Pre-treatment structural brain biomarkers predict response to repetitive transcranial magnetic stimulation in subjective tinnitus.

BACKGROUND

Variable efficacy of repetitive transcranial magnetic stimulation (rTMS) for tinnitus necessitates predictive biomarkers. Pre-treatment brain structural features may predict rTMS outcomes, given that tinnitus involves structural brain alterations and rTMS can induce neuroplastic changes.

OBJECTIVE

To identify pre-treatment brain structural biomarkers predictive of rTMS efficacy in subjective tinnitus.

METHODS

We prospectively enrolled 64 patients with subjective tinnitus and 18 healthy controls (HCs). Patients underwent a 2-week course of rTMS. High-resolution T1-weighted structural MRI (sMRI) was acquired, and 242 whole-brain morphometric features were extracted. Univariate analysis identified features differing between responders and non-responders, which subsequently were used to construct a machine learning model evaluated via 5-fold cross-validation and SHapley Additive exPlanations (SHAP) analysis. Feature significance was further interpreted through three-group comparisons among responders, non-responders, and HCs. Spearman correlation analyses were performed between structural features and clinical improvement scores (ΔVAS, ΔTHI) as well as baseline clinical measures.

RESULTS

Thirty-six patients (56.25%) were classified as responders. Ten regional features distinguished responders from non-responders, encompassing prefrontal, limbic, sensorimotor, and parietal networks. The predictive model (ExtraTreesGini_BAG_L1) achieved optimal performance (AUC = 0.85; accuracy = 0.77; precision = 0.71; recall = 0.97; F1-score = 0.82). SHAP analysis identified right pars triangularis of the inferior frontal gyrus (IFGtriang-R) gray matter volume (GMV) as the top predictor (positive influence). Three-group comparison revealed that IFGtriang-R GMV was significantly larger in responders (0.90 ± 0.08) than in both HCs (0.86 ± 0.06) and non-responders (0.86 ± 0.07), indicating a specific structural signature associated with positive treatment outcome. Spearman correlation analyses revealed that IFGtriang-R volume did not significantly correlate with ΔVAS or ΔTHI, and no structural feature showed a robust association with baseline clinical measures after Bonferroni correction.

CONCLUSION

Responders were characterized by relative enlargement of the IFGtriang-R, suggesting a threshold effect of neuroplastic reserve conducive to rTMS efficacy. Pre-treatment sMRI assessment of this region may facilitate patient stratification for rTMS treatment, advancing precision neuromodulation for tinnitus.