一种基于模糊核聚类的脑部磁共振图像分割算法

An algorithm for MRI brain image segmentation based on fuzzy kernel clustering

目的针对普通模糊核聚类算法(kernel fuzzy c-means clustering algorithm,KFCM)存在的随机选择初始聚类中心的问题,本文提出一种根据直方图得到确定的初始聚类中心的模糊核聚类算法,以更快速地分割脑部磁共振图像。方法首先利用区域生长法和形态学方法对原始脑部磁共振图像进行预处理,提取脑实质,然后计算出预处理图像的直方图,将直方图的4个峰值作为模糊核聚类的初始聚类中心,最后利用模糊核聚类算法对脑实质进行分割。结果本文算法能有效地提取出脑组织中的白质(white matter,WM)、灰质(grey matter,GM)和脑脊髓液(cerebral spinal fluid,CSF)。与普通模糊核聚类算法相比,该算法的目标函数能更快地达到平稳,从而缩短运行时间。结论本文算法与随机选择聚类中心的模糊核聚类算法相比,可减少迭代次数,更快地得到分割结果。

Objective To improve the random choice problem of the preliminary clustering centers in ordinary kernel fuzzy C-means clustering algorithm （ KFCM ） , this paper proposes a method which could get assured preliminary clustering centers according to histogram and segment MRI brain image quickly. Methods Firstly the original image was processed by using the region growing and the mathematical morphology techniques and brain parenchyma was extracted. Then the histogram of the pre-segmented image was calculated and the four recognized histogram peaks were chosen as the preliminary clustering centers of KFCM. Finally KFCM was applied to segment the brain parenchyma. Results The proposed method could abstract the white matter （WM）, gray matter （GM） and cerebral spinal fluid （CSF） from the brain image efficiently. The objective function of the proposed method tended to be steady more quickly than ordinary KFCM and the running time was shorter. Conclusions This proposed method can reduce the iteration numbers and quickly get segmentation results compared with the random choice centre of KFCM.