SCOAT-Net: A novel network for segmenting COVID-19 lung opacification from CT images

Shixuan Zhao,Zhidan Li,Yang Chen,Wei Zhao,Xingzhi Xie,Jun Liu,Di Zhao,Yongjie Li

DOI: https://doi.org/10.1016/j.patcog.2021.108109

IF: 8

2021-11-01

Pattern Recognition

Abstract:<p>Automatic segmentation of lung opacification from computed tomography (CT) images shows excellent potential for quickly and accurately quantifying the infection of Coronavirus disease 2019 (COVID-19) and judging the disease development and treatment response. However, some challenges still exist, including the complexity and variability features of the opacity regions, the small difference between the infected and healthy tissues, and the noise of CT images. Due to limited medical resources, it is impractical to obtain a large amount of data in a short time, which further hinders the training of deep learning models. To answer these challenges, we proposed a novel spatial- and channel-wise coarse-to-fine attention network (SCOAT-Net), inspired by the biological vision mechanism, for the segmentation of COVID-19 lung opacification from CT images. With the UNet++ as basic structure, our SCOAT-Net introduces the specially designed spatial-wise and channel-wise attention modules, which serve to collaboratively boost the attention learning of the network and extract the efficient features of the infected opacification regions at the pixel and channel levels. Experiments show that our proposed SCOAT-Net achieves better results compared to several state-of-the-art image segmentation networks and has acceptable generalization ability.</p>

computer science, artificial intelligence,engineering, electrical & electronic

What problem does this paper attempt to address?

The problem that this paper attempts to solve is: automatically segmenting the lung opaque areas caused by Coronavirus Disease 2019 (COVID - 19) from CT images. The main challenges faced by this task include:

1. **Difficulty in data acquisition**: Due to limited medical resources, it is difficult to obtain a large amount of data with accurate labels in a short time to train deep - learning models.
2. **Complexity and variability of the infected area**: The infected areas in CT slices show different sizes, positions, and texture features, and there are no obvious boundaries, which increases the difficulty of segmentation.
3. **Similarity to healthy tissues**: Due to the complexity of medical images, the lung opaque areas are very similar to other lung tissue structures and are difficult to distinguish.

To address these challenges, the authors propose a novel Spatial and Channel Coarse - to - Fine Attention Network (SCOAT - Net). Based on UNet++, this network introduces specially - designed spatial attention modules and channel attention modules to collaboratively enhance the network's attention - learning ability and extract effective features of the infected opaque areas at the pixel and channel levels. Experimental results show that the proposed SCOAT - Net achieves better results compared to other state - of - the - art image - segmentation networks and has good generalization ability.