融合多尺度空间特征的甲状腺结节超声图像分割  

Ultrasound Image Segmentation of Thyroid Nodules by Fusing Multi-scale Spatial Features

作　　者：崔少国 张宇楠 CUI Shao-guo;ZHANG Yu-nan(School of Computer and Information Sciences,Chongqing Normal University,Chongqing 401331,China)

机构地区：[1]重庆师范大学计算机与信息科学学院,重庆401331

出　　处：《计算机与现代化》2024年第3期72-77,共6页Computer and Modernization

基　　金：国家自然科学基金资助项目(62003065);重庆市科技局自然基金资助项目(2022NSCQ-MSX2933,2022TFII-OFX0262,cstc2019jscx-mbdxX0061);教育部人文社科规划基金资助项目(22YJA870005);重庆市教委重点项目(KJZD-K202200510);重庆市社会科学规划项目(2022NDYB119);重庆师范大学人才基金资助项目(20XLB004)。

摘　　要：甲状腺结节超声图像噪声严重,不同组织间对比度低,在现有的甲状腺结节超声图像分割算法中存在边缘信息模糊和小结节分割不准确的问题,因此本文提出一种融合多尺度空间特征的甲状腺结节超声图像分割算法。该算法以U-Net模型为基础,在编码部分,引入坐标注意力机制,将位置信息嵌入到通道注意力中实现模型对甲状腺结节区域的定位,同时融合多尺度特征模块提取空间特征,在下采样过程中使用卷积操作,保留更多的细节特征,并采用二值交叉熵损失和Dice系数损失作为综合损失。实验结果表明,本文算法模型相比基准模型U-Net而言,在F1评价指标上提升了9.9个百分点,在精确率上提高至92.8%,从而验证本文所提方法的可行性与有效性。The ultrasound images of thyroid nodules have serious noise and low contrast between different tissues.The existing ul⁃trasound image segmentation algorithm of thyroid nodules have some problems of blurred edge information and inaccurate segmen⁃tation of small nodules.Therefore this paper proposes an ultrasound image segmentation algorithm of thyroid nodules fused with multi-scale spatial features.Based on the U-Net model,the coordinate attention mechanism is introduced to embed the position information into the channel attention to achieve the model’s localization of the thyroid nodule region in the coding part.At the same time,the fused multiscale feature module extracts the spatial aspect features.To retain more detailed features,we uses con⁃volution operation in the process of down sampling and the binary cross-entropy loss and Dice coefficient loss as the comprehen⁃sive loss.The experimental results show that compared with the benchmark model U-Net,the proposed algorithm model im⁃proves the F1 evaluation index by 9.9 percentage points,and the accuracy rate is increased to 92.8%.Thus the feasibility and ef⁃fectiveness is verified.

关 键 词：甲状腺结节 U-Net 空洞卷积 多尺度特征 坐标注意力 

分 类 号：TP391.9[自动化与计算机技术—计算机应用技术]