Tissue segmentation of brain MRI
Pavel Dvořák, Karel Bartusek, Jan Mikulka · 2015
This work focuses on segmentation of magnetic resonance images of brain. The segmentation is based on assumption that in magnetic resonance images with high signal-to-noise ratio, the noise can be approximated by Gaussian. The method is tested on stand-alone simulated 2D MR images of healthy brain. The comparison between T1-weighted, T2-weighted and multi-parametric images is performed. The proposed algorithm is used to segment brain images into three different tissues. For the proposed method, the best results were achieved for stand-alone T1-weighted images, while stand-alone T2-weighted images show the worst results. The achieved results slightly vary for particular tissue.