基于纹理分析的磁共振图象区域分割

本文介绍了一种用于医学磁共振脑部图象中的主要解剖结构区域分割的纹理分析方法. 利用纹理的二阶统计参数和局部分形维数,组成特征空间后,对象元进行两步分类:首先利用 K平均方法进行二值决策分类,然后采用概率松弛方法获得象元对各类的隶属概率.

Region Segmentation Based on Texture Analysis in MRI Data

In this paper, a texture analysis method for medical MRI region segmentation is presented, aimed at discrimination of macro anatomical structure in the cerebral image data. Four second-order statistical texture parameters, computed by using an isotropic gray-value concurrence, and the local fractal dimension compose a feature space, then a two-step classification procedure is applied to this space: a K-Means method is firstly used for obtaining binary-decision and follows a probabilistic relaxation for computing the membership probability of each voxel.