Multivariate neuroanatomical classification of cognitive subtypes in schizophrenia: a support vector machine learning...

Area of Science:

• Neuroimaging
• Psychiatry
• Brain Anatomy

Background:

• Schizophrenia exhibits significant heterogeneity in structural brain abnormalities, complicating the identification of reliable neuroanatomical markers.
• Utilizing more homogenous clinical phenotypes, such as cognitive subtypes, may enhance the accuracy of predicting psychotic disorders based on brain disturbances.

Purpose of the Study:

• To investigate the utility of cognitive subtypes of schizophrenia ('cognitive deficit' and 'cognitively spared') in discriminating these subtypes from healthy controls and each other using neuroanatomical data.
• To determine if multivariate patterns of volumetric brain differences can accurately classify these clinical subtypes.

Main Methods:

• Support vector machine classification was applied to grey- and white-matter volume data.
• The study included 126 cognitively spared schizophrenia patients, 74 cognitive deficit schizophrenia patients, and 134 healthy controls.
• Classification accuracy was assessed with and without sex stratification.

Main Results:

• Cognitive subtypes were distinguished from healthy controls with up to 72% accuracy.
• Cross-validation between subtypes achieved 71% accuracy, indicating common neuroanatomical patterns differentiating both subtypes from controls.
• Classification accuracy significantly improved to 83% for females when the sample was stratified by sex, revealing sex-specific neuroanatomical patterns.

Conclusions:

• Volumetric brain differences between cognitive subtypes are minor in mixed-sex samples compared to common disease-associated changes.
• Distinguishing cognitive subtypes relies on sex-specific neuroanatomical patterns, particularly evident in females.
• Future efforts to differentiate schizophrenia subgroups using neuroanatomical features should consider sex-specific brain organization.