Gradient boosting decision-tree-based algorithm with neuroimaging for personalized treatment in depression

•Pretreatment imaging can predict remission with 77% weighted accuracy.•The predictive performance does not differ by sex or treatment assignment.•Pretreatment metabolism of left hippocampus is the most predictive of remission.•Outlier removal improves model performance for predicting remission.•Synthetic data generation is an effective way to address class imbalance.