Magnetic resonance imaging (MRI) techniques provide detailed anatomic information non-invasively and without the use of ionizing radiation. The development of new pulse sequences in MRI has allowed obtaining images with high clinical importance and thus joint analysis (multispectral MRI) is required for interpretation of these images. Fuzz rule-based systems can combine many inputs from widely varying sources so that they can be useful for description of tissues in MRI. In a fuzzy system an error free and optimized classifier can be obtained by genetic algorithms. In this paper, we have utilized a genetic fuzzy system for modeling different tissues in brain MRI and proposed a statistical pixel classification based on maximum likelihood (ML) and Bayesian classifiers as the final step of our segmentation process. Experiments were performed using simulated brain data (SBD) set. Provided numerical validation of the results demonstrate the strength of the proposed algorithm for medical image segmentation.