Multivariate Pattern Analysis of EEG-Based Functional Connectivity: A Study on the Identification of Depression

• Published in: IEEE access Vol. 7; pp. 92630 - 92641
• Main Authors: Peng, Hong, Xia, Chen, Wang, Zihan, Zhu, Jing, Zhang, Xin, Sun, Shuting, Li, Jianxiu, Huo, Xiaoning, Li, Xiaowei
• Format: Journal Article
• Language: English
• Published: Piscataway IEEE 2019; The Institute of Electrical and Electronics Engineers, Inc. (IEEE)
• Subjects: Biomarkers; Brain; Classifiers; Connectivity analysis; Correlation coefficients; Depression; EEG; Electroencephalography; Frequencies; functional connectivity; Functional magnetic resonance imaging; Indexes; Mental depression; multivariate pattern analysis; Oscillators; Pattern analysis; Permutations; Phase lag; resting-state; Support vector machines
• ISSN: 2169-3536; 2169-3536
• DOI: 10.1109/ACCESS.2019.2927121

Resting-state electroencephalography (EEG) studies have shown significant group differences in functional connectivity networks between patients with depression and healthy controls. The present study aims to identify the altered EEG resting-state functional connectivity patterns of depressed patients, which can be used to test the feasibility of distinguishing individuals with depression from healthy controls. In the present study, the phase lag index was employed to construct functional connectivity matrices. An altered Kendall rank correlation coefficient was used to identify the features with high discriminative power, and several classifiers were employed to classify a total of 27 depressed patients and 28 demographically matched healthy volunteers. Permutation tests were used to evaluate classifier performance. The best classification results demonstrate that more than 92% of subjects were correctly classified by binary linear SVM through leave-one-out cross-validation for the full frequency band, and the AUC was 0.98. Our findings suggest that the depression affects brain activity in nearly the whole cortex and that changes in brain oscillation patterns in the delta, theta, and beta frequency bands are more significant than those in the alpha frequency band. The current study sheds new light on the pathological mechanism of depression and suggests that EEG resting-state functional connectivity analysis may identify potentially effective biomarkers for its clinical diagnosis.