Improvement of segmentation in sub-cortical area in MR images of the brain

Abstract

In this study, a method that improves the segmentation quality in the subcortical area of the brain using Bayesian approach is proposed. The tissue probability density functions and volume ratios which are required during segmentation are calculated by using only the subcortical area while taking a single atlas as the reference, and the contribution of this prior information on segmentation is shown. As a difference from classical Bayesian approaches, the proposed method is applied only in the segmentation of subcortical region, while cortical segmentation is performed exclusively by another method whose success is widely accepted. Then these results are merged generating a hybrid solution.