The major concern of classification methods in psychiatric imaging context is the separation of groups (e.g., patients and controls) based solely on structural properties of the brain. For this purpose, our group uses algorithms from the field of machine learning, such as support vector machines (SVM). These techniques estimate borderlines for multidimensional brain attributes in direct comparison of two or more groups. In many cases, predicting the group membership of further (previously unclassified) individuals can be accomplished with a high degree of accuracy. Validation of predictions with SVMs can be done using a leave-one-out cross-validation method. The mean of all resulting accuracies (each subject is left out once) is a measure of goodness of the prediction in this sample. Testing the ‘trained machine’ on new sample data is required for generalization on a broader population of subjects. One aim of this classification procedure is to support objective diagnoses of certain psychiatric and neurological diseases.