[1]丛培璐,张冲,贠恺,等.基于多尺度特征提取与多特征融合的甲状腺结节超声影像分割[J].中国医学物理学杂志,2026,43(4):473-479.[doi:DOI:10.3969/j.issn.1005-202X.2026.04.008]
　CONG Peilu,ZHANG Chong,YUN Kai,et al.Thyroid nodules segmentation in ultrasound image based on multi-scale feature extraction and multi-feature fusion[J].Chinese Journal of Medical Physics,2026,43(4):473-479.[doi:DOI:10.3969/j.issn.1005-202X.2026.04.008]
点击复制

基于多尺度特征提取与多特征融合的甲状腺结节超声影像分割()

《中国医学物理学杂志》[ISSN:1005-202X/CN:44-1351/R]

卷:
43卷
期数:
2026年第4期
页码:
473-479
栏目:
医学影像物理
出版日期:
2026-04-28

文章信息/Info

Title:
Thyroid nodules segmentation in ultrasound image based on multi-scale feature extraction and multi-feature fusion
文章编号:
1005-202X（2026）04-0473-07
作者:
丛培璐1; 张冲1; 贠恺1; 赵爽2; 赵文华1; 马志庆1
1.山东中医药大学医学信息工程学院， 山东 济南 250355； 2.山东中医药大学实验中心， 山东 济南 250355
Author(s):
CONG Peilu1;  ZHANG Chong1;  YUN Kai1;  ZHAO Shuang2;  ZHAO Wenhua1;  MA Zhiqing1
1. School of Medical Information Engineering, Shandong University of Traditional Chinese Medicine, Jinan 250355, China 2. Experimental Center, Shandong University of Traditional Chinese Medicine, Jinan 250355, China
关键词:
甲状腺结节; 超声; 图像分割; 多尺度特征提取; 注意力
Keywords:
Keywords: thyroid nodule ultrasound image segmentation multi-scale feature extraction attention
分类号:
R318；R445
DOI:
DOI:10.3969/j.issn.1005-202X.2026.04.008
文献标志码:
A
摘要:
以传统的U型网络为架构，提出一种基于多尺度特征提取与多特征融合的甲状腺结节超声影像分割方法。首先，设计一种基于多个小卷积核叠加的特征提取策略。通过堆叠多个小尺寸的卷积核，模型能够在不同的感受野下捕捉图像中的细节特征和全局特征，从而实现多尺度特征的高效提取。其次，通过混合注意力机制，包括通道注意力与空间注意力，将不同阶段的特征图进行融合，以强化原有的跳跃连接。本文算法在甲状腺结节分割数据集TN3K和DDTI的95%豪斯多夫距离（HD95）分别为16.02和17.86 mm，F1分数分别为82.21%和75.74%，在所有对比算法中表现最佳。实验结果表明，该方法可以为临床医生提供辅助诊断。
Abstract:
Abstract: Based on the classical U-Net, a novel method incorporating multi-scale feature extraction and multi-feature fusion for thyroid nodule segmentation in ultrasound image is proposed. Specifically, a feature extraction strategy based on stacked small-sized convolutional kernels is designed. By stacking multiple small-sized convolutional kernels, the model can capture both detailed and global features of images under different receptive fields, thereby achieving efficient multi-scale feature extraction. Through a hybrid attention mechanism which includes both channel and spatial attention, feature maps from different stages are effectively fused, thereby enhancing the original skip connections. The proposed algorithm achieves 95% Hausdorff distance (HD95) of 16.02 and 17.86 mm on the TN3K and DDTI thyroid nodule segmentation datasets, respectively, along with F1-scores of 82.21% and 75.74%, outperforming all other compared methods. Experimental results demonstrate that this approach can provide valuable assistance to clinicians in diagnostic practice.

相似文献/References:

[1]傅 静,James F Cnota.超声不同数学公式评估中国胎儿体重的准确性分析[J].中国医学物理学杂志,2014,31(01):4709.[doi:10.3969/j.issn.1005-202X.2014.01.025]
[2]裘品丹,翁纽周.超声显像在脐尿管囊肿病灶中的特点及其诊断价值研究[J].中国医学物理学杂志,2014,31(04):5064.[doi:10.3969/j.issn.1005-202X.2014.04.019]
[3]陆祎,朱景雨,汪登斌,等.双源CT双能非线性融合技术显示甲状腺疾病[J].中国医学物理学杂志,2016,33(10):1047.[doi:10.3969/j.issn.1005-202X.2016.10.014]
　[J].Chinese Journal of Medical Physics,2016,33(4):1047.[doi:10.3969/j.issn.1005-202X.2016.10.014]
[4]陶超,项四平. 基于光声-超声协同的多模成像系统[J].中国医学物理学杂志,2016,33(12):1240.[doi:10.3969/j.issn.1005-202X.2016.12.013]
　[J].Chinese Journal of Medical Physics,2016,33(4):1240.[doi:10.3969/j.issn.1005-202X.2016.12.013]
[5]涂昊.超声与核磁共振成像对早产儿脑室周围白质软化诊断价值的对比性分析[J].中国医学物理学杂志,2016,33(12):1276.[doi:10.3969/j.issn.1005-202X.2016.12.021]
　[J].Chinese Journal of Medical Physics,2016,33(4):1276.[doi:10.3969/j.issn.1005-202X.2016.12.021]
[6]孟爽,王辉,谢蓄芬,等.超像素有偏观测模糊聚类的乳腺超声图像分割[J].中国医学物理学杂志,2017,34(7):693.[doi:10.3969/j.issn.1005-202X.2017.07.009]
　[J].Chinese Journal of Medical Physics,2017,34(4):693.[doi:10.3969/j.issn.1005-202X.2017.07.009]
[7]李俏颖,周定中. 超声与MRI诊断乳腺癌及不同病理分型影像特点分析[J].中国医学物理学杂志,2019,36(2):205.[doi:DOI:10.3969/j.issn.1005-202X.2019.02.016]
　LI Qiaoying,ZHOU Dingzhong. Effectiveness of ultrasound and MRI in the diagnosis of breast cancers and analysis of imaging features of different pathological types[J].Chinese Journal of Medical Physics,2019,36(4):205.[doi:DOI:10.3969/j.issn.1005-202X.2019.02.016]
[8]任慧雅,高宏博,李月秋,等. 石墨烯热疗对良性甲状腺结节的疗效[J].中国医学物理学杂志,2019,36(4):487.[doi:DOI:10.3969/j.issn.1005-202X.2019.04.023]
　REN Huiya,GAO Hongbo,LI Yueqiu,et al. Therapeutic effects of graphene thermal therapy for benign thyroid nodules[J].Chinese Journal of Medical Physics,2019,36(4):487.[doi:DOI:10.3969/j.issn.1005-202X.2019.04.023]
[9]申严,杨普查,卓宋明.床旁超声联合支气管镜在被动卧床肺炎患者住院管理评估指标中的价值[J].中国医学物理学杂志,2019,36(8):933.[doi:DOI:10.3969/j.issn.1005-202X.2019.08.014]
　SHEN Yan,YANG Pucha,ZHUO Songming.Value of bedside ultrasound combined with bronchoscopy in the index evaluation of inpatient management for bedridden patients with pneumonia[J].Chinese Journal of Medical Physics,2019,36(4):933.[doi:DOI:10.3969/j.issn.1005-202X.2019.08.014]
[10]王家友,黄燕,伍小勇. 不同切面超声筛查在早中期孕妇孕检胎儿颜面部筛查中的应用[J].中国医学物理学杂志,2019,36(10):1195.[doi:DOI:10.3969/j.issn.1005-202X.2019.10.016]
　WANG Jiayou,HUANG Yan,WU Xiaoyong. Application of ultrasound screening of different sections in fetal facial screening during early and mid-pregnancy[J].Chinese Journal of Medical Physics,2019,36(4):1195.[doi:DOI:10.3969/j.issn.1005-202X.2019.10.016]
[11]叶冯颖,杨文敏,李尚青,等.计算机辅助诊断软件联合多学科建立甲状腺结节恶性风险预测模型[J].中国医学物理学杂志,2021,38(1):54.[doi:DOI:10.3969/j.issn.1005-202X.2021.01.010]
　YE Fengying,YANG Wenmin,LI Shangqing,et al.A predictive model of thyroid nodules: based on the CAD software and multi-disciplinary indicators[J].Chinese Journal of Medical Physics,2021,38(4):54.[doi:DOI:10.3969/j.issn.1005-202X.2021.01.010]

备注/Memo

备注/Memo:
【收稿日期】2025-12-16 【基金项目】山东省医药卫生科技项目（202425020411）；山东中医药大学科学研究基金（KYZK2024Q30） 【作者简介】丛培璐，硕士研究生，研究方向：医学图像处理与分析，E-mail: 15662643020@163.com 【通信作者】马志庆，教授，研究方向：医学图像处理与分析，E-mail: mazhq126@163.com
更新日期/Last Update: 2026-04-29