基于改进模糊C均值算法的颈动脉超声图像分割  

Segmentation of Carotid Intima Media in Ultrasound Images Using Improved Fuzzy C Means Algorithm

作　　者：李锵 张琦珺 关欣 滕建辅 Li Qiang;Zhang Qijun;Guan Xin;Teng Jianfu(School of Microelectronics,Tianjin University,Tianjin 300072,China)

机构地区：[1]天津大学微电子学院,天津300072

出　　处：《天津大学学报（自然科学与工程技术版）》2018年第1期95-102,共8页Journal of Tianjin University：Science and Technology

基　　金：国家自然科学基金资助项目(61471263)~~

摘　　要：颈动脉的内中膜厚度(IMT)是预测心血管疾病(CVDs)病发程度的重要指标.本文研究并提出一种自动、高效的计算机辅助IMT测量算法,该算法依据先验知识自动提取感兴趣区域(ROI),并采用基于隐马尔可夫随机场(HMRF)模型改进的模糊C均值(FCM)算法分割图像,实现IMT的自动测量.实验结果表明,所提算法对超声图像噪声的鲁棒性较强,IMT自动测量结果与真实值(GT)有很高的一致性:两个数据集合的相关系数为98.52%,平均绝对误差为0.022 0?0.016 4 mm.Common carotid artery intima-media thickness(IMT),which is considered as an important indicator of the development of cardiovascular diseases(CVDs),is usually measured on ultrasound images.This paper proposed an automatic and efficient computer-aided IMT measurement system.With the proposed method,region of interest(ROI)is extracted automatically based on prior knowledge,and then improved fuzzy C means(FCM)based on hidden Markov random field(HMRF)were applied to segment ROI and to detect the boundary of intima-media for IMT measurement.Experimental results show that the proposed method has strong robustness against ultrasound artifacts,and achieves similar clinical parameters to the ground truth(GT).The correlation coefficient between the two databases reaches98.52%,,and the mean of the absolute error between them is0.0220±0.0164mm.

关 键 词：超声图像分割 内中膜厚度测量 模糊C均值 隐马尔可夫随机场模型 感兴趣区域 

分 类 号：TP391.7[自动化与计算机技术—计算机应用技术] R445.1[自动化与计算机技术—计算机科学与技术]