Chemical shift imaging (CSI), a method which samples 1H NMR-spectra from a grid of volume elements, produces an overwhelming amount of data. Each spectrum contains information about several metabolites in the sampled area. One approach for interpretation of this large amount of data is segmentation of the CSI grid in clusters which share the same features, followed by classification of each segment to a specific tissue type. We used mixture modeling to perform segmentation of CSI images for automatic identification of malignant areas in the human brain.