Abstract: Tissue segmentation of single and multi-spectral magnetic resonance (MR) images has been widely studied for the applications on normal aging brain, as well as on the diagnosis studies of Alzheimer's disease (AD), brain trauma and tumor in the recent years. But, the most of proposed methods in the published papers, the tissue segmentation was considered as problems of statistical decision, pattern classification, cluttering, image processing and analysis. The parameters used for tissue segmentation in those methods were the gray scalar/vector in single/multi-spectral images, which indirectly reflected the physical characteristics of the tissue. And those methods addressed the problems of tissue segmentation as the partitioning concourse of the components in a pixel in finite sets. So the results of the tissue segmentation obtained by conventional methods were unreasonable in some sense. This paper presents a new method of tissue segmentation based on the principle of spectroscopic decomposition of MR images, which consider the tissue segmentation as the problem of the estimation of MR physical parameters of the issues. This method can be used for suppressing not only fat MR signal in magnetic resonance imaging (MRI) but also water signal in magnetic resonance spectroscopy (MRS). Thus, this method is called the spatial and spectral MR imaging.