Abstract Background Artificial intelligence (AI) and machine learning (ML) offer scalable screening, objective monitoring, and precision treatment for Major Depressive Disorder (MDD), a leading cause of global disability and mortality, by integrating multimodal data. Objective To synthesize and critically appraise studies that applied AI/ML methods to diagnose, monitor, predict treatment response, or discover biomarkers for MDD. Methods A systematic search of the Scopus database was conducted to identify studies applying AI/ML in MDD. Due to heterogeneity in study designs, data modalities, AI models, and reported performance metrics, the results were narratively synthesized. Results Classical ML models (e.g., Support Vector Machines, Random Forest) were typically applied to smaller or interpretable feature sets, while deep learning architectures dominated high-dimensional Electroencephalogram (EEG), neuroimaging, and multimodal data, often reporting high internal performance. Diagnostic studies particularly EEG, Natural Language Processing, and single-site imaging–based frequently achieved accuracies > 90% or Area Under Curves (AUCs) > 0.85, whereas larger multisite and prognostic models showed more modest but more generalizable performance (AUC ~ 0.70–0.85); despite gains from multimodal fusion and large language models-based approaches, limited external validation remains a major barrier to clinical translation. Conclusions AI/ML could transform MDD care, but realizing this potential requires rigorous validation, ethical safeguards, and careful handling of data. Emerging methods like multimodal fusion, federated learning, and large language models can support personalized and scalable care.
Okesanya et al. (Tue,) studied this question.