Skip to content

EEG-based depression detection via temporal-frequency-dynamical feature fusion.

Nov 2026 · Psychiatry research. Neuroimaging · Vol 365, pp. 112388 · 0 citations · 53 references
Medicine

Abstract

Depression has become a major global mental health illness. Electroencephalography (EEG)-based depression detection, combined with machine learning techniques, provides a promising route for developing novel diagnostic or screening measures. Most existing works represent complex EEG signals using domain-specific information, ignoring the complementarity among the temporal, frequency, and nonlinear dynamical characteristics. This hinders their ability to comprehensively capture the dynamic variations of EEG, thereby restricting model performance. To address this problem, the current work proposes a sorting-guided multi-domain feature coupling framework for depression detection. Specifically, the proposed method extracts meaningful features from the temporal, frequency, and nonlinear dynamical perspectives to characterize EEG signals. Then, an ordered feature coupling strategy is introduced to interactively fuse multi-view characteristics, thereby yielding a better yet compact feature representation. The resulting features can be effectively classified using standard machine learning classifiers. The proposed framework is evaluated on two datasets for distinct classification tasks: discriminating patients with major depressive disorder from healthy controls using MODMA database and distinguishing questionnaire-defined high-risk from low-risk individuals for depression using EDRA database. The experimental results demonstrate the superiority of the presented approach. This study provides a novel perspective on distinguishing different types of EEG signals, laying the groundwork for future research.

View source

We use cookies to run the site and, with your consent, for analytics and to show ads. See our Cookie Policy.