Magnetic resonance imaging (MRI) is a non-invasive technique which provides with stable and reliable information of brain structure. Although utilizing MRI as an objective biological marker adjunct to clinical information would be useful for better early detection and treatment of schizophrenia, it has not been applied to the clinical diagnosis because of a considerable between-group overlap in each anatomical variable. However a combination of brain anatomical variables or a pattern of disease-related anatomical changes would possibly differentiate patients with schizophrenia from healthy subjects or patients with other psychiatric disorders. In this article, we review several MRI-based classification studies that showed favorable classification accuracy using measures of multiple regions of interest, voxel-based morphometry, and automatically parcellated cortical regions of the entire cerebral cortex.