Abstract
Background: Repetitive transcranial magnetic stimulation (rTMS) is an established2 intervention for treatment-resistant depression, but response rates remain highly variable and3 reliable predictors of outcome are lacking. Resting-state electroencephalography (EEG),4 combined with machine learning, represents a cost-effective and accessible approach for5 individualized treatment planning.6 7 Methods: One hundred patients with major depressive disorder (MDD; 51 responders, 498 non-responders) underwent pre-treatment resting-state EEG prior to rTMS. Recordings were9 obtained using either 8 or 30 electrodes. Features including band power, coherence,10 phase-lag index (PLI), and phase-locking value (PLV) were extracted. Machine learning models11 were constructed using two-stage feature selection (recursive feature elimination with support12 vector classifier and sequential backward selection) and linear discriminant analysis (LDA).13 Model performance was assessed with 5-fold cross-validation and 1,000 shuffle iterations.14 15 Results: Multi-feature EEG models consistently outperformed single-feature approaches in16 predicting rTMS treatment response. The best-performing model, based on only 8 electrodes,17 achieved an accuracy of 78.9% and an AUC of 73.3%, surpassing the 30-electrode18 configuration. PLI emerged as the strongest individual predictor, but combining multiple EEG19 features substantially improved classification. LDA provided the most stable performance20 across limited datasets.21 22 Conclusions: Resting-state EEG combined with machine learning can reliably predict rTMS23 treatment response in MDD. Notably, accurate prediction was achieved with a minimal24 8-electrode montage, supporting the clinical feasibility of low-cost EEG assessments. This approach may facilitate personalized treatment selection and improve rTMS outcomes in1 routine psychiatric care.
Authors
Chang C., Sack A.T., Chu Chang, C. &., & H.
https://doi.org/10.1101/2025.10.30.25339032