AI-Enabled Diagnostic Systems for Early Identification of Schizophrenia: Trends and Challenges

Abstract

Schizophrenia is a severe psychiatric disorder marked by disturbances in thought, perception, and behavior, resulting in long-term functional impairment and a substantial societal burden. Early identification is essential, as delayed diagnosis is highly associated with poorer prognosis, increased relapse risk, and prolonged untreated illness. Conventional diagnostic practice depends largely on clinical interviews and behavioral assessment, which are subjective, resource intensive, and limited in identifying early or prodromal stages. This research presents a structured review of recent AI-driven approaches for schizophrenia detection, encompassing machine learning, deep learning, and hybrid learning strategies. These methods have been applied to a wide range of data sources, including biological markers, behavioral traits, physiological recordings and medical imaging data. The reviewed studies report consistently high discriminatory performance across multiple data sources, demonstrating the potential to complement traditional clinical evaluation. However, common limitations were observed, including small and homogeneous cohorts, absence of longitudinal follow-up, sensitivity to preprocessing choices, limited transparency, and weak cross-site validation. Variability in data acquisition protocols and inconsistent reporting of clinical factors such as medication status and illness duration further restrict reproducibility and clinical translation. Additionally, many studies focus only on binary classification, ignoring symptom severity, disorder subtypes, and relapse patterns that are important in clinical practice. The absence of common benchmarks and shared evaluation protocols also limits fair comparison and wider adoption. By consolidating current findings and highlighting unresolved challenges, this work outlines key directions for developing reliable, interpretable, and scalable systems suitable for real clinical settings.