Objective:In order to analyze the causes of potential changes of default mode network(DMN)in the brain of patients with Parkinson's disease and the relationship between its clinical characteristics,and to explore how to extract its EEG characteristics and classify them accurately and effectively.Methods:26 subjects in Parkinson's disease group and 26 subjects in healthy control group were selected as experimental objects.Partial directed coherence(PDC)was applied to the data sequence of DMN related electrodes to obtain the effective connection between the control group and Parkinson's disease subjects.After statistical analysis,the six PDC connections with significant differences are obtained discussed in depth.Further,these connection values are grouped in feature sets and classfied.Results:Compared with the control group,the connections related to attention control decreased in patients with Parkinson's disease,and some connections related to working memory increased in patients with Parkinson's disease compared with healthy patients.At the same time,XGBoost algorithm is used to classify the feature set,and the average test accuracy is 76.5%.Conclusion:There is a significant relationship between non motor symptoms and DMN network in patients with Parkinson's disease at rest,which is manifested in attention control and memory function,which is closely related to the damage of BA area in DMN.Subsequently,the classification of the two categories of subjects also verified the effectiveness of PDC algorithm in DMN analysis,and provided a new way for the testing and prevention of Parkinson's patients.
国家科技期刊平台
NSTL
维普
万方数据
