Fully automated segmentation of MRI brain image using mixture Gaussian model and three-dimensional morphological filter

The method to segment a human brain of a magnetic resonance image is proposed in which tissue characteristic and structure are considered using two algorithms, one is mixture Gaussian model and the other is three-dimensional morphological filter. First, the image is segmented into five clusters of background, cerebro-spinal fluid, gray matter, white matter and others by mixture Gaussian model applied to a histogram of voxel values. Next, a three-dimensional morphological filter is invoked to modify misclustered voxels. All of steps can be done in automatically. The brain structures are segmented well in derived results and the proposed algorithm has a potential possibility for clinical applications.