Investigating possible subtypes of schizophrenia patients and controls based on brain cortical thickness

Abstract 

Schizophrenia is a heterogeneous disease in which different dimensions could be associated with localized subtypes in cortical thickness of the brain. Subtypes in data that includes patients and controls could be associated with patient/control could associate with patient/control groupings. Testing for subtypes provides a non-parametric investigation of group differences. Cortical thickness maps, generated from magnetic resonance images of 96 patients with schizophrenia and 106 controls, were co-registered and corrected for age-related thinning. At multiple map locations, the number of (sub)types best explaining cortical thickness in the patients, the controls, and both combined was determined. Grey matter volumes of selected regions were measured. Both patients and controls, considered independently, were predominantly homogeneous in cortical thickness. The few bimodal regions were similar in both groups. The combined subjects' cortical thickness was bimodal over 34% of the cortical mantle and otherwise unimodal. Further probing of these bimodal regions showed that subjects tending to belong to thinner modes were significantly more likely to be patients, and grey matter volumes of most bimodal regions were significantly smaller in patients. The study found no subtypes specific to patients. It suggested, however, that associations between abnormally thin cortex and schizophrenia are more widespread than shown by previously published results based on significance testing.

Keywords: Clustering, Non-parametric statistics, Vertex-based morphometry, K-means, Machine learning