Neural Network in Diagnosis of Alzheimer’s from Electroencephalography

Alzheimer’s disease (AD) is an irreversible, progressive brain disorder that slowly destroys memory, thinking skills, and eventually the ability to carry out the simplest tasks. Recent estimates indicate that the disorder may rank third, just behind heart disease and cancer, as a cause of death for older people. Early diagnosis of AD helps to ensure prescription of the medications when they are most useful. Early diagnosis of AD also allows prompt treatment of psychiatric symptoms such as depression or psychosis. Early diagnosis raises the chance of treating the disease at a nascent stage, before the patient suffers permanent brain damage. The purpose of the project is to assist neurologist in detecting and monitoring the Alzheimer’s diseases at the early stage by analyzing the EEG recordings. The project proposes the classification of EEG signal of patients suffering from Alzheimer’s diseases in the early stages using artificial neural networks. The classification is carried out by identifying the abnormalities of EEG signal. Different types of neural networks are used for classifying the EEG signals into 2 classes (Alzheimer’s or Normal) and this neural networks are compared with the parameter classification accuracy to predict which neural network model is best for classifying EEG signal. The input for the classifier is a vector which contains the features. The features relevant for distinguishing Alzheimer’s patients EEG is extracted from the EEG Signal. The Discrete Wavelet Transform (DWT) is employed for extracting features from EEG. Use of wavelet transform for the feature extraction, which is faster and enables better resolution and high performance for representation and visualization of the abnormal activity in EEG than other methods. The db4 wavelet selected enables it appropriate for detecting changes in EEG signals because of its smoothing feature. The Feed-forward neural network (FNN), Block based neural network (BBNN) and Convolutional neural network (CNN) are used as classifiers and a comparative study is conducted to choose best classifier.