Share:
Share this content in WeChat
X
• Clinical Article •
An interpretable machine-learning model based on structural and metabolic features of the cortical-striatal-cerebellar circuit for predicting recurrence risk after recovery in adolescents with first-episode major depressive disorder
WANG Jinxin  WANG Ranchao  GE Jiyuan  LI Yuefeng  CAO Zhihong  LUO Yifeng  LIU Yu 

DOI:10.12015/issn.1674-8034.2026.11.005.


[Abstract] Objective To evaluate an interpretable machine learning model based on structural and metabolic features of the cortical-striatal-cerebellar (CSC) circuit for predicting recurrence after recovery in adolescents with first-episode major depressive disorder (MDD).Materials and Methods In this prospective study, 200 adolescents with first-episode MDD who had achieved recovery were consecutively enrolled. After follow-up and imaging quality control, 182 were included in the final analysis. The follow-up period was 24 months. Regional brain volumes within the CSC circuit were derived from structural MRI, and metabolite measures were obtained using magnetic resonance spectroscopy. Key imaging features were selected in the training set using univariable Cox regression followed by LASSO-Cox regression. Cox, CoxBoost, and random survival forest (RSF) models were constructed. Model performance was evaluated in the test set using the concordance index, time-dependent area under the receiver operating characteristic curve, Brier score, and area under the precision-recall curve. Shapley additive explanations (SHAP) were used to interpret the model predictions.Results Six key features were retained: L-ACC-Glx, L-ACC-NAA, R-ACC-mI, R-BG-sV, L-Tha-sV, and L-Tha-Cho. The RSF model showed comparatively favorable overall performance in the test set, with a concordance index of 0.768. The time-dependent areas under the receiver operating characteristic curve at 6, 12, and 24 months were 0.790 [95% confidence interval (CI): 0.695 to 0.873], 0.775 (95% CI: 0.674 to 0.859), and 0.742 (95% CI: 0.636 to 0.833), respectively, and the corresponding Brier scores were 0.119, 0.158, and 0.207. SHAP analysis showed that lower L-ACC-Glx, L-ACC-NAA, L-Tha-sV, and R-BG-sV, together with higher R-ACC-mI and L-Tha-Cho, were associated with a higher risk of recurrence.Conclusions Multimodal imaging features of the CSC circuit may provide objective imaging information for stratifying recurrence risk in adolescents who have recovered from a first episode of MDD. The interpretable RSF-based survival model may support individualized follow-up during recovery.
[Keywords] cortical-striatal-cerebellar circuit;adolescents;first-episode major depressive disorder;magnetic resonance spectroscopy;random survival forest;survival analysis;interpretable machine learning

WANG Jinxin1   WANG Ranchao2   GE Jiyuan1   LI Yuefeng2   CAO Zhihong1   LUO Yifeng1   LIU Yu1*  

1 Department of Radiology, Yixing Hospital Affiliated to Jiangsu University, Yixing 214200, China

2 Department of Radiology, Affiliated People's Hospital of Jiangsu University, Zhenjiang 212002, China

Corresponding author: LIU Y, E-mail: liuyum0526@163.com

Conflicts of interest   None.

Received  2026-06-05
Accepted  2026-10-15
DOI: 10.12015/issn.1674-8034.2026.11.005
DOI:10.12015/issn.1674-8034.2026.11.005.

[1]
GBD 2019 MENTAL DISORDERS COLLABORATORS. Global, regional, and national burden of 12 mental disorders in 204 countries and territories, 1990-2019: a systematic analysis for the Global Burden of Disease Study 2019[J]. Lancet Psychiatry, 2022, 9(2): 137-150. DOI: 10.1016/S2215-0366(21)00395-3.
[2]
LU B Q, LIN L X, SU X J. Global burden of depression or depressive symptoms in children and adolescents: a systematic review and meta-analysis[J/OL]. J Affect Disord, 2024, 354: 553-562 [2026-06-04]. https://pubmed.ncbi.nlm.nih.gov/38490591/. DOI: 10.1016/j.jad.2024.03.074.
[3]
DESAI BOSTRÖM A E, CARS T, HELLNER C, et al. Recovery and recurrence from major depression in adolescence and adulthood[J]. Acta Psychiatr Scand, 2025, 151(5): 625-633. DOI: 10.1111/acps.13785.
[4]
TAGLIAFERRI S D, HAN L K M, KHETAN M, et al. Systematic review and meta-analysis: predictors of relapsing, recurrent, and chronic depression in young people[J]. J Am Acad Child Adolesc Psychiatry, 2026, 65(2): 206-230. DOI: 10.1016/j.jaac.2025.03.019.
[5]
WANG X Q, YANG K S, WANG M Y, et al. Research progress of magnetic resonance imaging in evaluating the efficacy of major depressive disorder[J]. Chin J CT MRI, 2025, 23(5): 191-194. DOI: 10.3969/j.issn.1672-5131.2025.05.056.
[6]
SONG X D, YAN S Y, LAI S K, et al. Gender differences of neurometabolic and neuroendocrine alternations and its lateralization in adolescents with major depressive disorder[J/OL]. BMC Psychiatry, 2024, 24(1): 949 [2026-06-04]. https://pubmed.ncbi.nlm.nih.gov/39731037/. DOI: 10.1186/s12888-024-06428-1.
[7]
MORIGUCHI S, TAKAMIYA A, NODA Y, et al. Glutamatergic neurometabolite levels in major depressive disorder: a systematic review and meta-analysis of proton magnetic resonance spectroscopy studies[J]. Mol Psychiatry, 2019, 24(7): 952-964. DOI: 10.1038/s41380-018-0252-9.
[8]
GRAY J P, MÜLLER V I, EICKHOFF S B, et al. Multimodal abnormalities of brain structure and function in major depressive disorder: a meta-analysis of neuroimaging studies[J]. Am J Psychiatry, 2020, 177(5): 422-434. DOI: 10.1176/appi.ajp.2019.19050560.
[9]
WU B L, ZHANG X, XIE H S, et al. Research Review: Shared and distinct structural and functional brain alterations in adolescents with major depressive disorder–a multimodal meta-analysis[J]. J Child Psychol Psychiatry, 2025, 66(7): 1084-1097. DOI: 10.1111/jcpp.14104.
[10]
WANG Y H, WEI J X, YAN Y S, et al. Disruption of multimodal brain networks and structural-functional coupling in adolescents with major depressive disorder[J/OL]. Neuropsychiatr Dis Treat, 2025, 21: 791-798 [2026-06-04]. https://pubmed.ncbi.nlm.nih.gov/40230783/. DOI: 10.2147/NDT.S515540.
[11]
LAI S K, ZHONG S M, SHAN Y Y, et al. Altered biochemical metabolism and its lateralization in the cortico-striato-cerebellar circuit of unmedicated bipolar II depression[J/OL]. J Affect Disord, 2019, 259: 82-90 [2026-06-04]. https://pubmed.ncbi.nlm.nih.gov/31442883/. DOI: 10.1016/j.jad.2019.07.021.
[12]
HOU X M, LIU R, ZHOU Y, et al. Shared and unique alterations of large-scale network connectivity in drug-free adolescent-onset and adult-onset major depressive disorder[J/OL]. Transl Psychiatry, 2024, 14(1): 255 [2026-06-04]. https://pubmed.ncbi.nlm.nih.gov/38866779/. DOI: 10.1038/s41398-024-02974-0.
[13]
WU R, LING A N, ZHANG G F. Advances in resting-state fMRI studies of depression in children and adolescents[J]. Chin J Magn Reson Imaging, 2024, 15(6): 149-152, 195. DOI: 10.12015/issn.1674-8034.2024.06.023.
[14]
MORIARTY A S, MEADER N, SNELL K I E, et al. Predicting relapse or recurrence of depression: systematic review of prognostic models[J]. Br J Psychiatry, 2022, 221(2): 448-458. DOI: 10.1192/bjp.2021.218.
[15]
KENNIS M, GERRITSEN L, VAN DALEN M, et al. Prospective biomarkers of major depressive disorder: a systematic review and meta-analysis[J]. Mol Psychiatry, 2020, 25(2): 321-338. DOI: 10.1038/s41380-019-0585-z.
[16]
LAM R W, RNIC K, NUNEZ J J, et al. Predicting Relapse of Depressive Episodes During Maintenance Treatment: The Canadian Biomarker Integration Network in Depression (CAN-BIND) Wellness Monitoring in Major Depressive Disorder Study: Prédire la rechute d'épisodes dépressifs pendant le traitement d'entretien: Une étude de suivi du Bien-être dans les troubles dépressifs majeurs du Réseau canadien d'intégration des biomarqueurs pour la dépression (CAN-BIND)[J]. Can J Psychiatry, 2025, 70(7): 565-573. DOI: 10.1177/07067437251337603.
[17]
SCHOENFELD D A. Sample-size formula for the proportional-hazards regression model[J]. Biometrics, 1983, 39(2): 499-503.
[18]
RAO U, CHEN L A. Characteristics, correlates, and outcomes of childhood and adolescent depressive disorders[J]. Dialogues Clin Neurosci, 2009, 11(1): 45-62. DOI: 10.31887/DCNS.2009.11.1/urao.
[19]
KENNARD B, SILVA S, VITIELLO B, et al. Remission and residual symptoms after short-term treatment in the Treatment of Adolescents with Depression Study (TADS)[J]. J Am Acad Child Adolesc Psychiatry, 2006, 45(12): 1404-1411. DOI: 10.1097/01.chi.0000242228.75516.21.
[20]
CURRY J, SILVA S, ROHDE P, et al. Recovery and recurrence following treatment for adolescent major depression[J]. Arch Gen Psychiatry, 2011, 68(3): 263-269. DOI: 10.1001/archgenpsychiatry.2010.150.
[21]
LIN A, ANDRONESI O, BOGNER W, et al. Minimum Reporting Standards for in vivo Magnetic Resonance Spectroscopy (MRSinMRS): Experts' consensus recommendations[J/OL]. NMR Biomed, 2021, 34(5): e4484 [2026-06-04]. https://pubmed.ncbi.nlm.nih.gov/33559967/. DOI: 10.1002/nbm.4484.
[22]
FAN H, YANG Z X, YU X Y, et al. Interpretable machine learning model to predict aggressive behavior in first-episode schizophrenia[J/OL]. J Psychiatr Res, 2025, 191: 349-355 [2026-06-04]. https://pubmed.ncbi.nlm.nih.gov/41045766/. DOI: 10.1016/j.jpsychires.2025.09.065.
[23]
BINDER H, SCHUMACHER M. Allowing for mandatory covariates in boosting estimation of sparse high-dimensional survival models[J/OL]. BMC Bioinform, 2008, 9(1): 14 [2026-06-04]. https://pubmed.ncbi.nlm.nih.gov/18186927/. DOI: 10.1186/1471-2105-9-14.
[24]
ISHWARAN H, KOGALUR U B, BLACKSTONE E H, et al. Random survival forests[J]. Ann Appl Stat, 2008, 2(3): 841-860. DOI: 10.1214/08-aoas169.
[25]
CHEN Y J, ZHAO W, YI S J, et al. The diagnostic performance of machine learning based on resting-state functional magnetic resonance imaging data for major depressive disorders: a systematic review and meta-analysis[J/OL]. Front Neurosci, 2023, 17: 1174080 [2026-06-04]. https://pubmed.ncbi.nlm.nih.gov/37811326/. DOI: 10.3389/fnins.2023.1174080.
[26]
BELOV V, ERWIN-GRABNER T, AGHAJANI M, et al. Multi-site benchmark classification of major depressive disorder using machine learning on cortical and subcortical measures[J/OL]. Sci Rep, 2024, 14(1): 1084 [2026-06-04]. https://pubmed.ncbi.nlm.nih.gov/38212349/. DOI: 10.1038/s41598-023-47934-8.
[27]
LUNDBERG S M, ERION G, CHEN H, et al. From local explanations to global understanding with explainable AI for trees[J]. Nat Mach Intell, 2020, 2(1): 56-67. DOI: 10.1038/s42256-019-0138-9.
[28]
WANG R C, WANG J X, LI Y, et al. Value of hippocampal morphological and metabolic parameters in predicting patients with non-stroke atrial fibrillation with mild cognitive impairment[J]. Chin J Radiol, 2026, 60(4): 404-412. DOI: 10.3760/cma.j.cn112149-20250410-00184.
[29]
BHATTARAI P, THAKURI D S, NIE Y Z, et al. Explainable AI-based Deep-SHAP for mapping the multivariate relationships between regional neuroimaging biomarkers and cognition[J/OL]. Eur J Radiol, 2024, 174: 111403 [2026-06-04]. https://pubmed.ncbi.nlm.nih.gov/38452732/. DOI: 10.1016/j.ejrad.2024.111403.
[30]
WU B L, ZHANG X, XIE H S, et al. Disrupted structural brain networks and structural-functional decoupling in first-episode drug-Naïve adolescent major depressive disorder[J]. J Adolesc Health, 2024, 74(5): 941-949. DOI: 10.1016/j.jadohealth.2024.01.015.
[31]
DING H, ZHANG Q, SHU Y P, et al. Vulnerable brain regions in adolescent major depressive disorder: a resting-state functional magnetic resonance imaging activation likelihood estimation meta-analysis[J]. World J Psychiatry, 2024, 14(3): 456-466. DOI: 10.5498/wjp.v14.i3.456.
[32]
KIM H, PARK J, KUHN M, et al. Neuroticism modulates functional connectivity of the midcingulate cortex during emotional conflict[J/OL]. Sci Rep, 2025, 15: 13095 [2026-06-04]. https://pubmed.ncbi.nlm.nih.gov/40240799/. DOI: 10.1038/s41598-025-97529-8.
[33]
CIPOLOTTI L, MOLE J, RUFFLE J K, et al. Cognitive control & the anterior cingulate cortex: Necessity & coherence[J/OL]. Cortex, 2025, 182: 87-99 [2026-06-04]. https://pubmed.ncbi.nlm.nih.gov/39645441/. DOI: 10.1016/j.cortex.2024.11.010.
[34]
TAO S W, DENG R H, WEI M H, et al. A meta-analysis of magnetic resonance spectroscopy studies on glutamatergic neurometabolite levels in major depressive disorder[J/OL]. Depress Anxiety, 2025, 2025: 5180077 [2026-06-04]. https://pubmed.ncbi.nlm.nih.gov/40778365/. DOI: 10.1155/da/5180077.
[35]
MA L Y, et al. γ-Aminobutyric acid and glutamate dysregulation in the dorsolateral prefrontal cortex of adolescents with first-episode major depressive disorder and the modulatory effects of repetitive transcranial magnetic stimulation[J/OL]. Psychoradiology, 2026, 6: kkaf034 [2026-06-04]. https://academic.oup.com/psyrad/article/doi/10.1093/psyrad/kkaf034/8329172. DOI: 10.1093/psyrad/kkaf034.
[36]
SACCARO L F, TASSONE M, TOZZI F, et al. Proton magnetic resonance spectroscopy of N-acetyl aspartate in first depressive episode and chronic major depressive disorder: a systematic review and meta-analysis[J/OL]. J Affect Disord, 2024, 355: 265-282 [2026-06-04]. https://pubmed.ncbi.nlm.nih.gov/38554884/. DOI: 10.1016/j.jad.2024.03.150.
[37]
ÖZ G, ALGER J R, BARKER P B, et al. Clinical proton MR spectroscopy in central nervous system disorders[J]. Radiology, 2014, 270(3): 658-679. DOI: 10.1148/radiol.13130531.
[38]
ZHAO X H, WU S Y, LI X, et al. Common neural deficits across reward functions in major depression: a meta-analysis of fMRI studies[J]. Psychol Med, 2024, 54(11): 2794-2806. DOI: 10.1017/S0033291724001235.
[39]
HAGEN J, RAMKIRAN S, SCHNELLBÄCHER G J, et al. Phenomena of hypo- and hyperconnectivity in basal ganglia-thalamo-cortical circuits linked to major depression: a 7T fMRI study[J]. Mol Psychiatry, 2025, 30(1): 158-167. DOI: 10.1038/s41380-024-02669-4.
[40]
ZHAO X, YIN Q R, ZHAO W, et al. Abnormal functional connectivity and structurefunction coupling of the nucleus accumbens in patients with major depressive disorder[J]. J Cent South Univ Med Sci, 2025(9): 1579-1589. DOI: 10.11817/j.issn.1672-7347.2025.250392.
[41]
MILLER C H, HAMILTON J P, SACCHET M D, et al. Meta-analysis of functional neuroimaging of major depressive disorder in youth[J/OL]. JAMA Psychiatry, 2015, 72(10): 1045 [2026-06-04]. https://pubmed.ncbi.nlm.nih.gov/26332700/. DOI: 10.1001/jamapsychiatry.2015.1376.

PREV Research on autism subtypes based on individualized extreme functional connectivity
NEXT A resting-state functional magnetic resonance imaging study of the relationship between cortical abnormal amplitude of low-frequency fluctuation and cognitive impairment in patients with vascular cognitive impairment
  



Tel & Fax: +8610-67113815    E-mail: editor@cjmri.cn