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
. 2022 Jan 3:8:755309.
doi: 10.3389/fmed.2021.755309. eCollection 2021.

Automatic Segmentation of Novel Coronavirus Pneumonia Lesions in CT Images Utilizing Deep-Supervised Ensemble Learning Network

Yuanyuan Peng  1   2 , Zixu Zhang  1 , Hongbin Tu  1   3 , Xiong Li  4
Affiliations

Automatic Segmentation of Novel Coronavirus Pneumonia Lesions in CT Images Utilizing Deep-Supervised Ensemble Learning Network

Yuanyuan Peng et al. Front Med (Lausanne). .

Abstract

Background: The novel coronavirus disease 2019 (COVID-19) has been spread widely in the world, causing a huge threat to the living environment of people. Objective: Under CT imaging, the structure features of COVID-19 lesions are complicated and varied greatly in different cases. To accurately locate COVID-19 lesions and assist doctors to make the best diagnosis and treatment plan, a deep-supervised ensemble learning network is presented for COVID-19 lesion segmentation in CT images. Methods: Since a large number of COVID-19 CT images and the corresponding lesion annotations are difficult to obtain, a transfer learning strategy is employed to make up for the shortcoming and alleviate the overfitting problem. Based on the reality that traditional single deep learning framework is difficult to extract complicated and varied COVID-19 lesion features effectively that may cause some lesions to be undetected. To overcome the problem, a deep-supervised ensemble learning network is presented to combine with local and global features for COVID-19 lesion segmentation. Results: The performance of the proposed method was validated in experiments with a publicly available dataset. Compared with manual annotations, the proposed method acquired a high intersection over union (IoU) of 0.7279 and a low Hausdorff distance (H) of 92.4604. Conclusion: A deep-supervised ensemble learning network was presented for coronavirus pneumonia lesion segmentation in CT images. The effectiveness of the proposed method was verified by visual inspection and quantitative evaluation. Experimental results indicated that the proposed method has a good performance in COVID-19 lesion segmentation.

Keywords: COVID-19 lesion segmentation; deep learning; deep-supervised ensemble learning network; local and global features; transfer learning; under CT imaging.

PubMed Disclaimer

Conflict of interest statement

HT was employed by company Technique Center, Hunan Great Wall Technology Information Co. Ltd. The remaining authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Figures

Figure 1
Figure 1
An intersection over union (IoU) criterion.
Figure 2
Figure 2
Hausdorff distance.
Figure 3
Figure 3
A pipeline for COVID-19 lesion segmentation in CT images.
Figure 4
Figure 4
UNet model.
Figure 5
Figure 5
Pyramid attention network (PAN).
Figure 6
Figure 6
DeepLabv3+ model.
Figure 7
Figure 7
Feature pyramid network (FPN) model.
Figure 8
Figure 8
Weighting parameters optimization. (A) w1 = 0.0. (B) w1 = 0.1. (C) w1 = 0.2. (D) w1 = 0.3. (E) w1 = 0.4. (F) w1 =0.5. (G) w1 = 0.6. (H) w1 = 0.7.
Figure 9
Figure 9
Segmentation of COVID-19 lesions with different deep learning methods. (A) CT slice. (B) Anotation. (C) DeepLabV3+. (D) Unet. (E) PAN. (F) FPN. (G) Linknet. (H) MAnet. (I) PSPnet. (J) The proposed method.
Figure 10
Figure 10
COVID-19 lesion segmentation with different weighting parameters. (A) CT slice. (B) Anotation. (C) [0.2, 0.1, 0.5, 0.2]. (D) [0.1, 0, 0.9, 0]. (E) [0.4, 0, 0.3, 0, 0.3]. (F) [0.5, 0.5, 0, 0]. (G) [0.1, 0, 0, 0.9]. (H) [0, 0, 0.2, 0.8]. (I) [0, 0.7, 0.1, 0.2]. (J) [0.2, 0.1, 0.6, 0.1].
Figure 11
Figure 11
Hausdorff distance with different methods. (A) CT slice. (B) Anotation. (C) DeepLabV3+. (D) Unet. (E) PAN. (F) FPN. (G) Linknet. (H) MAnet. (I) PSPnet. (J) The proposed method.
See this image and copyright information in PMC

Similar articles

See all similar articles

Cited by

References

    1. Alon D, Paitan Y, Robinson E, Ganor N, Lipovetsky J, Yerushalmi R, et al. . Downregulation of CD45 Signaling in COVID-19 patients is reversed by C24D, a novel CD45 targeting peptide. Front Med. (2021) 8:1251. 10.3389/fmed.2021.675963 - DOI - PMC - PubMed
    1. Murphy A, Pinkerton LM, Bruckner E, Risser HJ. The impact of the novel coronavirus disease 2019 on therapy service delivery for children with disabilities. J Pediatr. (2021) 231:168–77. 10.1016/j.jpeds.2020.12.060 - DOI - PMC - PubMed
    1. Comunale BA, Engineer L, Jiang Y, Andrews JC, Liu Q, Ji L, et al. . Poliovirus vaccination induces a humoral immune response that cross reacts with SARS-CoV-2. Front Med. (2021) 8:1285. 10.3389/fmed.2021.710010 - DOI - PMC - PubMed
    1. Li L, Cao Y, Fan J, Li T, Lang J, Zhang H, et al. . Impact of COVID-19 pandemic on the clinical activities in obstetrics and gynecology: a national survey in China. Front Med. (2021) 8:1225. 10.3389/fmed.2021.633477 - DOI - PMC - PubMed
    1. Meng Z, Wang T, Chen L, Chen X, Li L, Qin X, et al. . The effect of recombinant human interferon alpha nasal drops to prevent COVID-19 pneumonia for medical staff in an epidemic area. Curr Top Med Chem. (2021) 21:920–7. 10.2174/1568026621666210429083050 - DOI - PubMed

LinkOut - more resources