Advances in Schizophrenia Subtyping: A Systematic Review of Machine Learning Applications in Neuroimaging

Schizophrenia is a disorder with a wide range of symptoms, elusive etiology, and high variability. Consequently, diagnosis and treatment can be challenging. Individualized care is possible by data-driven subtyping made by recent developments in neuroimaging and machine learning (ML).

Following the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines, a systematic evaluation of articles published between January 2019 and September 2024 was conducted. Studies employing machine learning techniques (e.g., Support Vector Machine (SVM), clustering) in conjunction with neuroimaging techniques (e.g., MRI, fMRI, diffusion tensor imaging (DTI)) to categorize schizophrenia subtypes were identified through searches in PubMed, Web of Science, and Scopus databases. The inclusion criteria were met by 18 studies.

Different neuroanatomical, functional, and inflammatory subgroups of schizophrenia were found using machine learning-based subtyping. Studies revealed different subtypes distinguished by variations in inflammatory markers, functional connectivity, and gray matter volume. These subtypes highlighted the disorder's variability and were linked to various levels of symptom severity and treatment responsiveness. The lack of longitudinal data, methodological heterogeneity, and small sample sizes were among the limitations impeding the generalizability of the above findings.

Different neurobiological manifestations of schizophrenia are revealed by ML-based subtyping, providing valuable information for precision psychiatry and individualized treatment. These findings will be more solid and clinically applicable upon resolution of methodological issues via multimodal imaging in larger, multi-site, longitudinal studies.