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

Area of Science:

• Neuroscience
• Psychiatry
• Artificial Intelligence

Background:

• Schizophrenia presents diverse symptoms and challenges in diagnosis and treatment due to its complexity.
• Data-driven subtyping using neuroimaging and machine learning (ML) offers a path toward individualized patient care.
• Recent advancements enable the categorization of schizophrenia into distinct subtypes.

Purpose of the Study:

• To systematically review studies utilizing machine learning and neuroimaging for schizophrenia subtyping.
• To identify neurobiological markers differentiating schizophrenia subtypes.
• To assess the potential for precision psychiatry in schizophrenia treatment.

Main Methods:

• A systematic literature review adhering to PRISMA guidelines was conducted.
• Searches encompassed studies from January 2019 to September 2024 in PubMed, Web of Science, and Scopus.
• Included studies employed ML techniques (e.g., SVM, clustering) with neuroimaging (e.g., MRI, fMRI, DTI) for schizophrenia subtyping.

Main Results:

• Machine learning-based subtyping revealed distinct neuroanatomical, functional, and inflammatory subgroups of schizophrenia.
• Subtypes were characterized by variations in inflammatory markers, functional connectivity, and gray matter volume.
• These identified subtypes correlated with symptom severity and treatment response, underscoring schizophrenia's heterogeneity.

Conclusions:

• ML-based subtyping illuminates diverse neurobiological manifestations of schizophrenia, crucial for precision psychiatry.
• These findings support individualized treatment strategies for schizophrenia.
• Future research should address methodological limitations through multimodal imaging in larger, multi-site, longitudinal studies to enhance clinical applicability.