基于改进注意力W-Net的超声图像分割算法

Ultrasound Image Segmentation Algorithm Based on Improved Attention W-Net

摘要

针对超声成像过程中受噪声、伪像等因素干扰,导致图像中存在斑点噪声、区域模糊以及弱边界等纹理模糊不清,进而导致超声图像分割质量变差的问题,提出一种基于改进注意力W-Net的超声图像分割算法.首先,根据像素强度和边缘纹理的关联性构建纹理特征因子,应用纹理因子改进引导滤波算法,改进后的引导滤波算法能在平滑图像的同时,更好地保留图像中的纹理特征.其次,通过注意力机制将U-Net扩展为W-Net,在W-Net中融入改进的注意力机制,采用改进注意力W-Net对纹理保持后的超声图像进行分割,可在分割过程中更好地保持图像的纹理信息,从而提高分割的准确性和鲁棒性.实验结果表明,该算法的无参考图锐化因子(CPBD)值较高,面积差异百分数较低,超声图像分割效果好,可广泛应用于超声图像分割领域.

Abstract

Aiming at the problems of texture blurring such as speckle noise, region blurring, and weak boundaries in the image caused by factors such as noise and artifacts during ultrasound imaging, which led to a decrease in the quality of ultrasound image segmentation, we proposed an ultrasound image segmentation algorithm based on improved attention W-Net. Firstly, we constructed texture feature factors based on the correlation between pixel intensity and edge texture, and applied texture factors to improve the guided filtering algorithm. The improved guided filtering algorithm could smooth the image while better preserving the texture features in the image. Secondly, by extending U-Net to W-Net through attention mechanism, an improved attention mechanism was integrated into W-Net. The improved attention W-Net was used to segment ultrasound images with texture preservation, which could better preserve the texture information of the image during the segmentation process, thereby improving the accuracy and robustness of segmentation. Experimental results show that the proposed method has a relatively high cumulative probability of blur detection (CPBD) value, a low percentage of area difference, and good ultrasound image segmentation effect, which can be widely applied in the field of ultrasound image segmentation.

关键词

改进注意力W-Net / 超声图像 / 分割 / 纹理因子 / 改进引导滤波算法

Key words

improved attention W-Net / ultrasound image / division / texture factor / improved guidance filtering algorithm

参考文献

[1] 黄峥, 颜上取, 邹孝, 等. 基于超声图像语义分割的HIFU治疗目标区域提取方法[J]. 计算机辅助设计与图形学学报, 2022, 34(5):693-700.

[2] (Huang Z, Yan S Q, Zou X, et al. Extraction Method of Target Region for HIFU Therapy Based on Ultrasonic Image Semantic Segmentation[J]. Journal of Computer-Aided Design & Computer Graphics, 2022, 34(5):693-700.)

[3] 董志超, 温庆华, 代鹰, 等. 公转滑动轴承外圈划伤故障图像特征提取方法[J]. 煤炭工程, 2022, 54(增刊1):171-175.

[4] (Dong Z C, Wen Q H, Dai Y, et al. Image Feature Extraction Method for Scratch Fault of the Outer Ring of Revolution Sliding Bearing[J]. Coal Engineering, 2022, 54(Suppl 1):171-175.)

[5] 刘羽, 吴蓉蓉, 唐璐, 等. U-Net支气管超声弹性图像纵膈淋巴结分割[J]. 中国图象图形学报, 2022, 27(10):3082-3091.

[6] (Liu Y, Wu R R, Tang L, et al. U-Net-Based Mediastinal Lymph Node Segmentation Method in Bronchial Ultrasound Elastic Images[J]. Journal of Image and Graphics, 2022, 27(10):3082-3091.)

[7] 巫笠平, 段晓鹏, 马玉良, 等. 基于改进U-Net网络的多囊卵巢图像分割方法[J]. 中国生物医学工程学报, 2022, 41(6):663-671.

[8] (Wu L P, Duan X P, Ma Y L, et al. Image Segmentation of Polycystic Ovary Based on Improved U-Net Network[J]. Chinese Journal of Biomedical Engineering, 2022, 41(6):663-671.)

[9] 刘明坤, 张俊华, 李宗桂. 改进Mask R-CNN的甲状腺结节超声图像分割方法[J]. 计算机工程与应用, 2022, 58(16):219-225.

[10] (Liu M K, Zhang J H, Li Z G. Improved Mask R-CNN Method for Thyroid Nodules Segmentation in Ultrasound Images[J]. Computer Engineering and Applications, 2022, 58(16):219-225.)

[11] 崔文超, 徐德伟, 孙水发, 等. 基于边缘熵和局部FT分布的超声图像分割模型[J]. 图学学报, 2022, 43(2):263-272.

[12] (Cui W C, Xu D W, Sun S F, et al. Ultrasound Image Segmentation Model Based on Edge Entropy and Local FT Distribution[J]. Journal of Graphics, 2022, 43(2):263-272.)

[13] 石威, 李昕泽, 黄文昌, 等. 基于改进MultiResUNet网络的甲状腺超声图像分割[J]. 声学技术, 2022, 41(2):228-234.

[14] (Shi W, Li X Z, Huang W C, et al. Thyroid Ultrasound Image Segmentation Based on an Improved MultiResUNet Network[J]. Technical Acoustics, 2022, 41(2):228-234.)

[15] 周晓松, 赵涓涓. 多尺度监督U-Net甲状腺结节超声图像分割[J]. 太原理工大学学报, 2022, 53(6):1134-1142.

[16] (Zhou X S, Zhao J J. Multi-scale Supervised U-Net Ultrasound Image Segmentation of Thyroid Nodule[J]. Journal of Taiyuan University of Technology, 2022, 53(6):1134-1142.)

[17] 林天予, 宋亮, 高智凡, 等. 基于深度学习的二维心脏超声图像分割模型在小规模数据集上的性能评估[J]. 暨南大学学报(自然科学与医学版), 2022, 43(2):191-198.

[18] (Lin T Y, Song L, Gao Z F, et al. Evaluation of a Deep Learning-Based Model for 2-D Echocardiography Segmentation on Small Datasets[J]. Journal of Jinan University(Natural Science & Medicine Edition), 2022,43(2):191-198.)

基金资助

江西省自然科学基金(20212BAB206009)

江西中医药大学博士科研启动基金(2023BSZR007)