摘要
根据脑磁共振图像(magnetic resonance imaging,MRI)的灰度直方图中不同峰值对应不同的脑组织灰度分布的特点,提出了一种基于灰度直方图提取MR图像中脑组织的方法;首先,为了克服传统方法主观选择门限阈值等方面的不足,利用多项式曲线拟合灰度直方图获取区域生长门限阈值确定最优种子点,并结合形态学重构方法进行颅骨分离,对脑MR图像进行了预处理;然后,结合K均值聚类算法通过对灰度直方图多峰值的选取确定初始聚类中心,将颅骨剥离后的脑组织图像高效、精确地细化分割出脑脊液、灰质、白质;文中分别使用了模拟脑MRI数据以及真实脑部MRI数据进行测试,对分类结果进行定性、定量的分析,并与模糊C均值算法进行比较;结果表明,该方法提高了提取脑组织的工作效率和准确度。
This paper presents a tissue segmentation method based on the fact that the peak distribution of gray scale histogram in MR brain images corresponds to different brain tissue distribution. Firstly, in order to separate the skull, we propose a method which combines optimal threshold selection and adaptive regional growing algorithm. The optimal threshold is obtained by using polynomial curve fitting to histogram data, overcoming the drawbacks of subjective threshold selection associated with traditional methods. We then apply adaptive re gional growing algorithm to complete the separation of the skull. Secondly, the initial cluster centers are determined by selecting the peak val- ues of gray histogram, and then the K--means clustering algorithm is used to refine the segmentation of cerebrospinal fluid, gray matter, white matter. Finally, this method is tested on both simulated MRI data and human brain MRI images. We perform both qualitative and quantitative analyses in comparison with other image segmentation algorithms. Results show that the proposed algorithm can improve the ef ficiency and accuracy brain tissue segmentation.
出处
《计算机测量与控制》
2017年第11期170-173,178,共5页
Computer Measurement &Control
关键词
灰度直方图
曲线拟合
区域生长
K均值
脑组织提取
gray histogram
curve fitting
region growing
K-- means
brain tissue extraction