Shared and Unshared Feature Extraction in Major Depression During Music Listening Using Constrained Tensor Factorization.

Zhen Mu,Jing Xu,Jianlin Wu,Wenya Liu,Qing Zhang,Xiaoyu Wang,Yi Chang,Xiulin Wang,Fengyu Cong

doi:10.3389/fnhum.2021.799288

Abstract

Ongoing electroencephalography (EEG) signals are recorded as a mixture of stimulus-elicited EEG, spontaneous EEG and noises, which poses a huge challenge to current data analyzing techniques, especially when different groups of participants are expected to have common or highly correlated brain activities and some individual dynamics. In this study, we proposed a data-driven shared and unshared feature extraction framework based on nonnegative and coupled tensor factorization, which aims to conduct group-level analysis for the EEG signals from major depression disorder (MDD) patients and healthy controls (HC) when freely listening to music. Constrained tensor factorization not only preserves the multilinear structure of the data, but also considers the common and individual components between the data. The proposed framework, combined with music information retrieval, correlation analysis, and hierarchical clustering, facilitated the simultaneous extraction of shared and unshared spatio-temporal-spectral feature patterns between/in MDD and HC groups. Finally, we obtained two shared feature patterns between MDD and HC groups, and obtained totally three individual feature patterns from HC and MDD groups. The results showed that the MDD and HC groups triggered similar brain dynamics when listening to music, but at the same time, MDD patients also brought some changes in brain oscillatory network characteristics along with music perception. These changes may provide some basis for the clinical diagnosis and the treatment of MDD patients.

Highlights

Major depressive disorder (MDD) is a globally prevalent mental disorder with multifactorial causes (Belmaker and Agam, 2008; Gotlib and Joormann, 2010; Jia et al, 2010)
In terms of the number of components for the tensor data, a simple explained variance-based principal component analysis (PCA) method was adopted (Liu et al, 2021), and the number of principal components with 99% accumulated explained variance were assigned to RHC and RMDD, we set RHC = 26, RMDD = 36
We first carried out the proposed nonnegative coupled tensor factorization (NCTF)-ADMM algorithm 50 times on the two groups of ongoing EEG data, and through correlation analysis and hierarchical clustering, we totally obtained 5 clusters of shared and unshared component patterns between/in healthy controls (HC) and major depression disorder (MDD) groups

Summary

Introduction

Major depressive disorder (MDD) is a globally prevalent mental disorder with multifactorial causes (Belmaker and Agam, 2008; Gotlib and Joormann, 2010; Jia et al, 2010). The neural mechanisms of MDD have been widely explored using non-invasive neuroimaging techniques, like functional magnetic resonance imaging (fMRI), electroencephalogram (EEG), and magnetoencephalography (Gotlib and Hamilton, 2008; Kaiser et al, 2015). Few studies have correlated the music perceptive arousal with neural mechanisms in MDD, and the studies mainly explored the networks of brain connectivity at the source space (Liu et al, 2020, 2021; Zhu et al, 2021). The current findings are often inconsistent or even contradictory due to the different methodological approaches and involved participants, unified neural mechanisms of MDD (in music perception) can not be concluded (Zhi et al, 2018). We aim to investigate the biomarkers in MDD during music listening using EEG data at the sensor level

Objectives

Results

Conclusion