Fuzzy neural network model for assessment of Alzheimer-type dementia

A system for assessing dementia of the Alzheimer type (DAT) from electroencephalogram (EEG) data by means of fuzzy neural networks (FNNs) was investigated. The system consisted of two FNN models, one to discriminate DAT patients from normal subjects and the other to estimate the severity of the DAT patients' symptoms. EEG data were collected using 15 electrodes attached to the scalp. The power spectra were calculated by the fast Fourier transform. For each electrode, the power spectrum was divided into 9 frequency bands and relative power values were calculated. The θ₁ (4.0-6.0 Hz), θ₂ (6.0-8.0 Hz), and α (8.0-13.0 Hz) band data were used as the network input values. DAT severity was assessed by the Mini-Mental State (MMS) examination administered to each patient and the results were used as the output. The FNN model for DAT patient discrimination correctly distinguished 94% of the DAT patients from normal subjects. The FNN model for severity estimation gave an average error of 2.57 points out of 30 in the MMS scores. The FNNs were found to be useful tools for discriminating DAT patients from normal subjects as well as for estimating quantitatively the severity of DAT symptoms from EEG data.