基于模糊均值聚类算法的灰度不均匀脑MR图像的分割

Segmentation of brain MR images with inhomogeneity based on a modified fuzzy C-mean clustering algorithm

磁共振图像经常被含有缓慢变化的灰度不均匀场所破坏,不均匀场会造成同一组织的灰度发生变化,从而影响计算机辅助诊断的准确性。传统的基于灰度信息的分割方法对具有不均匀场的磁共振图像分割效果往往并不理想。文章改进了基于灰度信息的模糊C均值(FCM)算法,将偏移场模型、代表图像空间信息的邻域控制信息和最小二乘曲面拟合方法有机结合,能同时实现图像的校正和聚类,适用于灰度不均匀脑部磁共振图像的分割,分割精度明显优于已有的基于FCM的分割方法。

Magnetic resonance （MR） images are often corrupted by slowly intensity inhomogeneity for the same tissue over the image domain. The conventional segmentation method for the corrupted MR images is based on the gray information of images and these results are not satisfying. This paper presents a modified fuzzy c-mean （FCM） clustering based segmentation algorithm. It simultaneously carries out the intensity inhomogenity correction while segmenting the images by incorporated the bias model, area control information, curve and surface fitting based on least-squares method. Compared to previous FCM methods, the proposed method is more suitable to segment the corrupted brain MRI data with inhomogeneity and is able to provide higher segmentation accuracy.