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 Oct 10;14(1):173-184.
doi: 10.1007/s13534-023-00330-7. eCollection 2024 Jan.

MS-CFNet: a multi-scale clinical studying-based and feature extraction-guided network for breast fibroadenoma segmentation in ultrasonography

Yongxin Guo  1   2 , Yufeng Zhou  1   2   3
Affiliations

MS-CFNet: a multi-scale clinical studying-based and feature extraction-guided network for breast fibroadenoma segmentation in ultrasonography

Yongxin Guo et al. Biomed Eng Lett. .

Abstract

Segmenting breast tumors in ultrasonography is challenging due to the low image quality and presence of artifacts. Radiologists' studying and diagnosis skills are integrated with artificial intelligence to establish a clinical learning-based deep learning network in order to robustly extract and delineate features of breast fibroadenoma. The spatial local feature contrast (SLFC) module captures overall tumor contours, while the channel recursive gated attention (CRGA) module enhances edge perception through high-dimensional information interaction. Additionally, full-scale feature fusion and enhanced deep supervision are applied to improve model stability and performance. To achieve smoother boundaries, we introduce a new loss function (cosh-smooth) that penalizes and finely tunes tumor edges. Our dataset comprises 1016 clinical ultrasound images of breast fibroadenoma with labeled masks, alongside a publicly available dataset of 246 ones. Segmentation performance is evaluated using the Dice similarity coefficient (DSC) and mean intersection over union (MIOU). Extensive experiments demonstrate that our proposed MS-CFNet outperforms state-of-the-art methods. Compared to TransUNet as a baseline model, MS-CFNet improves by 1.47% in DSC and 2.56% in MIOU. The promising result of MS-CFNet is attributed to the integration of radiologists' clinical diagnosis procedure and the bionic mindset, enhancing the network's ability to recognize and segment breast fibroadenomas effectively.

Keywords: Breast fibroadenomas; Multi-scale clinical studying-based network; Multi-scale feature extraction-guided network; Segmentation; Sonography.

PubMed Disclaimer

Conflict of interest statement

Conflict of interestThe authors have no conflict of interest to state.

Figures

Fig. 1
Fig. 1
a Representative ultrasound breast fibroadenoma images and b the corresponding ground truth of tumor boundary in both local UBF and public BUSI datasets
Fig. 2
Fig. 2
a The network architecture of MS-CFNet, b the schematic representation of patch partition and linear embedding, and c the architecture of transformer layer
Fig. 3
Fig. 3
Structure of the spatial local feature contrast (SLFC) module
Fig. 4
Fig. 4
Structure of the channel recursive gated attention (CRGA) module
Fig. 5
Fig. 5
The process of full-scale deep supervision at the second decoder layer
Fig. 6
Fig. 6
Visual comparison of the ground truth with segmentation results using the backbone and with the inclusion of SC, SC + SLFC, SC + SLFC + CRGA, SC + SLFC + CRGA + Deep for three representative ultrasound images. The white areas are the corresponding breast fibroadenoma segmentation results. The arrow highlights the region where lesion segmentation is challenging
Fig. 7
Fig. 7
Comparison of breast fibroadenoma segmentation results with different models, including ours, Swin-Unet, TransUNet, DeepLab V3 + , and U-net3 + for four representative cases. The arrow highlights the region where lesion segmentation is challenging
Fig. 8
Fig. 8
Statistical significance tests which performed based on the DSC metrics for each models compared with ours. Note: p-values were calculated: *p < 0.05; **p < 0.001
Fig. 9
Fig. 9
Statistical significance tests which performed based on the DSC metrics for each models compared with ours. Note: p-values were calculated: *p < 0.05; **p < 0.001
See this image and copyright information in PMC

Similar articles

See all similar articles

Cited by

References

    1. Cerrato F, Labow BI. Diagnosis and management of fibroadenomas in the adolescent breast. In: Seminars in plastic surgery. Thieme Medical Publishers, New York; 2013. p. 023–5. - PMC - PubMed
    1. Rodden AM. Common breast concerns. Primary Care: Clin Off Pract. 2009;36:103–113. doi: 10.1016/j.pop.2008.10.006. - DOI - PubMed
    1. Santen RJ, Mansel R. Benign breast disorders. N Engl J Med. 2005;353:275–285. doi: 10.1056/NEJMra035692. - DOI - PubMed
    1. Jayasinghe Y, Simmons PS. Fibroadenomas in adolescence. Curr Opin Obstet Gynecol. 2009;21:402–406. doi: 10.1097/GCO.0b013e32832fa06b. - DOI - PubMed
    1. Huang Q, Luo Y, Zhang Q. Breast ultrasound image segmentation: a survey. Int J Comput Assist Radiol Surg. 2017;12:493–507. doi: 10.1007/s11548-016-1513-1. - DOI - PubMed

LinkOut - more resources