Unraveling schizophrenia replicable functional connectivity disruption patterns across sites

Area of Science:

• Neuroimaging
• Psychiatric Disorders
• Data Harmonization

Background:

• Functional connectivity (FC) disruptions are characteristic of schizophrenia but exhibit heterogeneous patterns across sites, hindering clinical generalization.
• Addressing site-related heterogeneity is crucial for developing robust neuroimaging biomarkers for schizophrenia.

Purpose of the Study:

• To compare four site-effect correction methods for detecting aberrant functional connectivity in schizophrenia.
• To identify replicable FC disruption subnets across sites.
• To evaluate the clinical relevance and diagnostic potential of these subnets.

Main Methods:

• Utilized nodal-based FC data from 340 schizophrenia patients (SZ) and 348 normal controls (NC) across seven scanners.
• Compared ComBat harmonization with three other methods for site-effect correction.
• Applied hierarchical clustering to identify FC disruption subnets and correlated them with clinical symptoms.
• Assessed the diagnostic performance of subnet-specific FC measures using receiver operating characteristic (ROC) analysis.

Main Results:

• ComBat harmonization demonstrated superior performance (F1 score: 0.806 ± 0.145) in detecting aberrant FCs compared to other methods.
• Identified three replicable FC disruption subnets: sensory hypo-connectivity (Net1), thalamus-striatum-ventral attention hypo-connectivity (Net2), and thalamus-sensory hyper-connectivity (Net3).
• Net1 composite FC negatively correlated with hostility and disorientation (p < .05).
• Subnet-specific FCs achieved a robust Area Under the ROC Curve (AUC) of 0.728 for SZ vs. NC classification.

Conclusions:

• ComBat harmonization is the most robust method for detecting aberrant functional connectivity in schizophrenia.
• The identified three subnet-specific FC measures represent potential replicable neuroimaging biomarkers for schizophrenia.
• These subnet measures offer meaningful discrimination between schizophrenia patients and controls.