基于水平集的脑部MR图像混合分割算法

A Hybrid Segmentation Method for Brain MRI Based on Level Set

脑部组织的分割与提取是脑部医学图像三维重建及可视化能准确表达其相应组织部位的前提。针对脑部MRI图像数据的特点,本实验设计方法,一方面将Canny算子作用于原图像得到比原图更明确的边缘信息,并将其结果引入水平集函数中;另一方面利用阈值分割方法对原图进行分割,将其结果作为水平集方法中的初始水平集,避免了手工勾画轮廓的操作,有效地减少了由手工操作带来的分割误差。本方法将Canny算子精确定位边缘的优点与水平集追踪物体拓扑结构改变的特点相结合,结果表明,3种算法的有效结合可得到很好的目标分割结果。

Segmenting and extracting brain tissues from medical images are premises of 3D reconstruction and visualization which accurately express the corresponding tissues.Considering the features of brain MRI,the hybrid Segmentation Method used Canny operator acting on the original image,which got the more clear edge information than the original image,and then imported the result into the leve lset function.the method segmented the original images using the threshold segmentation method,and took the segmentation results as the initial level set in the level set method,which avoided the operation of hand-sketched contours and effectively reduced the manual segmentation error.the Method combined the advantages of Canny operator which could orient accurately the edge and the characteristics of Level set method which could track the object topology changes,the experiments shows that the effective combination of the three algorithms can get a good target segmentation results.