A scientometric analysis of machine learning in schizophrenia neuroimaging: Trends and insights (2012-2024)

Machine learning applications in schizophrenia neuroimaging research have undergone significant evolution since 2012. However, a comprehensive scientometric analysis of this field has not yet been conducted. This study analyzed 315 original research articles, accumulating 10,181 citations, from the Web of Science Core Collection (WOSCC) (2012-2024) using R-bibliometrix and CiteSpace. The annual publication output showed steady growth, reaching a peak of 57 publications in 2021, followed by a gradual decline. Network analysis revealed a thematic evolution from conventional machine learning applications toward explainable artificial intelligence, and a methodlogical shift from single-modality imaging to integrated multimodal approaches (modularity Q = 0.7536, silhouette S = 0.9039). China led in publication count (115), whereas the United States had the highest citation impact (4414 citations). Recent research clusters (2021-2024) highlighted areas such as risk assessment, brain age prediction, and treatment response monitoring. These findings provide strategic guidance for future research directions in neuroimaging-based machine learning applications for schizophrenia, emphasizing the critical need for methodological standardization to enhance clinical application.