Artificial Intelligence in Schizophrenia Spectrum Disorders: Current Use and Future Perspectives. A Systematic Mapping Literature Review

Background: Artificial intelligence (AI) techniques are increasingly applied in psychiatric research, yet their use in schizophrenia spectrum disorders (SSDs) remains heterogeneous. This systematic mapping literature review aimed to examine the current applications of AI systems in SSDs. AI approaches were categorized into natural language processing (NLP), machine learning (ML), and deep learning (DL). Key study characteristics were systematically analyzed, including publication year, type of AI algorithm, study population, clinical outcomes, and instrumental techniques. Trends in journal impact factor (IF) were also evaluated. Methods: A systematic mapping approach was employed to identify and categorize relevant studies investigating AI applications in SSDs. Included studies were analyzed according to predefined evaluation criteria to provide an overview of methodological trends and research focus within the field. Results: A total of 853 studies met the inclusion criteria. Machine Learning emerged as the most frequently utilized AI approach. AI methods were predominantly applied to diagnostic and differential diagnostic tasks in SSDs. The most commonly employed instrumental techniques were magnetic resonance imaging (MRI) and electroencephalography (EEG). Journal Impact Factor varied significantly across study characteristics, with higher IFs observed in studies focusing specifically on schizophrenia, reporting diagnostic outcomes, or employing MRI or EEG-based measures. Conclusions: This review represents the first systematic mapping literature review to comprehensively examine AI applications in schizophrenia spectrum disorders. The findings may support clinicians and researchers in planning, implementing, and evaluating AI-based methodologies in SSD research and clinical contexts. Nevertheless, further studies are needed to establish the clinical validity, robustness, and translational applicability of these algorithms.