Disrupted Global Brain Dynamics in Adolescents With Comorbid Anxiety and Depression: Neural Mechanisms and
Shangfeng Han1, Yaohui Lin1, Jie Gao2,3
1Department of Psychology and Center for Brain and Cognitive Sciences, School of Education, Guangzhou University, Guangzhou, China, gzhu.edu.cn.
Background:
Adolescents with comorbid anxiety and depression (ACAD) represent a significant mental health challenge that necessitates the identification of objective neurobiological markers for accurate diagnosis and intervention. Previous electroencephalography (EEG) studies have often relied on single-domain or single-feature analyses, which may fail to capture the multidimensional nature of the underlying neuropathology. This study aimed to characterize neurophysiological alterations in adolescents with ACAD using a multidimensional EEG analytical framework integrating spectral power, functional connectivity (FC), and microstate (MS) dynamics.
Methods:
Forty adolescents with ACAD and 42 healthy controls (HCs) were included. Resting-state EEG data were analyzed to extract spectral power, weighted phase lag index-based FC, and MS parameters. Three types of support vector machine (SVM) classification were implemented, including single-feature SVM analysis, multivariate SVM (MV-SVM) without principal component (PC) analysis (PCA), and PCA-based SVM, to distinguish the independent discriminative value of individual features from the effects of multivariate feature integration and PCA-based dimensionality reduction.
Results:
Adolescents with ACAD showed marginally reduced delta power, altered alpha-band FC, and disrupted MS dynamics, including altered MS class expression and abnormal transition probabilities. Among individual MS features, the transition probability from MS B to MS C showed the highest independent discriminative value, achieving an accuracy of 75% and an area under the receiver operating characteristic (ROC) curve (AUC) of 0.86. Multivariate integration of all MS parameters substantially improved classification performance, with the MV-SVM achieving an accuracy of 0.95 and an AUC of 0.98. The PCA -SVM model showed comparable performance, with an accuracy of 0.94 and an AUC of 0.98, while reducing correlated MS features to three PCs explaining more than 90% of the variance.
Conclusion:
These findings suggest that adolescent ACAD is characterized by global dynamic network dysregulation rather than isolated local abnormalities. Integrated MS dynamics, particularly when modeled in a multivariate framework, may serve as candidate biomarkers for distinguishing adolescents with ACAD from HCs. This multidimensional EEG approach may contribute to early detection, risk stratification, and targeted intervention in adolescent mental health.
