基于动态规划的IVUS图像序列分割  

Segmentation of IVUS image sequences based on dynamic programming

作　　者：王岭[1,2] 陈曦[2] 董峰[1] 杨国文[2] 王福[2] WANG Ling CHEN Xi DONG Feng YANG Guo-wen WANG Fu(School of Electrical Engineering and Automation, Tianjin University, Tianjin 300072, China Tianjin Navigational Instruments Institute, Tianjin 300456, China)

机构地区：[1]天津大学电气与自动化工程学院,天津300072 [2]天津航海仪器研究所,天津300451

出　　处：《天津理工大学学报》2016年第5期16-20,共5页Journal of Tianjin University of Technology

摘　　要：血管内超声(IVUS)诊断是临床上对冠状动脉粥样硬化的重要检测手段之一.由于超声图像感兴趣区域受斑点噪声影响较大,传统的分割方法很难得到令人满意的内外膜分割效果.首先对IVUS帧序列图像进行了时域降噪和空间滤波等预处理,然后基于优化的活动轮廓模型将图像的分割转化为目标函数的最优化求解问题,利用动态规划算法求得全局最优值.实验结果表明,该算法不仅准确性和可靠性较高,而且是一种全局收敛的最优算法,对IVUS序列图像处理的可重复性和鲁棒性较好.Intravascular ultrasound（IVUS）constitutes a valuable technique for the diagnosis of coronary atherosclerosis in clinic. We can＇t get a popular segmentation result through conventional means because IVUS frames contain noise and the actual boundaries of regions of interest are difficult to be identified. Firstly, we use a spatio-temporal filter to reduce the effect of speckles and make the image enhanced. Then translate the problem of image segmentation to the problem of the minimum resolution of energy function. Finally,we use dynamic programming to resolute it. The experimental results show that the method is simple, accurate,reproducible and robust for sequential IVUS frames, and the algorithm is a kind of global feasibility one.

关 键 词：血管内超声 活动轮廓模型 动态规划 图像分割 斑点噪声 

分 类 号：TP391[自动化与计算机技术—计算机应用技术] R816.2[自动化与计算机技术—计算机科学与技术]