Diagnosis of dementias using partial least squares discriminant analysis.

Twenty patients with Alzheimer's disease (AD), 34 with senile dementia of the Alzheimer type (SDAT) and 12 with vascular dementia (VAD) were carefully investigated. The diagnoses were made according to the DSM-III-R criteria and for AD/SDAT also according to the NINCDS-ADRDA criteria. The data set consisted of 29 descriptor variables obtained at the examination, and the data were fitted to a partial least squares model using dummy variables for the response variables (i.e. clinical diagnoses). The distribution of the scores over the descriptor variable space revealed adequate classification regarding all three diagnoses, with some overlap between SDAT and AD and between VAD and SDAT. The study demonstrates a clear difference in symptomatology between AD and SDAT with no influence of age per se or the severity of dementia.