Difference Analysis of Brain Network Working Memory Data with EEG Sub-Sequence Feature Vector as Node

摘要

Many studies have shown that microstates are related to psychological processes. This study investigated the microphysical activities of brain microstates and working memory, and found that there are significant differences in microstates at different stages of working memory and there are also significant differences in microstates between normal people and patients. However, the microstate can only analyze the EEG signal from a global perspective, and cannot understand the EEG signal characteristics of each channel in detail. This study proposes a new analysis method based on microstates to analyze EEG signals, including segmentation, feature extraction, EEG signal feature selection, channel network construction, and channel network characteristic analysis. In the segmentation, the microstates used to segment the multi-channel signal, and divide each channel signal into sub-sequences of different lengths. The feature extraction mainly uses the features frequently used in EEG signals, including statistical features, nonlinearities, and entropy characteristics. In this experiment, a sequence forward selection algorithm is used to select a set of effective features that best represent the EEG signal. In the channel construction network, the Pearson correlation coefficient is used to calculate the correlation between each sub-segment of each channel of the working memory. Finally, the network attributes of the networks and the similarity between each channel are analyzed. It is found that there are significant differences between normal people and patients in the network properties constructed and the channel network, which provides a basis for the diagnosis and treatment of patients with schizophrenia.