverse-transcription polymerase chain reaction (RT-PCR) testing, and the other is chest computed tomography (CT), where the latter shows higher sensitivity to COVID-19 and can be used as an effective screening or diagnostic tool in epidemic areas [2], [3], [4]. Unfortunately, manual diagnosis using CT images requires many radiologists to read the subtle imaging features of COVID-19, which is extremely labor-intensive and subjective [5].