Analysis of bivariate time series of magnetoencephalography for depression Based on Granger Kernel Method

Konferenz: BIBE 2019 - The Third International Conference on Biological Information and Biomedical Engineering
20.06.2019 - 22.06.2019 in Hangzhou, China

Tagungsband: BIBE 2019

Seiten: 6Sprache: EnglischTyp: PDF

Persönliche VDE-Mitglieder erhalten auf diesen Artikel 10% Rabatt

Hu, Jialin; Wang, Jun (Smart Health Big Data Analysis and Location Services Engineering Lab of Jiangsu Province, Nanjing University of Posts and Telecommunications, Nanjing, China)
Yan, Wei (Department of Psychiatry, The Affiliated Brain Hospital of Nanjing Medical University, Nanjing, China)
Li, Jin (College of Physics and Information Technology,Shaanxi Normal University, Xi’an, China)
Hou, Fengzhen (School of Science, China Pharmaceutical University, Nanjing, China)

Granger causality is widely used to analyze dynamic systems, among which the linear granger causality method is a very popular method to analyze the causality between bivariate time series. In this paper, an improved granger causality method, kernel granger causality method, is adopted to analyze the causality of the time series of symmetrical channels in the left and right frontal areas of magnetoencephalography (MEG) in normal people and depressed patients by specifically using the inhomogeneous polynomial(IP) kernel method and Gaussian kernel. The results showed that both methods showed that the causality index from the right frontal area to the left frontal area is greater than in the opposite direction in both healthy subjects and depressed patients. Different from Gaussian kernel method, the results of Inhomogeneous polynomial kernel also showed that the causality index of depressed patients was more susceptible to stimulation. Specifically, the causality index of both directions for depressed patients was higher than that of normal people under negative and neutral stimulation. By comparing the two different kernel methods, we found that the inhomogeneous polynomial kernel is a more appropriate method to distinguish depressed patients from normal people than the Gaussian kernel in our study.