改进模糊C-均值分割算法在多发性硬化症MR脑部图像中的应用

背景：脑部MR图像是一种无纹理的图像，未被噪声污染的脑部MR图像的灰度值具有分片为常数的特点。因此，在聚类过程中灰度值有趋向于在同一个分割区域中相对接近的性质。目的：寻找一个能够自动分割多发性硬化症病灶的模糊C-均值改进方法，为临床对于多发性硬化症的判断提供更方便的工具。方法：考虑到脑部MR图像相邻象素属于同一分类的概率相近的特性，在迭代过程中对8邻域数据集进行滤波以降低噪声对聚类精度的影响，提出了一种改进的模糊C-均值聚类算法。就是将模糊C-均值聚类算法迭代过程中得到的灰度值看作一个数据集，用每个象素邻域象素的灰度值修正该象素的模糊隶属度取值，从而达到利用空间信息抑制噪声的目的。结果与结论：选取了10个多发性硬化症患者的脑部MRI图像进行试验。通过对多发性硬化症患者MR T1脑部图像和T2液体衰减反转回复脑部图像的分割实验，结果显示该算法能够有效分割多发性硬化症病灶，与其他方法所做的多发性硬化症病灶分割相比，本算法更易于实现，运算时间短，同时结果与临床医生的勾画比较重叠率较高，对其临床辅助诊断具有重要作用。

Segmentation of multiple sclerosis lesions in brain magnetic resonance images with modified fuzzy C-means algorithm

BACKGROUND: Brain magnetic resonance image is a non-texture image, characterized as piecewise constant for the gray value of MR images. Therefore, the gray value in clustering process has tended to relatively close in the same area.OBJECTIVE: To find a modified fuzzy C-means (FCM) algorithm method to segment the multiple sclerosis (MS) automatically that can support a tool to confirm MS easily.METHODS: A novel modified FCM framework is proposed by filtering membership data sets in the iterate process of FCM. The proposed algorithm denoise by making use of the property that the probability of the neighboring pixels which belong to the same cluster are similar.RESULTS AND CONCLUSION: We test our method on brain MR T1 and T2 fluid-attenuated inversion recovery images of 10 patients with MS. The testing experiments on brain MR images show that the proposed algorithm is able to segment the images correctly, which is important to assist the diagnosis of MS in clinic.