Classifying patients and controls using multi-dimensional scaling and exploring the metric of semantic space

Multi-dimensional scaling (MDS) has been used to visualize the organization of semantic memory in normal control subjects and in psychiatric conditions such as schizophrenia and Alzheimer's disease. However, the potential for such techniques to classify subjects into diagnostic groups has not been realized. This study attempted to tackle this by developing classification statistics and by exploring the dimensional organization of semantic space using models with different underlying metrics. The test data were from controls and patients with early onset schizophrenia. The results indicated subtly altered semantic organization in schizophrenia, sufficient for novel classification statistics to correctly classify subjects as either patient or control with >80% accuracy.