点分布模型约束的主动轮廓及其在脑MR图像分割中的应用

针对脑部MR图像结构复杂，使用传统的主动轮廓进行分割时，对初始化位置敏感，且易受非目标轮廓干扰，陷入局部极值等问题，提出了利用点分布模型进行形状约束的梯度矢量流主动轮廓模型。该方法在训练样本集的基础上，建立反映待分割目标轮廓先验形状信息的点分布模型，以模型的均值作为参考初始化主动轮廓，并使用模型提供的参数限制主动轮廓运动过程中的形变范围，有效地解决了使用传统主动轮廓对复杂图像往往不能收敛到期望形状的问题。此方法成功运用于脑MR图像中胼胝体和大脑灰质的分割，取得了满意的效果，验证了此方法的有效性和鲁棒性。

Brain MR Image Segmentation with PDM constrained Active Contour Model

Due to the complicated structures of brain MR images,problems were often encountered using traditional active contour models(ACM),such as sensitive to initialization,convergence to the local extremum subject to the allurement of non-target contour.Thus a new gradient vector flow(GVF) ACM with its shape constrained by the point distribution model(PDM) was proposed.In this method,a PDM was built based on the training set,which included the pre-known shape information about the target contour.The initial contour of ACM was estimated by the mean shape of PDM,and the deformation method of the ACM was also restrained by the parameter supplied by PDM.It solved the problem that traditional ACM could not converge to the expected shapes in the images with complicated structures.Satisfied results have been acquired while using this method on the segmentation of corpus callosum and gray matter of brain MR images,and its efficiency and robustness were also verified.