Hierarchical method for brain MRI segmentation based on using atlas information and least square support vector machine

In this paper, an automatic method for segmentation of cerebral magnetic resonance (MR) images based on using a hierarchical approach is proposed. In this study, a combination of brain probabilistic atlas as a priori information and support vector machines (SV M) is employed. Here, least-square SV M (LS-SV M) as a powerful supervised learning method with high generalization characteristics is used to generate brain tissue probabilities. The proposed method is applied to BrainW eb simulated data and IBSR real data. Quantitative and qualitative results obtained from simulations demonstrate excellent performance of the applied method in segmenting brain tissues into three categories of cerebrospinal fluid (CSF), white matter (W M) and grey matter (GM).