CLASSIFICATION AND STAGING OF DEMENTIA OF THE ALZHEIMER-TYPE - A COMPARISON BETWEEN NEURAL NETWORKS AND LINEAR DISCRIMINANT-ANALYSIS

Objective: To examine the utility of artificial neural networks (ANNs) for differentiating patients with Alzheimer disease from healthy cont rol subjects and for staging the degree of dementia. Design: Compariso n of the classification abilities of ANNs with the statistical techniq ue of linear discriminant analysis (LDA) using the results of 11 neuro psychological tests as predictors. Participants: Ninety-two patients w ith a diagnosis of probable Alzheimer disease (referred from a geriatr ic clinic) and 43 elderly control subjects (independently solicited). The patients met National Institute of Neurological and Communicative Disorders and Stroke-Alzheimer's Disease and Related Disorders Associa tion criteria for probable dementia, with clinical ratings of dementia severity derived from the Cambridge Examination for Mental Disorders of the Elderly (CAMDEX). Main Outcome Measures: Classifications betwee n and within groups were determined by using LDA and ANNs, and more de tailed comparisons of the 2 methods were performed by using chi(2) ana lyses and unweighted and weighted kappa statistics. Results: Linear di scriminant analysis correctly identified 71.9% of cases. Artificial ne ural networks, trained to classify the subjects using the same data, c orrectly classified 91.1% of the cases. Subsidiary analyses showed tha t although both techniques effectively discriminated between the contr ol subjects and patients with dementia, the ANNs were more powerful in discriminating severity levels within the dementia population. The an alyses for goodness of fit revealed that the ANN classification produc ed a better fit to the actual data. A comparison of the weighted propo rtion of agreement between the criterion and predictor variables also showed that the ANNs clearly outperformed LDA in classification accura cy for the full data set and patients-only data set. Conclusion: The r esults demonstrate the utility of ANNs for group classification of pat ients with Alzheimer disease and elderly controls and for staging deme ntia severity using neuropsychological data.