Skip to main page content
U.S. flag

An official website of the United States government

Dot gov

The .gov means it’s official.
Federal government websites often end in .gov or .mil. Before sharing sensitive information, make sure you’re on a federal government site.

Https

The site is secure.
The https:// ensures that you are connecting to the official website and that any information you provide is encrypted and transmitted securely.

Access keys NCBI Homepage MyNCBI Homepage Main Content Main Navigation
. 2023 Sep:86:104939.
doi: 10.1016/j.bspc.2023.104939. Epub 2023 Apr 10.

Fully feature fusion based neural network for COVID-19 lesion segmentation in CT images

Wei Li  1 , Yangyong Cao  2 , Shanshan Wang  2 , Bolun Wan  2
Affiliations

Fully feature fusion based neural network for COVID-19 lesion segmentation in CT images

Wei Li et al. Biomed Signal Process Control. 2023 Sep.

Abstract

Coronavirus Disease 2019 (COVID-19) spreads around the world, seriously affecting people's health. Computed tomography (CT) images contain rich semantic information as an auxiliary diagnosis method. However, the automatic segmentation of COVID-19 lesions in CT images faces several challenges, including inconsistency in size and shape of the lesion, the high variability of the lesion, and the low contrast of pixel values between the lesion and normal tissue surrounding the lesion. Therefore, this paper proposes a Fully Feature Fusion Based Neural Network for COVID-19 Lesion Segmentation in CT Images (F3-Net). F3-Net uses an encoder-decoder architecture. In F3-Net, the Multiple Scale Module (MSM) can sense features of different scales, and Dense Path Module (DPM) is used to eliminate the semantic gap between features. The Attention Fusion Module (AFM) is the attention module, which can better fuse the multiple features. Furthermore, we proposed an improved loss function L o s s C o v i d - B C E that pays more attention to the lesions based on the prior knowledge of the distribution of COVID-19 lesions in the lungs. Finally, we verified the superior performance of F3-Net on a COVID-19 segmentation dataset, experiments demonstrate that the proposed model can segment COVID-19 lesions more accurately in CT images than benchmarks of state of the art.

Keywords: COVID-19; CT images; Deep network; Image segmentation; Multi-scale.

PubMed Disclaimer

Conflict of interest statement

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Figures

Fig. 1
Fig. 1
The architecture of F3-Net, which adopts encoder–decoder architecture and mainly includes four types of modules, namely Base Module (BM), Multiple Scale Module (MSM), Dense Path Module (DPM), Attention Fusion Module (AFM). The input data is the original CT images, and the output results are the lesion areas.
Fig. 2
Fig. 2
Base Module is utilized to extract features mainly through serial convolution, batch normalization and relu operations. Its input and output are features.
Fig. 3
Fig. 3
Multiple Scale Module is utilized to better deal with lesions at multiple scales. It mainly implements the extraction of features of different scales through two methods, dilated convolution and serial pooling, convolution, and up sampling. Different colored blocks Si in the figure represent features of different scales.
Fig. 4
Fig. 4
Dense Path Module is utilized to eliminate the semantic gap between encoder and decoder features. Its input is Mi and its output is Di. It contains several serial Dense Modules.
Fig. 5
Fig. 5
Attention Fusion Module contains an attention mechanism, which can make features more harmoniously merged. (a) is the application of AFM in F3-Net. (b) is an expanded form that can fuse multiple features.
Fig. 6
Fig. 6
Statistics of our dataset. (a)(b) The distribution map of lesion pixels in different dimensions. (c) The heat map of the pixels of the lesion. (d) The heat map of LossCovid−BCE.
Fig. 7
Fig. 7
Visual comparison of COVID-19 lesions segmentation results of different methods on our test set.
See this image and copyright information in PMC

Similar articles

See all similar articles

References

    1. Zu Zi Yue, Jiang Meng Di, Xu Peng Peng, Chen Wen, Ni Qian Qian, Lu Guang Ming, Zhang Long Jiang. Coronavirus disease 2019 (COVID-19): A perspective from China. Radiology. 2020;296(2):E15–E25. - PMC - PubMed
    1. Ai Tao, Yang Zhenlu, Hou Hongyan, Zhan Chenao, Chen Chong, Lv Wenzhi, Tao Qian, Sun Ziyong, Xia Liming. Correlation of chest CT and RT-PCR testing for coronavirus disease 2019 (COVID-19) in China: A report of 1014 cases. Radiology. 2020;296(2):E32–E40. - PMC - PubMed
    1. Wang Shuai, Kang Bo, Ma Jinlu, Zeng Xianjun, Xiao Mingming, Guo Jia, Cai Mengjiao, Yang Jingyi, Li Yaodong, Meng Xiangfei, et al. A deep learning algorithm using CT images to screen for corona virus disease (COVID-19) Eur. Radiol. 2021:1–9. - PMC - PubMed
    1. Wang Youxin, Hou Haifeng, Wang Wenrui, Wang Wei. Combination of CT and RT-PCR in the screening or diagnosis of COVID-19. J. Glob. Health. 2020;10(1) - PMC - PubMed
    1. Salehi Sana, Abedi Aidin, Balakrishnan Sudheer, Gholamrezanezhad Ali. Coronavirus disease 2019 (COVID-19): A systematic review of imaging findings in 919 patients. Am. J. Roentgenol. 2020;215(1):87–93. - PubMed