A bibliometric analysis of artificial intelligence applications to schizophrenia research (2005 - 2025).

Abstract

Artificial intelligence (AI) has been increasingly applied in schizophrenia research, yet a bibliometric overview of its global development and evolutionary trends remains lacking. Based on the Web of Science Core Collection (WoSCC), a bibliometric analysis was conducted on AI‑related schizophrenia studies published from 2005 to 2025. A total of 1,839 publications were included, and trends, countries, institutions, authors, journals, and research hotspots were analysed. Annual publications grew rapidly from 2 in 2005 to 374 in 2025, consistent with the expansion of AI research. The U.S., China, and England were the most productive and influential countries. The U.S. dominated global collaboration, while Chinese institutions showed strong domestic links but limited international partnerships. Leading institutions included the University of London, King's College London, and Harvard University. Nikolaos Koutsouleris and Vince D. Calhoun were the top authors. Key journals included Frontiers in Psychiatry, Schizophrenia Research, and Schizophrenia Bulletin, with NeuroImage as the top co-cited journal. Keyword analysis showed early focus on machine learning, risk assessment, and brain biomarkers based on functional magnetic resonance imaging (fMRI) and electroencephalogram (EEG). After 2016, deep learning became dominant in diagnosis. Recently, natural language processing has emerged for risk prediction, indicating a shift toward multimodal AI applications. AI applications in schizophrenia have expanded exponentially with a clear evolution from machine learning to deep learning. The global collaboration pattern is imbalanced, and future research will advance toward multi‑modal, predictive, and precise intelligence.