A validation framework for MR image segmentation
Xia Zheng, Mo Dai, Mingquan Zhou · 2008
A validation framework for MR image segmentation is proposed in this paper. It includes three stages: intensity inhomogeneity (IIH) correction, noise suppression without blurring structures and tissue classification. Based on MR brain images, in the first stage, an improved process is used to implement IIH correction. Subsequently, a new enhancement method on moments for noise removal and edge sharpening is introduced in the second stage. It owes much to properties of Gauss-Hermite moments (GHMs). In the third stage, FCM is used to classify two different tissues: white matter (WM) and gray matter (GM). For cerebrospinal fluid (CSF), it comes from subtraction result between T1 and T2 weighted images. Examples on simulated images have been reported to show the efficiency of this framework.