基于各向异性扩散活动轮廓模型的左心室MRI分割  

Left Ventricle MRI Segmentation Based on Anisotropic Diffusion Snake Model

作　　者：祖克举[1] 周昌雄[1] 张尤赛[1] 

机构地区：[1]江苏科技大学电子信息学院,江苏镇江212002

出　　处：《计算机测量与控制》2007年第3期339-341,共3页Computer Measurement &Control

基　　金：江苏省社会发展项目"医学体数据的三维可视化应用研究"(2005DX015J)

摘　　要：结合各向异性扩散算法与梯度矢量流活动轮廓模型,提出了基于各向异性扩散活动轮廓模型并应用于心脏核磁共振图像分割;模型采用各向异性扩散方程构造活动轮廓模型的外部能量函数,得到边界更加清晰的分段平滑图像,运用梯度矢量流将边缘图梯度散射到平坦区域,可以有效抑制噪声,同时保持了目标边界;对左心室核磁共振图像的分割实验表明,该模型可以克服噪声和伪影的干扰,与原梯度矢量流模型相比具有更高的精确性和可靠性,有利于实现自动分割。Integrating anisotropic diffusion and gradient vector flow （GVF）, an anisotropic diffusion active contour model is proposed, and is used to segment left ventricle MR image. In this model, external energy function is constructed by using anisotropic diffusion equation in order to obtain piecewise image with clearer edges. Using a computational diffusion process, gradient map is extended farther away from the edge and into homogeneous regions. The new model smoothes the constant regions while preserves the edges. This model is tested on left ventricle MR image and the results demonstrate that it can overcome noise and pseudo image, and has better accuracy and reliability than original GVF snake model for the segmentation of LV MR image, facilitating computer automated segmentation.

关 键 词：各向异性扩散 活动轮廓模型 梯度矢量流 MRI分割 

分 类 号：TP319.41[自动化与计算机技术—计算机软件与理论]