RAU: An Interpretable Automatic Infection Diagnosis of COVID-19 Pneumonia with Residual Attention U-Net

ABSTRACT

The novel coronavirus disease 2019 (COVID-19) has been spreading rapidly around the world and caused a significant impact on public health and economy. However, there is still lack of studies on effectively quantifying the different lung infection areas caused by COVID-19. As a basic but challenging task of the diagnostic framework, distinguish infection areas in computed tomography (CT) images and help radiologists to determine the severity of the infection rapidly. To this end, we proposed a novel deep learning algorithm for automated infection diagnosis of multiple COVID-19 Pneumonia. Specifically, we use the aggregated residual network to learn a robust and expressive feature representation and apply the soft attention mechanism to improve the capability of the model to distinguish a variety of symptoms of the COVID-19. With a public CT image dataset, the proposed method achieves 0.91 DSC which is 14.6% higher than selected baselines. Experimental results demonstrate the outstanding performance of our proposed model for the automated segmentation of COVID-19 Chest CT images. Our study provides a promising deep learning-based segmentation tool to lay a foundation to facilitate the quantitative diagnosis of COVID-19 lung infection in CT images. © 2021, Springer Nature Switzerland AG.