fNIRS-Based characterization of adolescent depression using dynamic functional connectivity biomarkers in a verbal fluency task

TL;DR

This study uses brain imaging to identify dynamic connectivity patterns in adolescents with depression during a language task, offering potential biomarkers for diagnosis and symptom assessment.

Contribution

Introduces dynamic functional connectivity biomarkers from fNIRS data to characterize adolescent depression during a verbal fluency task.

Findings

A random forest model achieved 86.32% accuracy in distinguishing adolescents with MDD from healthy controls.

Specific brain pathways showed significant differences in connectivity related to depressive and anxious symptoms.

Dynamic connectivity features in certain brain regions predicted anhedonic symptoms in adolescents with MDD.

Abstract

The human brain is a dynamic neural system with time-varying functional connectivity (FC) strengths between brain regions. Evidence indicates that adolescents with major depressive disorder (MDD) exhibit decreased average FC strength in cognitive tasks. Nevertheless, research focused on dynamic FC analysis in this population remains limited. This study aims to identify cognitive task-related dynamic FC features as valuable biomarkers to characterize clinical symptoms in adolescents with MDD.

A total of 83 adolescents with MDD and 78 age/sex-matched healthy controls (HCs) were recruited. We utilized functional near-infrared spectroscopy (fNIRS) to record brain functional data from participants while they performed the verbal fluency task (VFT). An analytical framework for fNIRS data was proposed, in which the average FC strength values over the entire VFT duration and the principal components (PCs) of dynamic (time-varying) FC strength values were extracted as static and dynamic FC features, respectively. A random forest model was built to distinguish adolescents with MDD from HCs. Statistical analyses of the FC features were conducted to identify between-group differences, as well as their relationships with clinical symptoms in adolescents with MDD.

The random forest model achieved an accuracy of 86.32% (95% confidence interval: 83.75%-89.38%) for distinguishing adolescents with MDD from HCs. Significant between-group differences emerged in several FC features (false discovery rate-corrected q < 0.05). For adolescents with MDD, the average FC strength value in the right dorsolateral prefrontal cortex (DLPFC) ~ right medial prefrontal cortex (mPFC) pathway was a significant predictor of depressive and anxious symptoms; the 3rd PC of dynamic FC strength values in the left DLPFC ~ left temporal lobe (TL) pathway and the 5th PC of dynamic FC strength values in the right mPFC ~ right TL pathway were significant predictors of anhedonic symptoms.

VFT-related static and dynamic FC features in specific brain pathways are potential biomarkers for characterizing clinical symptoms in adolescents with MDD. The developed random forest model holds promise as a diagnostic tool for MDD in adolescents.

Not applicable.

The online version contains supplementary material available at 10.1186/s12888-026-07799-3.

Figures