EEG Phase Space Feature And Phase Amplitude Coupling Analysis For Mild Cognitive Impairment

Cognitive disorder is closely connected with physiological age which seriously affect patients daily living, and the rapid recession of cognitive function in the elderly can lead to Alzheimer’s disease. At present, there is no substantive breakthrough in the treatment of Alzheimer disease, so more and more research focus on early diagnoses and prevention of mild cognitive impairment(MCI). Diabetes mellitus would likely be an important risk factor for cognitive decline and then it will develop to Alzheimer’s disease. As a kind of fast and noninvasive method, Electroencephalogram(EEG) oscillation can help us detect the patients with MCI, which is meaningful to our community. Accordingly, it is of realistic significance to study the EEG of Diabetes mellitus with mild cognitive impairment and help us to explore the mechanism of cognitive disorder. In this paper the phase space reconstruction based on Hilbert-Huang Transformation algorithm and Phase amplitude coupling algorithm are studied, and the actual EEGs collected from diabetic with MCI and extract EEG features are analyzed for Semi-Automated Diagnosis.Firstly, the paper studied the phase space reconstruction based on empirical mode decomposition algorithm and then the algorithm is simulated with dataset which is publicly available online, and the results of statistical test indicate that the features extracted from 2dimensions and 3dimensions had significant difference in two group people. Then the methods had also been used to analyze the actual diabetic brain EEGs and figure out the changes associated with mild cognitive impairment. The non-parametric statistical method, Kruskal-Wallis test, was employed to analysis the features from amnestic mild cognitive impairment(aMCI) groups and normal mild cognitive impairment(nMCI) groups. A feature extraction technique was based on maximization of the area under the curve, and which supports Vector Machines and was used to implement Semi-Automated Diagnosis, the accuracy rate was 79.17%.Secondly, this paper use the method of phase and amplitude coupling to study coupling characteristic of the spontaneous EEG signals in different frequency band. Weutilize simulated signals to analysis the method of phase and amplitude coupling(PAC)and the method of phase and amplitude coupling with statistical(SPAC) in the aspect of width of window and signal to noise ratio. The results showed that the SPAC had a better performance than the PAC in the aspect of width of window and signal to noise ratio.Finally, SPAC was employed to analyze the difference of spontaneous EEG signals between diabetes mellitus patients with mild cognitive impairment, parameter termed percentage coupling energy(PCE) was proposed, the PCE statistical results showed that there is a significant difference in frontal region temporal region and parietal region between aMCI and nMCI groups. Then a feature extraction technique based on maximization of the area under the curve, support Vector Machines and compressed sensing classifier were used to implement Semi-Automated Diagnosis and the accuracy rate was 73.33% and 83.33% respectively.