Resting EEG functional connectivity and graph theoretical measures for discrimination of depression

This study aims to clarify whether classification accuracy between major depressive disorder (MDD) and healthy subjects increases with complementary use of graph theoretical measures to functional connectivity. Electroencephalography (EEG) signals were recorded from 37 unmedicated MDD subjects (21 female, 16 male) and 37 gender and age matched control subjects. Signals were recorded during eyes-closed resting state from 30 EEG channels. Six frequency bands were analysed: delta (1-4 Hz), theta (4-8 Hz), alpha (8-12 Hz), beta (12-30 Hz), gamma (30-45 Hz) and total (1-45 Hz) frequency band. Coherence for estimation of functional connectivity and three graph theory measures, clustering coefficient (C), characteristic path length (L) and small-worldness (S), were calculated. Feature selection was conducted with two algorithms: genetic algorithm (GA) and sequential feature selection (SFS). Support Vector Machine (SVM) with 10-fold cross-validation was used for classification. Repeated cross-validation test and McNemar's test were used to compare classification accuracies. The results of the study indicate that adding graph theoretical measures to functional connectivity does not significantly increase classification accuracy for distinguishing MDD and healthy subjects. These results may suggest causal connection between abnormalities in functional network organization and functional connectivity.