Common and differential EEG microstate of major depressive disorder patients with and without response to rTMS treatment
Zongya Zhao,Xiangying Ran,Junming Wang,Shiyang Lv,Mengyue Qiu,Yanxiang Niu,Chang Wang,Yongtao Xu,Zhixian Gao,Wu Ren,Xuezhi Zhou,Xiaofeng Fan,Jinggui Song,Yi Yu
DOI: https://doi.org/10.1016/j.jad.2024.09.040
2024-12-15
Abstract:Objective: Repetitive transcranial magnetic stimulation (rTMS) has recently emerged as a novel treatment option for patients with major depressive disorder (MDD), but clinical observations reveal variability in patient's responses to rTMS. Therefore, it is clinically significant to investigate the baseline neuroimaging differences between patients with (Responder) and without (NonResponder) response to rTMS treatment and predict rTMS treatment outcomes based on baseline neuroimaging data. Method: Baseline resting-state EEG data and Beck Depression Inventory (BDI) were collected from 74 rTMS Responder, 43 NonResponder, and 47 matched healthy controls (HC). EEG microstate analysis was applied to analyze common and differential microstate characteristics of Responder and NonResponder. In addition, the microstate temporal parameters were sent to four machine learning models to classify Responder from NonResponder. Result: There exists some common and differential EEG microstate characteristics for Responder and NonResponder. Specifically, compared to the HC group, both Responder and NonResponder exhibited a significant increase in the occurrence of microstate A. Only Responder showed an increase in the coverage of microstate A, occurrence of microstate D, transition probability (TP) from A to D, D to A, and C to A, and a decrease in the duration of microstates B and E, TP from A to B and C to B compared to HC. Only NonResponder exhibited a significant decrease in the duration of microstate D, TP from C to D, and an increase in the occurrence of microstate E, TP from C to E compared to HC. The primary differences between the Responder and NonResponder are that Responder had higher parameters for microstate D, TP from other microstates to D, and lower parameters for microstate E, TP from other microstates to E compared to NonResponder. Baseline parameters of microstate D showed significant correlation with Beck Depression Inventory (BDI) reduction rate. Additionally, these microstate features were sent to four machine learning models to predict rTMS treatment response and classification results indicate that an excellent predicting performance (accuracy = 97.35 %, precision = 96.31 %, recall = 100 %, F1 score = 98.06 %) was obtained when using AdaBoost model. These results suggest that baseline resting-state EEG microstate parameters could serve as robust indicators for predicting the effectiveness of rTMS treatment. Conclusion: This study reveals significant baseline EEG microstate differences between rTMS Responder, NonResponder, and healthy controls. Microstates D and E in baseline EEG can serve as potential biomarkers for predicting rTMS treatment outcomes in MDD patients. These findings may aid in identifying patients likely to respond to rTMS, optimizing treatment plans and reducing trial-and-error approaches in therapy selection.