Predicting Treatment Response of rTMS in Major Depressive Disorder Using a Explainable Machine Learning Model Based on EEG and Clinical Features.

Abstract

Objective: Major Depressive Disorder (MDD) is highly heterogeneous in response to rTMS, and identifying predictive biomarkers is essential for personalized treatment. However, current researches either only uses EEG or clinical features, lacks interpretability, or have small sample sizes.

Method: This study included 74 MDD patients who responded (Responder) and 43 MDD who did not respond (NonResponder) to rTMS. Eight baseline EEG metrics and clinical features were sent to seven machine learning models to classify Responder from NonResponder. SHAP was used to interpret feature contributions.

Result: Combining phase locking value (PLV) and clinical features with Support Vector Machine (SVM) achieved optimal classification performance (accuracy=97.33%). SHAP revealed that delta and beta band functional connectivity (F3-P7, F3-P4, P3-P8, T7-Cz) significantly influenced predictions and differed between groups.

Conclusion: This study developed an explainable predictive framework to predict rTMS response in MDD, enhancing the accuracy of rTMS response prediction and supporting personalized treatment in MDD.