Fig. 1

Flowchart of the study. Firstly, sMRI (3D-T₁WI and DTI) was performed with two scanners in adolescent MDD subjects. FreeSurfer/ANTs hybrid segmentation toolkit of Mindboggle software, CAT 12 suite of SPM software and Diffusion Toolbox of FSL software were used for preprocessing. After that, conventional indicators and radiomics features of shape parameters of GM based on SBM and VBM analysis and diffusion properties of WM were extracted and then harmonized with ComBat technique. After receiving SSRIs or SNRIs for 2 weeks, the subjects were divided into ADM improvers (SSRIs improvers and SNRIs improvers) and non-improvers according to reduction rate of the HAM-D₁₇ score. Finally, the features were decreased and filtered based on ANOVA and RFE successively and those with high prediction power were employed to construct models based on RBF-SVM. The performance was estimated using LOO-CV and ROC curve. sMRI, structural MRI. 3D-T₁WI, three-dimensional T1 weighted imaging. DTI, diffusion tensor imaging. MDD, major depressive disorder. SPM, statistical parametric mapping. FSL, FMRIB’s software library. SBM, surface-based morphology. VBM, voxel-based morphology. GM, gray matter. WM, white matter. ANOVA, analysis of variance. RFE, recursive feature elimination. SVM, support vector mechanism. RBF, radial basis function kernel. LOO-CV, leave-one-out cross-validation. ROC, receiver operator characteristic. ADM, antidepressant medication. SSRIs, selective serotonin reuptake inhibitors. SNRIs, serotonin norepinephrine reuptake inhibitors. HAM-D₁₇, Hamilton Depression Rating Scale, 17 item