Non-Linear Dynamic Methods Researching Eeg Signals In Mild Cognitive Impairment Of Diabete Patients

Diabetes mellitus and cognitive impairment complications are gradually increasing year by year, and seriously affecting the patients’ life. Diabetes mellitus would likely be an important risk factor for cognitive decline and dementia. It can deepen the understanding of diabetes impairment mechanism by researching the electroencephalogram related to cognitive function in diabetes mellitus, moreover, it also has a significance for the diagnose and interventions of cognitive function in patients in early stage. To investigate the EEG nonlinear characteristic of the amnestic mild cognitive impairment in diabetes mellitus, the paper studied some kinds of entropies based on empirical mode decomposition(EMD) and proposed the symbolic time series recurrence plots analysis(SRP) with EEG data obtained from diabetes and non-diabetes.Firstly, the paper explored six entropies combining EMD, including approximate entropy, sample entropy, fuzzy entropy, permutation entropy, power spectrum entropy and wavelet entropy. These algorithms were simulated with the Logistic model, and the results showed that FEn has a better noise resistance. One way analysis of variance, a feature extraction technique based on maximization of the area under the curve, and support Vector Machines were used to do analyze the of nonlinear characteristic between the amnestic mild cognitive impairment and normal control group in dabetes mellitus. Pearson’s linear correlation was employed to study associations between these entropies and cognitive functions. The results demonstrated that fuzzy entropy based on EMD had a better classification accuracy. Frontal, temporal and occipital regions were highest ranking salient regions. Entropy based on EMD was correlated to neuropsychological test in temporal and occipital regions.Then, the symbolic mode combined with recurrence plots was proposed. It needs not to consider a long and stationary data. And it can directly analyze the distribution structure of symbolic pattern by a symbolic encoding method. SRP was simulated with the Logistic model and the neural mass model of signal channel and double dynamic. Simulation results showed that the determinism variable based on SRP could better reflect the change of model parameter than the recurrence plots based on order.Finally, SRP also analyzed the amnestic mild cognitive impairment in diabetes mellitus and without diabetes mellitus. Results concluded that the determinism variable based on SRP had a better performance in classification and found a possible association between diabetes and cognitive function under the condition of cognitive match. There was a correlation between determinism variable and neuropsychological test in the region of frontal, temporal and occipital regions.