The purpose of this study was to distinguish the changes of MR signals during relaxation and contraction of muscles. First, MR images were acquired in relaxation and contraction states. The subject clasped his hands in relaxation state and unclasped in contraction state. Next, the images were segmented using mixture Gaussian distributions and EM algorithm. Finally, we evaluated statistical values gotten from mixture Gaussian distributions. As a result, mixing coefficients were different during relaxation and contraction. The experimental results indicated that the proposed analysis has the potential to discriminate between two states.