Summary: A new EEG-based method distinguishes anxious from non-anxious depression with approximately 91% accuracy when brain activity is recorded with the eyes closed.
Source: University of Surrey
Researchers at the University of Surrey have developed a brain-signal processing approach that can reliably differentiate anxious depression from non-anxious depression using short resting-state EEG recordings.
In the study, classification accuracy reached about 91% in the eyes-closed condition. This finding points to a potential objective biomarker that could help clinicians identify patients who experience both depressive and anxiety symptoms, a subgroup known to have more severe symptoms, greater side effects, and higher resistance to treatment than patients with depression alone.
Lead author Hesam Shakouh Alaei from the University of Surrey explained the clinical importance: “Patients with anxious depression often have more severe symptoms and side effects with a higher resistance to treatment than patients with non-anxious depression. This is why it is critical to distinguish these two cohorts from each other. The hope is that these insights will help health professionals recognize anxious and non-anxious depression, and to treat accordingly.”
The team recorded five minutes of resting electroencephalogram (EEG) activity from each participant under both eyes-open and eyes-closed conditions. The sample included 15 individuals diagnosed with anxious depression and 9 individuals with non-anxious depression. Using source reconstruction, the researchers estimated electrical activity across 68 cortical and subcortical regions.
To model directed information flow within the brain, the study used the directed transfer function to build directed brain networks from the EEG source estimates. From those networks the researchers derived graph theory metrics—including connectivity strength in outgoing and incoming directions and betweenness centrality (a measure of nodal importance within a network).
Significant network features were selected statistically using the Mann–Whitney U test. Those discriminative features were then fed into a Support Vector Machine (SVM) classifier to separate anxious from non-anxious depression. The highest performance was achieved using outward connectivity strength features in the eyes-closed condition, reaching 91.66% accuracy, an F-score of 87.5%, and 100% specificity.
Notable patterns emerged from the analyses. Patients with anxious depression showed stronger directed connectivity in the right hemisphere, with particularly robust differences when recorded eyes closed. The study also found that features within the beta frequency band showed the largest group differences. In addition, anxious depression was associated with increased inward and outward connectivity and reduced nodal centrality in posterior regions of the default mode network.
About this depression research news
Author: Katherine Ingram
Source: University of Surrey
Contact: Katherine Ingram – University of Surrey
Image: The image is in the public domain
Original Research: Closed access. “Directed brain network analysis in anxious and non-anxious depression based on EEG source reconstruction and graph theory” by Hesam Shakouh Alaei. Biomedical Signal Processing and Control
Abstract
Directed brain network analysis in anxious and non-anxious depression based on EEG source reconstruction and graph theory
Because anxious depression combines substantial symptoms of both anxiety and major depression, it often presents with greater severity, more treatment side effects, and higher treatment resistance than non-anxious depression. Identifying reliable biomarkers that distinguish these subtypes can improve diagnosis and guide treatment choices.
This study recorded five minutes of resting EEG from 15 patients with anxious depression and 9 patients with non-anxious depression in both eyes-open and eyes-closed states. Sixty-eight brain regions were reconstructed from scalp EEG using exact low-resolution brain electromagnetic tomography (eLORETA), and directed transfer function was applied to estimate directed connectivity between regions.
Graph-theory-based measures—especially outward connectivity strength and betweenness centrality—were calculated from the directed networks. The most discriminative features were identified with the Mann–Whitney U test and used to train a Support Vector Machine classifier to separate anxious and non-anxious depressive groups.
The classifier produced its best results in the eyes-closed condition: outward connectivity strength features produced 91.66% accuracy, an 87.5% F-score, and 100% specificity. The anxious depressed group displayed increased connectivity, particularly outward connectivity in the right hemisphere, with many distinguishing features concentrated in the beta band. Posterior default mode network regions showed increased directed connectivity alongside reduced nodal centrality.
These preliminary but promising findings suggest that directed EEG network measures could offer objective markers for recognizing anxious depression, potentially aiding more tailored clinical assessment and intervention.