EEG Signal Analysis using Empirical Mode Decomposition and Time-Frequency Method

Abstract

The Electroencephalogram (EEG) signals are widely used by physicians for interpretation and identification of physiological and pathological phenomena. The subtle features inherent in the EEG signals are to be extracted for the proper diagnosis of diseases. This work aims at introducing techniques which could be employed for the extraction of features inorder to distinguish mild cognitive impairment Alzheimer Disease (MCI-AD) from healthy controls under resting and cognitive eyes closed condition. The EEG signals are analyzed using time domain and time-frequency domain methods. Empirical Mode Decomposition (EMD), a powerful adaptive technique, is employed for decomposing the highly non-linear and non-stationary EEG signals, acquired from MCI-AD patients and controls. EMD, being a time-domain analysis technique, decomposes the signals into finite number of intrinsic mode functions (IMF) which have variations in both amplitude and frequency. Features are extracted from the decomposed modes using the complexity measure of multiscale entropy (MSE). Six-level multiresolution wavelet decomposition is carried out using db10 as the mother wavelet on the acquired EEG signals. Feature extraction is performed on the wavelet decomposed signals utilizing the irregularity statistic of Conditional entropy (CE). These features are extracted for different frequency bands of gamma, beta, alpha, theta and delta. Lower complexity was observed in MCI-AD patients under cognitive task conditions than the resting state. Loss of irregularity in EEG signals of MCI-AD under cognitive task was revealed through the studies of CE. The results of the study reveal that a decline in cognitive and executive functioning occurs in early stages of AD.