基于多尺度水平集的MR图像海马区分割方法

Hippocampus region segmentation method in MR images based on multi-scale level set

针对MR图像中海马区存在灰度不均匀性,基于区域动态轮廓的C-V模型只利用区域信息无法获得准确的海马区分割问题,结合多尺度边缘约束的演化思想和图像区域的全局信息,提出了一种结合边缘和区域信息的多尺度水平集MR(magnetic resonance)图像海马区分割方法。首先,在C-V模型的基础上采用内部约束能量项,消除水平集的重初始化,提高分割速度;其次,改进水平集函数中外部能量项的图像区域全局信息,解决由于灰度不均匀所引起的分割不准确问题;最后,在水平集函数的外部能量项中加入基于多尺度图像边缘的梯度信息,作为边缘约束停止项,使分割效果达到优化。实验结果表明,该算法对存在灰度不均匀性的图像海马区分割速度快、准确率高。

Aiming at the problems that intensity inhomogeneity exists in hippocampus region in magnetic resonance （MR） images, and the C-V model based on regional dynamic contour only uses region information and cannot obtain accurate hippocampus region segmentation result, a new hippocampus region segmentation method in MR images based on level set is proposed in this paper, which combines edge and region information. The method considers the evolution thought of multi-scale edge restraint and global information of image region. Firstly, on the basis of C-V model, the internal constrained energy item is used to eliminate the re-initialization of the level set to improve the segmentation speed. Secondly, the image region global information of the external energy item in the level set function is improved to solve the inaccurate segmentation problem caused by image intensity inhomogeneity. Finally, the gradient information based on the multi-scale image edge is added in the external energy item of the level set function as edge constraint stop item to optimize the segmentation result. Experimental results show that this method is fast and accurate in intensity inhomogeneity hippocampus region segmentation in MR images.