Non-linear dynamical analysis of the EEG in Alzheimer's disease with optimal embedding dimension

Analysis of regularity in the EEG background activity of Alzheimer's disease patients with Approximate Entropy

“…We infer that brains affected by AD show a more regular and less complex electrophysiological behaviour in the parietal region. This confirms findings associated with the fact that a diffuse slowing of the background activity may be found in the EEGs of patients with AD (Markand, 1990) and that the dynamic processes underlying the EEG recording are less complex for AD patients than for normal subjects (Besthorn et al, 1995(Besthorn et al, , 1997Jeong et al, 1998;Jelles et al, 1999;Jeong et al, 2001a). This reduction of irregularity could be explained by a decrease of dynamical complexity of part of the brain in AD patients.…”

“…Most of these studies have been carried out applying the D 2 (Stam et al, 1995;Besthorn et al, 1995Besthorn et al, , 1997Jeong et al, 1998;Jelles et al, 1999;Jeong et al, 2001a) or the first positive Lyapunov exponent (Jeong et al, 1998(Jeong et al, , 2001a. Measures such as D 2 , K-S entropy, the Lyapunov spectrum and related algorithms have been much studied in the presence of noise and limited data.…”

“…However, the most frequently used non-linear method is the correlation dimension (D 2 ), a measure of dimensional complexity of the underlying system (Grassberger and Procaccia, 1983a). Results showed that AD patients had lower D 2 values than controls (Stam et al, 1995;Jeong et al, 1998Jeong et al, , 2001a. Furthermore, Besthorn et al (1995) showed that a lower D 2 was correlated with increased severity of dementia and also found that this method correctly classified AD patients and controls with an accuracy of 70% (Besthorn et al, 1997).…”

Analysis of MEG Background Activity in Alzheimer’s Disease Using Nonlinear Methods and ANFIS

“…22,39 The first non-linear methods used to study the brain recordings from AD patients were correlation dimension (D2) and the first Lyapunov exponent (L1). Jeong et al 20 showed that AD patients exhibit significantly lower D2 and L1 values than controls in most EEG channels. Using D2, another study revealed a decreased complexity of the MEG background activity in AD patients in the low frequency bands, and an increase in the high bands.…”

“…Reliable estimation of L1 and D2 requires a large number of data points and stationary and noise-free time series. 8,20 These requirements are difficult to fulfill for physiological data. Hence, other non-linear methods are necessary to study brain recordings.…”