Dissociation between neuroanatomical and symptomatic subtypes in schizophrenia

Area of Science:

• Neuroscience
• Psychiatry
• Machine Learning in Medicine

Background:

• Schizophrenia is a complex disorder with significant clinical and biological heterogeneity.
• Identifying distinct subtypes is crucial for improving diagnostic accuracy and enabling precise therapeutic strategies.
• Understanding the biological basis of clinical heterogeneity in schizophrenia subtypes remains a key challenge.

Purpose of the Study:

• To investigate the association between machine learning (ML)-based neuroanatomical and symptomatic subtypes of schizophrenia.
• To determine if distinct neuroanatomical patterns correlate with specific clinical presentations in schizophrenia patients.

Main Methods:

• Utilized K-means and hierarchical clustering methods on gray matter volume (GMV) data from 314 schizophrenia patients and 257 healthy controls.
• Employed the Positive and Negative Syndrome Scale (PANSS) to define symptomatic subtypes.
• Analyzed neuroanatomical and symptomatic subtypes across four independent sites.

Main Results:

• ML identified two neuroanatomical subtypes: one with focally increased GMV and another with widespread reduced GMV compared to controls.
• Two symptomatic subtypes were identified based on PANSS scores, with subtype-1 exhibiting more severe symptoms than subtype-2.
• A significant dissociation was found between neuroanatomical and symptomatic subtypes, with no correlation between GMV patterns and symptom severity, and vice versa.
• These dissociation patterns were consistently observed across diverse patient populations and disease progressions.

Conclusions:

• The study reveals a replicable dissociation between ML-derived neuroanatomical and symptomatic subtypes of schizophrenia.
• This finding offers a novel perspective on understanding the complex heterogeneity of schizophrenia.
• The results suggest that distinct biological pathways may underlie different clinical presentations in schizophrenia.