Skull Stripping of MRI Using Clustering and 2D Region Growing Method

The segmentation of brain tissue finds an important role in many clinical applications. In this paper we propose a new skull stripping method for Magnetic Resonance Images of human head scan. We use clustering based technique to find the brain boundary inside the skull. First we find the clusters based on the intensity. Then the centroids of these clusters are found. These centroids are then connected, which form the brain boundary. We use 2D region growing method to connect all the clusters. By the same region growing method we remove the skull area and the clusters which are outside of the skull region. Finally we refine the edge of the brain by connecting the values of maximum intensity values.