Artificial intelligence technologies in psychiatry: assessment of potential and barriers in the diagnosis of depressive states. A systematic review

Introduction. In the context of increasing levels of anxiety, the introduction of digital technologies with an emphasis on artificial intelligence as key tools to improve the situation by increasing the effectiveness of diagnosis and control of treatment of mental illnesses is of particular importance. The purpose of the study is to systematize data on the current state, prospects and limitations of the applicability of AI systems in the diagnosis of mental disorders, primarily depressive states, and to assess the maturity of existing technologies in order to implement them in real clinical practice. Materials and methods. The literature was searched in the PubMed (MEDLINE) database. The systematic review includes 74 original articles published in English between January 2021 and March 2026. Thematic blocks of search queries were used to identify publications exploring the possibilities of diagnosing depression using speech, voice, acoustics, facial expressions, and computer vision. The articles were subjected to a multi-stage selection: elimination of repetitions, preliminary rapid analysis by titles and annotations, and a detailed study of selected articles. Results. The conducted research demonstrates the prospects of using AI, especially multimodal approaches, in the field of detecting mental disorders, including depressive states. At the same time, a number of critical limitations have been identified that hinder the widespread clinical implementation of this technology. These include: the lack of research performed in real-world clinical practice, the potential bias of algorithms against different gender, age and ethnic groups, the difficulty in interpreting AI recommendations, and the lack of direct comparisons with the professional assessment of qualified specialists. It is also necessary to take into account the geographical concentration of research in the Asian region, which underlines the importance of conducting further research with more representative samples in order to ensure generalizability of the results. Conclusion. A promising direction is the development of hybrid models that integrate AI capabilities with the experience and knowledge of clinicians. Successful solution of these tasks will allow us to realize the potential of AI for early detection of diseases, personalization of therapy and improvement of the quality of life of patients with depression.