Discriminant analysis of MRI measures as a method to determine the presence of dementia of the Alzheimer type.

Multivariate discriminant analysis of brain volumes obtained from semiautomated magnetic resonance image (MRI) quantification was used in an attempt to identify demented patients very early in the course of the disease. Temporal and posterior frontal brain volumes were quantified from MRIs in a cross-sectional study of 31 male and female patients with dementia of the Alzheimer type (DAT) and 29 age- and sex-matched healthy comparison subjects. Mean scores on the Folstein Mini-Mental State Examination (MMS) were in the mild range for the DAT group (20 +/- 6.6), but patients with moderate and severe dementia were also included (MMS range of entire DAT group = 4-28). Significant mean differences in frontal and temporal lobe brain volumes were found between the DAT group and the age- and sex-matched healthy comparison group, but the sensitivity of any single measure was limited to 87% with a specificity of 83%. Initial multivariate discriminant analysis revealed significant gender differences among the healthy subjects, but not the DAT patients. The large group size allowed for subsequent discriminant analyses to be performed by gender. All healthy subjects and DAT patients were correctly classified by gender-specific discriminant functions. The male discriminant variables included brain volume, age, and temporal lobe measures. Inclusion of age in the male discriminant function accounted for age-related brain atrophy, a finding that may have emerged as a consequence of the broad age range of the male DAT population (50-81 years). The male discriminant function was also successfully applied to an independent group of mildly demented subjects that included patients for whom the diagnosis of dementia was uncertain but verified by follow-up clinical evaluations. Measures of temporal lobe brain matter and temporal lobe cerebrospinal fluid volumes were the significant discriminator variables for the women. Quantitative MRI and multivariate discriminant analysis showed promise in distinguishing the dementing process from healthy aging in a group of 60 subjects. Moreover, while not diagnostic of DAT, the approach appeared to offer additional information about the probability of a diagnosis being later confirmed in patients with very mild dementia for whom the clinical identification of DAT is uncertain.