Complexity Analysis Of Depression EEG Signals Based On The Two-dimensional Amplitude-period Features

Major depressive disorder(MDD) is a common affective disorder, It is mainly characterized by depressed mood. Currently the causes and mechanism of MDD is not yet fully study clearly, fully understanding the mechanism of MDD, improving diagnosis and treatment of depression, is still very important scientific and clinical issues. As the reaction of a large number of neural activity in the scalp, EEG contains a lot of physiological and pathological information, and has significant nonlinear features. Thus, the application of nonlinear methods of EEG analysis has become an important direction of EEG studies.Based on information entropy and sample entropy theories, the present study proposed two basic algorithms to calculate the two-dimension oscillation entropies. Simulation results showed that the two-dimensional information entropy(D2En) could effectively describe the distribution of the oscillation pattern in amplitudeperiod space, and the sample entropy(D2SEn) could effectively reflect the probability of new pattern generation in the amplitude-period space. When the complexity of the waveform mainly comes from the interaction effect of amplitudes and periods, the two-dimensional oscillation entropies are more effective than one-dimensional cases.Results of resting-state EEG analysis showed that, in the Alpha band, D2 Ens in the left parietal region and bilateral occipital regions in MDD group were significantly higher than control group, indicating that in the Alpha band, distribution patterns of the EEGs in MDD group are more dispersed in the amplitude-period space. In the Alpha band, D2 SEn values of MDD group in left parietal and left occipital regions were significantly reduced compared with the control group, indicating that in the Alpha band, the probability of new pattern generations of EEGs in MDD group are significantly decreased. The two-dimensional oscillation entropies of resting-state EEGs can become the potential biomarkers of MDD.Based on the self-related emotional word judgment task, the Beta Band EEG in task state was analyzed. The results of D2 En showed that, D2 Ens of the MDD and control groups had significant person effect and valence effect, indicating that D2 Ens of EEG signals can effectively distinguish different kinds of cognitive processes. The results of D2 SEn show that, D2 SEn values of the two groups had significant valence×group interaction effect, indicating that, for different kinds of valence conditions, the EEG modes between MDD and control groups can probably differ from each other. This feature can become the potential biomarker of MDD.