Multi-state neural dynamics encode antidepressant response: Fusion of resting and task networks

Although antidepressants demonstrate broad efficacy for major depressive disorder (MDD), treatment responses vary considerably across individuals. Consequently, pretreatment identification of eventual Responders (Res) versus Non-responders (Non), alongside patient-specific prediction of symptomatic improvement under a given regimen, remains clinically imperative. Capitalizing on the complementary information provided by resting and task EEGs regarding MDD-related brain abnormalities, we pooled baseline data from four distinct centers to construct brain networks differentiating Res from Non, with subsequent extraction of spatial topological features. By fusing the spatial topological features (resting and task), we classified Res versus Non and forecasted long-term medication response. Baseline features distinguished Res from Non with 96.30% accuracy and enabled reliable patient-specific prediction of eight-week outcomes. Collectively, these results indicated that resting and task EEGs provided synergistic information, enhancing both diagnostic classification and prognostic prediction. This multidimensional approach provided a promising translational framework that, with further validation, may guide personalized antidepressant selection and potentially accelerate treatment response and reduce healthcare costs.