The present study investigated whether different types of motor imageries can be classified based on the location of the activation peaks or the multivariate pattern analysis (MVPA) of functional magnetic resonance imaging (fMRI) and compared the difference between visual motor imagery (VI) and kinesthetic motor imagery (KI). During fMRI scanning sessions, 25 participants imagined four movements included in the Motor Imagery Questionnaire-Revised (MIQ-R): knee lift, jump, arm movement, and waist bend. ⋯ Our results show that the imagined movements can be classified using both the location of the activation peak and the spatial activation patterns within the sensorimotor cortex, and MVPA performs better than the activation peak classification. Furthermore, our result reveals that the KI group achieved a higher MVPA decoding accuracy within the left primary somatosensory cortex than the VI group, suggesting that the modality of motor imagery differently affects the classification performance in distinct brain regions.