Support Vector Machine-Based Schizophrenia Classification Using Morphological Information from Amygdaloid and...

Area of Science:

• Neuroimaging
• Psychiatric Disorders
• Machine Learning

Background:

• Structural abnormalities in the hippocampus and amygdala are observed in schizophrenia.
• The utility of these subcortical region morphologies for machine learning-based schizophrenia classification remains largely unexplored.

Purpose of the Study:

• To investigate the feasibility of using amygdaloid and hippocampal subregion volumes for schizophrenia classification via machine learning.
• To assess the diagnostic potential of these specific neuroanatomical features.

Main Methods:

• Utilized T1 structural MRI data from 57 schizophrenia patients and 69 healthy controls.
• Extracted volumes of 26 hippocampal and 20 amygdaloid subregions.
• Employed a Sequential Backward Elimination (SBE) algorithm for feature selection and a linear Support Vector Machine (SVM) classifier.

Main Results:

• The SBE-SVM model achieved an 81.75% classification accuracy.
• Demonstrated a sensitivity of 84.21% and a specificity of 81.16%.
• Achieved an Area Under the Curve (AUC) of 0.8241 (p < 0.001).

Conclusions:

• Structural changes in the hippocampus and amygdala are evident in schizophrenia.
• Morphological data from these subregions show promise for machine learning-based schizophrenia classification.
• This approach offers a potential tool for objective diagnostic assessment in psychiatry.