Neurophysiological mechanisms and predictive modeling of SSRI treatment response in depression disorder based on multidimensional EEG features

J Affect Disord. 2025 Oct 16:120424. doi: 10.1016/j.jad.2025.120424. Online ahead of print.

ABSTRACT

BACKGROUND: Depression exhibits significant heterogeneity in antidepressant treatment response. This study aimed to develop an Electroencephalography (EEG)-based machine learning model integrating multidimensional features to predict selective serotonin reuptake inhibitors (SSRI) efficacy and elucidate neurophysiological mechanisms.

METHODS: Resting-state EEG data were recorded from 27 patients diagnosed with depression (Dataset I), and an independent validation cohort of 5 additional patients (Dataset II) was recruited from the same hospital. Patients were classified into drug-effective and drug-ineffective groups based on Hamilton Depression Rating Scale-17 (HAMD-17) score reduction rates. Three categories of EEG features were extracted using multiple time windows, including relative power (RP), fuzzy entropy (FE), and phase lag index (PLI). A machine learning framework integrating recursive feature elimination (RFE) with four classifiers (XGBoost, SVM, RF, and LightGBM) was developed for feature selection and optimal subset identification. The selected features were then used to explore neural mechanisms of pharmacological response, focusing on neurophysiological differences between responders and non-responders.

RESULTS: The proposed SVM-RFE model achieved 96.83 % accuracy using 12-second EEG windows for SSRI efficacy prediction on Dataset I. Independent validation on Dataset II confirmed strong generalizability of the optimized feature subset. Treatment responders demonstrated higher Beta2 power, increased high-frequency functional connectivity (81 % long-range), and predominant engagement of frontal networks.

CONCLUSIONS: The proposed framework achieved superior classification performance for SSRI efficacy prediction and demonstrated excellent generalization on an independent validation set. Beta2 oscillations and long-range connectivity may serve as reliable biomarkers of SSRI treatment response, offering insights into the neurophysiological basis of antidepressant efficacy.

PMID:41109421 | DOI:10.1016/j.jad.2025.120424