Two distinct neuroanatomical subtypes of schizophrenia revealed using machine learning

Area of Science:

• Neuroimaging
• Psychiatry
• Machine Learning

Background:

• Schizophrenia exhibits significant neurobiological heterogeneity, complicating research and treatment.
• Understanding these differences is crucial for developing targeted interventions.

Purpose of the Study:

• To investigate neuroanatomical subtypes in schizophrenia using advanced machine learning.
• To identify distinct patterns of brain structure associated with schizophrenia.

Main Methods:

• Utilized semi-supervised machine learning on structural MRI data from a multi-institutional cohort.
• Analyzed regional grey matter, white matter, and CSF volumes in schizophrenia patients (n=307) and healthy controls (n=364).

Main Results:

• Identified two reproducible neuroanatomical subtypes of schizophrenia.
• Subtype 1: Widespread grey matter reduction (prominent in thalamus, nucleus accumbens, medial temporal/prefrontal cortices). Correlated with longer illness duration and lower educational attainment.
• Subtype 2: Increased basal ganglia and internal capsule volumes; otherwise normal brain volumes. Not associated with illness duration or antipsychotic dose.

Conclusions:

• Two distinct neuroanatomical subtypes of schizophrenia exist, challenging the view of uniform brain volume loss.
• Subtypes suggest potentially different neuropathological processes and etiologies.
• Findings support tailored clinical trial enrichment, stratification, and precision diagnostics for schizophrenia.