北京生物医学工程

基于改进的CV-RSF模型的甲状腺结节超声图像自适应分割算法

Adaptive segmentation algorithm for thyroid nodules based on improved CV-RSF model

作者：邵蒙恩严加勇崔崤峣于振坤

单位：上海理工大学医疗器械与食品学院(上海200093) 上海健康医学院附属周浦医院(上海201318) 中科院苏州生物医学工程技术研究所(苏州215163) 南京同仁医院(南京211102)

分类号：R318.04

出版年·卷·期（页码）：2020·39·3（251-256）

摘要：

目的甲状腺结节超声图像精确的分割对甲状腺结节的良恶性诊断尤为重要。目前，对于甲状腺结节超声图像的分割，有学者提出利用主动轮廓模型分割算法，但是由于活动轮廓分割算法需要手动设置迭代次数，未实现模型的自适应性。因此，本文提出了一种基于改进的无边缘主动轮廓-局部区域可控的拟合(Chan-Vese-region scalable fitting，CV-RSF)模型的甲状腺结节超声图像自适应分割算法。一方面，避免手动设置模型的迭代次数，提高分割效率；另一方面，通过对模型的改进，提高分割的准确率。方法选取南京同仁医院12例患者的甲状腺结节超声图像用于实验。首先，在无边缘主动轮廓(Chan-Vese，CV)模型中，引入一个基于梯度的边缘引导函数，根据面积变化率，自适应地获取甲状腺结节的粗分割轮廓；然后，将粗分割轮廓作为局部区域可控的拟合(region-scalable fitting，RSF)模型的初始轮廓，并根据面积变化率，自适应地获取甲状腺结节最终分割结果。将改进模型分割的结果与CV模型、RSF模型分割的结果进行比较，并分析甲状腺结节边缘清晰度对分割结果的影响。结果本文模型算法分割结果的平均迭代次数、平均面积重叠率、平均Hausdorff分别达到了134、90.34%、9.77，均优于CV模型、RSF模型的分割算法。结论该算法有效地分割出边缘清晰和不清晰的甲状腺结节超声图像，并解决手动设置迭代次数的问题，从而实现甲状腺结节的有效、准确、自动分割。

Objective The accurate segmentation of thyroid nodule ultrasound images is particularly important for the diagnosis of benign and malignant thyroid nodules. Currently, the segmentation of thyroid nodule ultrasound images have been proposed by some scholars who are using an active contour model segmentation algorithm. However, the active contour segmentation algorithm requires manual setting of the number of iterations so that model's adaptability has not been achieved. Therefore, this paper proposes an adaptive segmentation algorithm for thyroid nodule ultrasound images based on an improved edgeless active contour-local region controllable fitting (Chan-Vese-Region scalable fitting (CV-RSF) model); On the one hand, it avoids manually setting the number of iterations of the model to improve the segmentation efficiency; On the other hand, it improves the accuracy of the segmentation by improving the model. Methods Ultrasound images of thyroid nodules from 12 patients in Nanjing Tongren Hospital are used for the experiment. Firstly, in a Chan-Vese (CV) model, a gradient-based edge guidance function is introduced to adaptively obtain the rough segmented contour of the thyroid nodule based on the area change rate; Then, the rough segmented contour is regarded as the initial contour of a region-scalable fitting (RSF) model. At last, the final segmentation results of thyroid nodules are adaptively obtained based on the area change rate. The results of the improved model segmentation are compared with the results of the CV model and RSF model segmentation, and the influence of thyroid nodule edge sharpness on segmentation results are analyzed. Results In this study, ultrasound images of thyroid nodules from 12 patients in Nanjing Tongren Hospital are used for the experiment. The proposed model is superior to the CV model and RSF model, which achieves an average number of iterations of 134，an average area overlap ratio of 90.34% and an average Hausdorff distance of 9.77．Conclusions This algorithm effectively segments thyroid nodule ultrasound images with clear and unclear edges. Meanwhile, it does not need to manually set the number of iterations to achieve effective, accurate and automatic segmentation of thyroid nodules.

参考文献：

[1] 王昕，李亮，才晓东.改进LIC模型的甲状腺结节超声图像分割算法[J].吉林大学学报(信息科学版)，2017,35(1): 68-75.

Wang X, Li L, Cai XD. Ultrasound image segmentation algorithm for thyroid nodules based on improved LIC model [J]. Journal of Jilin University (Information Science Edition), 2017,35 (1): 68-75.

[2] 龙亚超. 超声甲状腺结节自动分级相关技术研究[D].成都：电子科技大学，2017.

Long YC. Research on automatic stratification for thyroid nodules in ultrasound images [D]. Chengdu: University of Electronic Science and Technology of China, 2017.

[3]中华医学会内分泌学分会，中华医学会外科学分会，中国抗癌协会头颈肿瘤专业委员会，等.甲状腺结节和分化型甲状腺癌诊治指南（2012）[J]. 中国肿瘤临床,2012，39（17）:1249-1278.

[4] 史晓光.2017版甲状腺癌诊疗指南怎么说[N].健康报，2017-08-01(2).

[5] 王萍，刘健，岳文胜，等.2016版《甲状腺结节超声诊断指南》在甲状腺良恶性结节鉴别中的应用[J].实用医学影像杂志，2017,18(6): 466-468.

Wang P, Liu J, Yue WS, et al. The application of diagnosis guider of thyroid nodules in 2016 with ultrasound in the differentiation of benign and malignant thyroid nodules [J]. Journal of Practical Medical Imaging, 2017, 18 (6): 466 -468.

[6] J.H.Yoon，K.Han，E.K.Kim，et al.甲状腺小结节的诊断与管理:甲状腺结节6项指南的比较研究[J].国际医学放射学杂志，2017,(4):467-468.

J.H.Yoon，K.Han，E.K.Kim，et al. Diagnosis and management of small thyroid nodules: a comparative study with six guidelines for thyroid [J]. International Journal of Medical Radiology, 2017, (04): 467-468.

[7] Li C M，Xu C，Gui C and Fox M D. Distance regularized level set evolution and its application to image segmentation [J]. IEEE Trans. Image Process, 2010,19(12):3243–3254.

[8] Chan T，Vese L. Active contours without edges [J]. IEEE Transactions on Image Processing,2001,10(2)：266-277.

[9] Li C，Kao C Y，Gore J C，et al. Minimization of region-scalable fitting energy for image segmentation [J]. IEEE Transactions on Image Processing, 2008,17(10):1940-1949.

[10] 刘瑞娟，何传江，原野.融合局部和全局图像信息的活动轮廓模型[J].计算机辅助设计与图形学学报，2012,24(3)：364-371.

Liu RJ, He CJ, Yuan Y. Active contours driven by local and global image fitting energy [J]. Journal of Computer-Aided Design & Computer Graphics, 2012,24 (3): 364-371.

[11] 郑伟，张晶，李凯玄，郝冬梅.结合局部信息改进的C-V超声图像分割模型[J].光电工程，2015,42(08):66-72.

Zheng W, Zhang J, Li KX, et al. The improved C-V ultrasound image segmentation model of combining local information [J]. Opto-Electronic Engineering, 2015,42 (08): 66-72.

[12] 王相海，金弋博.高光谱海岸带区域分割的活动轮廓模型[J].中国图象图形学报，2013,18(8):1031-1037.

Wang XH, Jin YB. The active contour model for segmentation of coastal Hyperspectral remote sensing image [J]. Journal of Image and Graphics, 2013,18 (8): 1031-1037.

[13] 郑斌，徐峰，郭进祥，刘立波.基于Chan-Vese模型的甲状腺结节超声图像分割方法[J].电脑与电信，2018,6:4-5.

Zheng B, Xu F, Guo JX, et al. Thyroid nodule ultrasound image segmentation method based on Chan-Vese model [J]. Computer & Telecommunication, 2018,6: 4-5.

[14] 徐文杰.超声甲状腺结节图像分割算法研究[D].长春：长春工业大学，2017.

Xu WeJ. Ultrasonic thyroid nodule image segmentation algorithms [D].Changchun: Changchun University of Technology, 2017.

[15] 张晶.基于活动轮廓模型的甲状腺结节分割算法研究[D].保定：河北大学，2016.

Zhang J. Study of segmentation methods of thyroid nodules based on active contour model [D].Baoding: Hebei University, 2017.

[16] 赵杰，祁永梅，潘正勇.结合全局和双核局部拟合的活动轮廓分割模型[J].计算机应用，2013,33(04):1092-1095.

Zhao J, Qi YM, Pan ZY. Active contour segmentation model of combining global and dual-core local fitting energy [J]. Journal of Computer Applications, 2013,33 (04): 1092-1095.

[17] Shattuck D W，Sandorleahy S R， Schaper K A，et al. Magnetic resonance image tissue classification using a partial volume model [J]. Neuroimage, 2001, 13(5):856-876.

[18] Taha A，Hanbury A. Metrics for evaluating 3D medical image segmentation: analysis，selection，and tool [J]. BMC medical image, 201, 15( 1) : 29．

[19] Cardenes R, de Luis-Garcia R，Bach-Cuadra M．A multidimensional segmentation evaluation for medical image data [J]. Computer Methods and Programs in Biomedicine, 2009, 96 ( 2) : 108－124．

服务与反馈：

【文章下载】【加入收藏】

提示：您还未登录，请登录！点此登录