Artificial Intelligence in Multimodal Neuroimaging Research on Depression: A Bibliometric Analysis
Open Access
- 25 October 2026
- journal article
- Published by Korean Neuropsychiatric Association in Psychiatry Investigation
- Vol. 23 (10) , 1219-1231
Abstract
Objective To conduct a systematic bibliometric review of artificial intelligence (AI)-based multimodal neuroimaging research in depression, mapping its evolution, collaborative networks, and intellectual foundations to guide future translation.Methods Records were retrieved from Web of Science Core Collection (search date: 1 September 2025) using terms for depression, neuroimaging modalities, and AI/machine learning (ML)/deep learning. English, peer-reviewed articles/reviews (2001–2025) were screened; 1,232 studies were retained after deduplication and filtering. Analyses included annual output trends, co-authorship networks (countries/institutions/authors), keyword co-occurrence clustering with burst detection, and co-citation/bibliographic coupling using VOSviewer (v1.6.20) and Bibliometrix (v4.5.1).Results Annual publications rose from <20/year to 202 in 2024; 2025 counts (n=173) reflect January–August only. The international network spans 42 countries/regions linked by 320 co-authorship ties (total link strength=1,228; density of approximately 0.37). USA, China, and the UK constitute the epistemic core, with expanding participation from India, Brazil, and Iran. Institutional and author overlays indicate a small-world, hierarchical organization (205 institutions; 10 collaboration communities; author network 168 researchers) characterized by a productive core and long tail. Six keyword clusters delineate a functional workflow: 1) biomarkers/phenotypes, 2) functional connectivity, 3) machine-learning/computational approaches, 4) multimodal/precision psychiatry, 5) diagnostic classification/algorithm development, and 6) structural imaging/comorbidity. Co-citation and coupling reveal a progression from early ML and connectivity foundations to deep/graph models, multimodal fusion, explainability, and treatment-oriented prediction.Conclusion AI-based multimodal neuroimaging in depression has evolved into a coherent, translationally focused field. Current trends emphasize model validation, interpretability, and cross-site generalization, supported by expanding international collaboration. Future priorities include harmonized data sharing, explainable modeling, and theory-informed clinical translation.Keywords
Funding Information
- Sanming Project of Medicine in Shenzen Municipality (SZZYSM202311020)
This publication has 36 references indexed in Scilit:
- MSARAE: Multiscale adversarial regularized autoencoders for cortical network classificationMedical Image Analysis, 2025
- Generalizability of Treatment Outcome Prediction Across Antidepressant Treatment Trials in DepressionJAMA Network Open, 2025
- DPD (DePression Detection) Net: a deep neural network for multimodal depression detectionHealth Information Science and Systems, 2024
- Global burden of depression or depressive symptoms in children and adolescents: A systematic review and meta-analysisJournal of Affective Disorders, 2024
- Robust discriminant feature extraction for automatic depression recognitionBiomedical Signal Processing and Control, 2023
- Computer assisted identification of stress, anxiety, depression (SAD) in students: A state-of-the-art reviewMedical Engineering & Physics, 2022
- Multiple Holdouts With Stability: Improving the Generalizability of Machine Learning Analyses of Brain–Behavior RelationshipsBiological Psychiatry, 2020
- Altered resting-state functional connectivity in late-life depression: A cross-sectional studyJournal of Affective Disorders, 2016
- A machine learning approach using EEG data to predict response to SSRI treatment for major depressive disorderClinical Neurophysiology, 2013
- Unsupervised classification of major depression using functional connectivity MRIHuman Brain Mapping, 2013