Abstract
Accumulating evidence shows that patients with schizophrenia have reduced cortical volume in several regions including anterior cingulate, medial and inferior prefrontal regions, temporal lobe, hippocampus, amygdala, thalamus and insula. Various researchers have made efforts to discriminate schizophrenia from controls using structural magnetic resonance imaging (MRI) . In early days researchers performed manual tracing of regions of interest and multivariate discrimination analysis. Recent mainstream is to use automated preprocessing such as voxel-based morphometry or cortical thickness analysis and extract features with machine learning technique. Many achieve 70~90% of accuracy, but clinical application is still on the way. Establishing a simple way for preprocessing, minimizing the difference across MRI scanners, and finding the optimal combination with some other tests are necessary for the breakthrough of clinical application.
