Classifying schizophrenia using functional MRI and investigating underlying functional phenomena.

Area of Science:

• Neuroscience
• Psychiatry
• Medical Imaging

Background:

• Schizophrenia (SZ) is associated with functional brain abnormalities, but their precise relationship with disease progression is not fully understood.
• Existing research highlights altered activity in specific brain regions in SZ patients.

Purpose of the Study:

• To investigate the relationships between brain functional abnormalities and schizophrenia progression.
• To identify key neuroimaging features predictive of schizophrenia using machine learning.
• To explore the causal relationships within functional brain networks in schizophrenia.

Main Methods:

• Utilized resting-state functional magnetic resonance imaging (fMRI) data from 56 schizophrenia patients and 56 healthy controls.
• Analyzed fractional amplitude of low-frequency fluctuations (fALFF), regional homogeneity (ReHo), and degree centrality (DC) as neuroimaging metrics.
• Applied machine learning classifiers, Louvain community detection, and structural equation modeling to identify predictive features and causal pathways.

Main Results:

• Machine learning models achieved high prediction accuracy (average 0.9241, best SVM 0.9464) using fALFF, ReHo, and DC.
• Abnormal ReHo in the right middle frontal gyrus was the most significant feature for classification and directly impacted schizophrenia.
• Identified two functional clusters (FClus) with internal causal influences, one positively and one negatively associated with schizophrenia onset and progression.

Conclusions:

• Discovered interactions within functional brain clusters that may influence schizophrenia onset and progression.
• The contribution of features to classification models may reflect direct impact, not necessarily overall disease process importance.
• Findings provide insights into the complex functional network dynamics underlying schizophrenia.