结合水平集和区域生长的脑MR图像分割

Brain MR image segmentation by unifying level set and region growth

本文提出了结合改进的水平集和区域生长方法实现脑MR图像分割,并根据不同组织成像特征和组织结构特点采用不同算法分割进行了探索。主要步骤:首先用改进的水平集算法实现图像中骨组织和脑脊液(CSF)的提取;然后,依据直方图确定脑灰质(GM)、脑白质(WM)的近似灰度值,自动定位种子点后进行区域生长,实现脑灰质和脑白质的分离。实验结果表明,该方法充分利用了脑MR图像中的区域信息和边界信息,与传统单一算法分割脑MR图像相比,具有更强的鲁棒性和准确性。

An MR image segmentation method is presented by unifying level set and region growth. It is discussed that different arithmetic separate different tissues based on the imaging characters and tissue structures. First, separated the skull and Cerebrospinal （CSF） from the MR image using developed level set. Second, the approximate gray-value of gray matter （ GM ） and white matter （ WM ） were achieved using histogram. After the seeds were located automatically, region growth was used to separate white matter from gray matter. Experimental results indicated this approach made full use of the region information and boundary information of MR images. Compared with traditional methods using singular algorithm to separate a brain MR image, this method is characterized by robustness and accurateness.