CLASSIFICATION OF QUANTITATIVE EEG DATA BY AN ARTIFICIAL NEURAL-NETWORK - A PRELIMINARY-STUDY

Previous studies from different laboratories have suggested that qEEG could be useful for distinguishing dementia from normality. Our aims w ere: (1) to study the ability of qEEG to distinguish dementia among di fferent pathological conditions in ambulatory settings; (2) to compare the ability of classical statistical analysis and of neural networks in classifying qEEG data. We were able to obtain a multiple discrimina nt function using a training set of patients, which classified correct ly more than 91% of the qEEGs from an independent group of patients, w ith less than 5% of false positives. Kohonen's neural network was trai ned with the same set of patients. This unsupervised learning artifici al neural network performed the classification of the independent samp le with an accuracy comparable to that of the multiple discriminant fu nction. Our results suggest that the use of unsupervised learning algo rithms could be an interesting alternative in the classification of da ta obtained from psychiatric patients where definition of their clinic al profile is not always a simple task.