Identification of patterns of gray matter abnormalities in schizophrenia using source-based morphometry and bagging

Area of Science:

• Neuroimaging
• Psychiatry
• Machine Learning

Background:

• Schizophrenia diagnosis lacks objective tests or biomarkers, hindering understanding.
• Structural magnetic resonance imaging (sMRI) studies on gray matter concentration (GMC) show inconsistent findings.
• Identifying reliable brain abnormality patterns is crucial for schizophrenia research.

Purpose of the Study:

• To develop a machine learning approach for detecting consistent patterns of GMC differences in schizophrenia.
• To identify specific brain regions exhibiting significant GMC variations between patients and controls.
• To improve classification accuracy for distinguishing schizophrenia patients from healthy individuals.

Main Methods:

• Utilized multi-site sMRI data from the Mind Clinical Imaging Consortium (124 controls, 110 patients).
• Applied a machine learning approach incorporating resampling techniques.
• Employed source-based morphometry to detect GMC differences.

Main Results:

• The proposed method achieved a higher classification rate compared to existing algorithms.
• Identified brain regions with consistent GMC differences between schizophrenia patients and controls.
• Detected patterns align with existing neuroimaging findings in schizophrenia literature.

Conclusions:

• The machine learning approach effectively identifies brain regions with significant GMC differences in schizophrenia.
• Results suggest potential for developing objective diagnostic tools for schizophrenia.
• The multi-site data approach indicates potential for replicability across different datasets.