跳到正文
原文
Nature 健康护理主题·· 8 天前AI 评分30

DSCA-Net 网络用于超声子宫肌瘤分割

A deep semantic and channel-spatial attention network for ultrasound uterine fibroid segmentation

AI 导读

DSCA-Net 是一种用于超声子宫肌瘤分割的深度语义和通道空间注意力网络,通过引入深层语义注意力门模块、通道空间注意力精炼模块和残差特征精炼模块,提高了分割精度。

正文

We’re sharing this article early to provide faster access to peer-reviewed, accepted research. It is citable and carries a permanent DOI. This version is subject to further edits and will be replaced automatically by the final Version of Record. All legal disclaimers apply.

Abstract

Accurate segmentation of uterine fibroids in ultrasound images is of great importance for computer-aided diagnosis and clinical assessment. However, this task remains difficult because ultrasound images often suffer from speckle noise, weak tissue contrast, indistinct lesion boundaries, and considerable variations in fibroid morphology. To improve segmentation accuracy under these challenging conditions, this study proposes DSCA-Net, a deep semantic and channel-spatial attention network for uterine fibroid segmentation. Specifically, the proposed network follows an encoder-decoder framework and incorporates several dedicated feature enhancement components. Firstly, a deep semantic attention gate module (DSAGM) is introduced into the deeper encoding stages to strengthen high-level semantic representation and guide the network to focus on fibroid-related regions. Secondly, a channel-spatial attention refinement module (CSARM) is embedded in the skip pathways to selectively refine encoder features before feature fusion, thereby reducing the interference of irrelevant background information. In the decoder, a residual feature refinement module (RFRM) is designed to enhance the recovery of spatial details and improve the fusion of multi-level features. In addition, a weighted efficient channel attention module (WECAM) is applied near the prediction layer to further recalibrate informative channel responses and stabilize feature activation. Experimental results on the uterine fibroid ultrasound dataset collected from Quzhou Hospital of Traditional Chinese Medicine, the public brain tumor dataset, and the digital database thyroid image (DDTI) dataset demonstrate the effectiveness and generalization capability of DSCA-Net across different medical imaging tasks. On the uterine fibroid ultrasound dataset, DSCA-Net achieved Dice, Mcc, and Jaccard scores of 0.9174, 0.9140, and 0.8479. On the public brain tumor dataset, the corresponding scores were 0.8473, 0.8464, and 0.7509, while on the DDTI dataset, DSCA-Net obtained Dice, Mcc, and Jaccard values of 0.7767, 0.7417, and 0.6371. In addition, comprehensive ablation experiments were conducted to verify the effectiveness of each key component in the proposed network. Computational efficiency was also evaluated by comparing model parameter counts and inference times with those of representative segmentation methods. DSCA-Net contains 15.07 million parameters and requires 163 ms per inference step under the reported experimental setting. These results demonstrate the segmentation performance of DSCA-Net across three medical imaging datasets and provide a quantitative assessment of its computational cost.

Similar content being viewed by others

Funding

This work was supported by the National Natural Science Foundation of China (No. 62102227), Joint Fund of Zhejiang Provincial Natural Science Foundation of China (No. ZCLTGS24E0601), Science and Technology Major Projects of Quzhou (2025K076).

Author information

Authors and Affiliations

  1. College of Mechanical Engineering, Quzhou University, Quzhou, 324000, China

    Qisong Zhou & Xiaokang Ding

  2. Department of Obstetrics and Gynecology, Quzhou Hospital of Traditional Chinese Medicine, Quzhou TCM Hospital at the Junction of Four Provinces Affiliated to Zhejiang Chinese Medical University, 324000, Quzhou, China

    Yun Jiang

Authors

  1. Qisong Zhou
  2. Xiaokang Ding
  3. Yun Jiang

Corresponding authors

Correspondence to Xiaokang Ding or Yun Jiang.

Ethics declarations

Competing interests

The authors declare no competing interests.

Additional information

Publisher's note

Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Rights and permissions

Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/.

Reprints and permissions

About this article

Check for updates. Verify currency and authenticity via CrossMark

Cite this article

Zhou, Q., Ding, X. & Jiang, Y. A deep semantic and channel-spatial attention network for ultrasound uterine fibroid segmentation. Sci Rep (2026). https://doi.org/10.1038/s41598-026-79906-z

Download citation

  • Received: 24 July 2026

  • Accepted: 04 November 2026

  • Published: 12 November 2026

  • DOI: https://doi.org/10.1038/s41598-026-79906-z

Keywords

来源:Nature 健康护理主题 · nature.com