自适应滤波结合改进T-snake模型的甲状腺超声图自动分割方法  

Automatic segmentation of thyroid ultrasound images based on adaptive filtering combined with T-snake model

作　　者：周春瑜 程显毅 Zhou Chunyu;Cheng Xianyi(School of Electrical Engineering,Nantong University,Nantong Jiangsu 226006,China)

机构地区：[1]南通大学电气工程学院,江苏南通226006

出　　处：《计算机应用研究》2020年第3期944-946,共3页Application Research of Computers

基　　金：国家自然科学基金资助项目(61171132);江苏省自然科学基金资助项目(BK20151272);江苏省“333工程”资助项目(BRA2017475);江苏省“青蓝工程”资助项目。

摘　　要：提出一种基于T-snake模型的甲状腺超声波图像分割的新方法。首先,结合基于窗口的各向异性扩散滤波方法与自适应加权中值滤波算法有效地消除甲状腺超声波图像斑点噪声;其次,以传统T-snake模型为基础,增加自适应区域能量和膨胀力对非连续边界与弱边界进行有效提取,实现甲状腺超声波图像的自动分割;最后设定模型参数,使用临床数据进行实验。结果证明,应用该方法得到自动分割结果的平均相对差异度小于5%,平均相对重叠度大于91%,验证了其可行性。This paper proposed a new segmentation method of thyroid ultrasonic image based on T-snake model.Firstly,a window-based anisotropic diffusion filter combined with an adaptive weighted median filter,effectively eliminated speckle noise in thyroid ultrasonic images.Secondly,based on the traditional T-snake model and the adaptive region energy and expansion force,the method extracted the discontinuous boundary and weak boundary effectively,and realized the automatic segmentation of thyroid ultrasonic image.Finally,it set the model parameters and experimented with the clinical data.The results show that the average relative difference is less than 5%and the average relative overlap is more than 91%.It verifies the feasibility of the proposed method.

关 键 词：甲状腺超声波图像 图像分割 T-snake 自适应 加权中值滤波 各向异性扩散滤波 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]