从数据整合到临床转化：机器学习在抑郁症诊断中的应用

From data integration to clinical translation: a review of machine learning applications in depression diagnosis

摘要

抑郁症起病隐匿、症状异质性明显，现行诊断仍主要依赖临床访谈与量表评估，存在主观性较强、早期识别不足以及对分型和预后判断能力有限等问题。随着神经影像、脑电图、语音与数字行为、临床量表及多组学等客观数据的不断积累，机器学习为抑郁症客观识别提供了新的研究路径。文章围绕抑郁症客观识别这一核心问题，综述了机器学习在神经影像及其他单模态客观数据中的应用进展，进一步总结了多模态数据在抑郁症诊断、分型、病程评估和疗效预测中的整合价值，并对常见融合策略、模型可解释性及临床转化问题进行了归纳分析。

Abstract

Depression is characterized by an insidious onset and marked symptom heterogeneity. Current diagnostic practice still mainly relies predominantly on clinical interviews and rating scales, which are constrained by substantial subjectivity, limited capacity for early identification, and insufficient performance in subtype differentiation and prognostic evaluation. With the growing availability of objective data, including neuroimaging, electroencephalography, speech and digital behavioral features, clinical scales, and multi-omics profiles, machine learning has opened new avenues for the objective identification of depression. Focusing on this central issue, the present review summarizes recent advances in the application of machine learning to neuroimaging and other unimodal objective data, further discusses the integrative value of multimodal data in the diagnosis, subtyping, disease-course assessment, and treatment-response prediction of depression, and provides a systematic overview of common fusion strategies, model interpretability, and issues related to clinical translation.

关键词

抑郁症 / 机器学习 / 神经影像学 / 多模态数据 / 辅助诊断

Key words

depression / machine learning / neuroimaging / multimodal data / auxiliary diagnosis

参考文献

[1] 王颖, 侯宇威, 张哲. 中国人群抑郁症疾病负担的年龄—时期—队列分析[J]. 四川精神卫生, 2025, 38(3):254-260.

[2] WANG Y, HOU Y W, ZHANG Z. Age—period—cohort analysis of the disease burden of depression in the Chinese population[J]. Sichuan Journal of Mental Health, 2025, 38(3):254-260.

[3] LIN S K, KUO P H, HSU C Y, et al. The effects of Lactobacillus plantarum PS128 in patients with major depressive disorder: an eight—week double—blind, placebo—controlled study[J]. Asian J Psychiatr, 2024, 101:104210.

[4] XU R F, LIU Z J, OUYANG S, et al. Machine learning—driven development of a stratified CES—D screening system: optimizing depression assessment through adaptive item selection[J]. BMC Psychiatry, 2025, 25(1):286.

[5] GAN X, LI X, CAI Y, et al. Metabolic features of adolescent major depressive disorder: a comparative study between treatment—resistant depression and first—episode drug—naive depression[J]. Psychoneuroendocrinology, 2024, 167:107086.

[6] BATAIL J M, COROUGE I, BLANCHARD T, et al. Inflammatory and MRI perfusion biomarkers in predicting persistence of depression: a 6—month longitudinal study[J]. Transl Psychiatry, 2025, 15(1):370.

[7] YIN B, CAI Y, TENG T, et al. Identifying plasma metabolic characteristics of major depressive disorder, bipolar disorder, and schizophrenia in adolescents[J]. Transl Psychiatry, 2024, 14(1):163.

[8] ENKHBAYAR D, KO J, OH S, et al. Explainable artificial intelligence models for predicting depression based on polysomnographic phenotypes[J]. Bioengineering, 2025, 12(2):186.

[9] HU Y X, SHI J Y, XIA G Y, et al. Analysis of functional connectivity changes in attention networks and default mode networks in patients with depression and insomnia[J]. Sleep Breath, 2024, 28(4):1731-1742.

[10] GRIMM B, YILMAM P, TALBOT B, et al. Multimodal machine learning for video based single question mental health assessment[J]. NPJ Digit Med, 2025, 9(1):65.

[11] CLAAS F, MICAH C, NILS O, et al. Systematic misestimation of machine learning performance in neuroimaging studies of depression[J]. Neuropsychopharmacology, 2021, 46(8):1510-1517.

[12] GUO Y, CHU T, LI Q, et al. Diagnosis of major depressive disorder based on individualized brain functional and structural connectivity[J]. J Magn Reson Imaging, 2025, 61(4):1712-1725.

[13] DAI P, HUANG K, HU T, et al. Altered effective connectivity in patients with drug—naïve first—episode, recurrent, and medicated major depressive disorder: a multi—site fMRI study[J]. Behav Brain Res, 2025, 495:115756.

[14] GUO P, LIU X, SUN Y, et al. Effects of sMRI—guided rTMS on brain function and structure in major depressive disorder[J]. J Affect Disord, 2025, 390:119772.

[15] VU T, DAWADI R, YAMAMOTO M, et al. Prediction of depressive disorder using machine learning approaches: findings from the NHANES[J]. BMC medical informatics and decision making, 2025, 25(1):83.

[16] LUO Y, SHEN Y, FAN X. EEG microstates in adolescent depression: effects of depression severity and overall symptoms[J]. J Affect Disord, 2025, 390:119819.

[17] MULC D, VUKOJEVIC J, KALAFATIC E, et al. Opportunities and challenges for clinical practice in detecting depression using EEG and machine learning[J]. Sensors, 2025, 25(2): 409-409.

[18] YANG CY, CHEN Y Z. Support vector machine classification of patients with depression based on resting—state electroencephalography[J]. Asian Biomed (Res Rev News), 2024, 18(5):212-223.

[19] DENIER N, WALTHER S, BREIT S, et al. Electroconvulsive therapy induces remodeling of hippocampal co—activation with the default mode network in patients with depression[J]. Neuroimage Clin, 2023, 38:103404.

[20] ALOTAIBI N M, BAKHEET D M. Predicting depression severity using effective and functional brain connectivity of the electroencephalography signals[J]. Comput Biol Med, 2025, 190:110045.

[21] RAVAN M, NOROOZI A, GEDIYA H, et al. Using deep learning and pretreatment EEG to predict response to sertraline, bupropion, and placebo[J]. Clin Neurophysiol, 2024, 167:198-208.

[22] BAUER J F, GERCZUK M, SCHINDLER—GMELCH L, et al. Validation of machine learning—based assessment of major depressive disorder from paralinguistic speech characteristics in routine care[J]. Depress Anxiety, 2024, 2024(1):9667377.

[23] IKÄHEIMONEN A, LUONG N, BARYSHNIKOV I, et al. Predicting and monitoring symptoms in patients diagnosed with depression using smartphone data: observational study[J]. J Med Internet Res, 2024, 26:e56874.

[24] EDER J, PFEIFFER L, WICHERT P S, et al. Deconstructing Depression by machine learning: the POKAL—PSY study[J]. Eur Arch Psychiatry Clin Neurosci, 2023, 274(5):1153-1165.

[25] ALOTAIBI N M, ALHOTHALI A M, ALI M S. Multi—atlas ensemble graph neural network model for major depressive disorder detection using functional MRI data[J]. Front Comput Neurosci, 2025, 19:1537284.

[26] SAMPRIT B, YIYUAN W, S K B, et al. Trajectories of remitted psychotic depression: identification of predictors of worsening by machine learning[J]. Psychol med, 2024, 54(6):1142-1151.

[27] MARDINI M T, KHALIL G E, BAI C, et al. Identifying adolescent depression and anxiety through real—world data and social determinants of health: machine learning model development and validation[J]. JMIR Ment Health, 2025, 12:e66665.

[28] MA S, NIE Z, ZHANG M, et al. Towards precision psychiatry: metabolomics identifies three biological subtypes of depression[J]. PLoS Digit Health, 2025, 4(12):e0001125.

[29] TOZZI L, ZHANG X, PINES A, et al. Personalized brain circuit scores identify clinically distinct biotypes in depression and anxiety[J]. Nat Med, 2024, 30(7):2076-2087.

[30] JIAO Y, ZHAO K, WEI X, et al. Deep graph learning of multimodal brain networks defines treatment—predictive signatures in major depression[J]. Mol Psychiatry, 2025, 30(9):3963-3974.

[31] HAN W, WU X, WANG L, et al. Altered brain function in treatment—resistant depression patients: a resting—state functional magnetic resonance imaging study[J]. Neurosci Lett, 2024, 842:138004.

[32] LEE D Y, KIM N, PARK C, et al. Explainable multimodal prediction of treatment—resistance in patients with depression leveraging brain morphometry and natural language processing[J]. Psychiatry Res, 2024, 334:115817.

[33] QAYYUM A, RAZZAK I, TANVEER M, et al. High—density electroencephalography and speech signal based deep framework for clinical depression diagnosis[J]. IEEE/ACM Trans Comput Biol Bioinform, 2023, 20(4):2587-2597.

[34] WEBER J, WEBER M, LOPEZ A. Depression diagnosis from patient interviews using multimodal machine learning[J]. Front Psychiatry, 2025, 16:1694762.

[35] KIM J, MA S P, CHEN M L, et al. Optimizing large language models for detecting symptoms of depression/anxiety in chronic diseases patient communications[J]. NPJ Digit Med, 2025, 8(1):580.

[36] YANG W, WANG X, KANG C, et al. Establishment of a risk prediction model for suicide attempts in first—episode and drug naïve patients with major depressive disorder[J]. Asian J Psychiatr, 2023, 88:103732.

[37] SCHNELLBÄCHER G J, RAJKUMAR R, VESELINOVI C T, et al. Structural alterations as a predictor of depression—a7—Tesla MRI—based multidimensional approach[J]. Mol Psychiatry, 2025, 30(6):2517-2524.

[38] COLOMBO F, CALESELLA F, BRAVI B, et al. Multimodal brain—derived subtypes of major depressive disorder differentiate patients for anergic symptoms, immune—inflammatory markers, history of childhood trauma and treatment—resistance[J]. Eur Neuropsychopharmacol, 2024, 85:45-57.

[39] LIU R, HOU X, LIU S, et al. Predicting antidepressant response via local—global graph neural network and neuroimaging biomarkers[J]. NPJ Digit Med, 2025, 8(1):515.

[40] MAZURKA R, CUNNINGHAM S, HASSEL S, et al. Relation of hippocampal volume and SGK1 gene expression to treatment remission in major depression is moderated by childhood maltreatment: a CAN—BIND—1 report[J]. Eur Neuropsychopharmacol, 2024, 78:71-80.

[41] WANG M, WANG H, FENG Z, et al. Predicting depression among chinese patients with narcolepsy type 1: a machine—learning approach[J]. Nat Sci Sleep, 2024, 16:1419-1429.

基金资助

国家自然科学基金(82574736)

国家级大学生创新训练计划项目(202510315005)