基于边缘流的改进梯度矢量流算法及其在淋巴结超声图像分割中的应用  

An Improved Gradient Vector Flow(GVF) Snake Model for Lymphatic Ultrasonic Image Segmentation Based on Edge Flow

作　　者：张俊华[1] 汪源源[1] 施心陵[2] 董怡[3] 王怡[3] 

机构地区：[1]复旦大学电子工程系,上海200433 [2]云南大学电子工程系,云南昆明650091 [3]复旦大学附属华山医院,上海200040

出　　处：《航天医学与医学工程》2006年第3期212-216,共5页Space Medicine & Medical Engineering

基　　金：国家重点基础研究规划基金(2005CB724303);上海市曙光计划(2003-901)

摘　　要：目的提出一种基于边缘流的梯度矢量流(gradientvectorflow,GVF)形变模型的图像分割方法,并用于淋巴结超声图像的分割。方法综合图像灰度和纹理特征构造边缘流,使每点的边缘流矢量指向最近的边缘,再由边缘流扩散得到GVF场作为形变模型的外部势力,引导模型形变实现图像分割。结果在给定4个标记点的条件下,实现了对淋巴结超声图像的半自动分割。结论将边缘流引入GVF将明显改善对低对比度超声图像的分割效果。Objective To put forward an improved gradient vector flow （GVF） snake model based on the edge flow for segmentation of medical ultrasonic images of the lymph node. Method An edge flow vector was constructed at each pixel location basing on the integration of image intensity and texture characteristics to make the edge flow vector point to the direction of potential boundarypixtels. The GVF field obtained by the diffusion of the edge flow vector was regarded as the exte（nal potential force of the snake model. Then the curve propagation was guided by the improved GVF snake model. Result A semi-automatic segmentation of ultrasonic images of the lymph node was accomplished under the condition that the four marks were provided by an expert. Conclusion By the integration of the edge flow into the GVF snake model, the segmentation quality is obviously improved for low contrast ultrasonic images.

关 键 词：梯度矢量流 边缘流 医学超声图像 图像分割 

分 类 号：TP391[自动化与计算机技术—计算机应用技术] R319[自动化与计算机技术—计算机科学与技术]