Studies on brain tissue segmentation methods from MR images

Abstract

The medical image processing and analysis of various objects of newlineinterest are essential process used in most of the real world domains. The newlinebrain Magnetic Resonance (MR) image segmentation process is the complex newlineprocess, when availability of Intensity Non-Uniformity (INU). Though, newlineplenty of methods were developed to address the problem of accurate brain newlineMR image segmentation, still it is a challenging task to efficiently newlineapproximate the INU and enhance superior outcomes in segmentation newlineprocess. This thesis presents three different contributions for approximating newlineINU and brain tissue segmentation. First, a generalized rough set method newlinebased local intensity correction for image segmentation. The important tasks newlinein brain MR image segmentation are initialization of the rough fuzzy regions, newlineINU approximation and rectification, and tissue segmentation. Basically, newlinebrain MR image has formulated as linear form. In every local region, a local newlinelinear function is derived. The rough fuzzy regions are derived from lower newlineand upper approximations estimated based on rough set theory. The newlinecoefficients of linear equations are solved to approximate and correct the newlineINU. Further, this approach has been extended to approximate INU and brain newlinetissue segmentation simultaneously. newline newline