Stationary epoch-based entropy estimation for early diagnosis of Alzheimer's disease

Abstract

Several studies showed that EEG signal of Alzheimer's disease patients is less complex than that of healthy subjects. In this article, we propose to characterize the complexity of the EEG signal by an entropy measure based on local density estimation by a Hidden Markov Model. We first show that this measure leads to consistent results qualitatively and quantitatively (in terms of classification accuracy). Indeed, it discriminates AD patients, at an early stage of Alzheimer's disease, from healthy subjects: a classification accuracy of 80% is reached on a dataset including EEG data recorded in different conditions. Based on this measure, we also show that parietal and temporal regions are the first regions affected by complexity loss in the early stage of Alzheimer's disease.

A Comparative Study of Functional Connectivity Measures for Brain Network Analysis in the Context of AD Detection with EEG

“…We considered three widely used measures: phase lag index (PLI), magnitude square coherence (MSC) and mutual information (MI), relying on different mathematical concepts. We also considered the epoch-based entropy measure (EpEn), already presented in previous works [49][50][51][52] but not largely used in the literature.…”

“…In this work, we propose to use an entropy-based entropy measure, called "epochbased entropy" (EpEn), already introduced and published in [49][50][51][52]. This measure relies on the statistical modeling of EEG signals with a Hidden Markov Model (HMM), which considers the propriety of nonlinearity, nonstationarity and multidimensionality of EEG signals.…”

“…This measure relies on the statistical modeling of EEG signals with a Hidden Markov Model (HMM), which considers the propriety of nonlinearity, nonstationarity and multidimensionality of EEG signals. In [49][50][51][52], we demonstrated that EpEn leads to a better characterization of the dynamic system underlying the observed EEG signals for AD detection since HMM allows a refined statistical modeling of the spatiotemporal changes in EEG time series. More specifically, EpEn quantifies the information content of EEG signals, both at the time and spatial levels, using local density estimation by an HMM, on inter-channel quasi-stationary epochs (segments).…”

Meta‐heuristic multi‐ and many‐objective optimization techniques for solution of machine learning problems

“…Later on, Albukhanajer, Briffa, and Jin (2014) used the same idea to extract features from noisy images. Xue, Zhang, and Browne (2013) proposed two multi-objective versions of the well-known binary particle swarm optimization, one based on mutual information, and the other based on entropy (Houmani, Dreyfus, & Vialatte, 2015). The results obtained from six benchmark data sets have shown the proposed approaches evolve to the Pareto front.…”

Diagnosis of Alzheimer’s Disease by Time-Dependent Power Spectrum Descriptors and Convolutional Neural Network Using EEG Signal

“…EEG signals from AD patients have been shown to show lower (lower complexity) values of certain tests than signals from age-matched control subjects. Other information-theoretical methods, in specific entropy-based approaches, have appeared as theoretically useful EEG indicators for AD: epoch-based entropy [ 43 , 44 ], sample entropy [ 45 ], Tsallis entropy [ 46 ], approximate entropy [ 47 , 48 ], multiscale entropy [ 49 ], and complexity of Lempel-Ziv [ 50 ]. These approaches relate the strength of a signal to unpredictability: irregular signals are more complicated than regular ones because they are erratic.…”