Multiple functional connectivity networks fusion for schizophrenia diagnosis

Area of Science:

• Neuroscience
• Medical Imaging
• Computational Psychiatry

Background:

• Accurate schizophrenia diagnosis is critical for patient care and prognosis.
• Existing methods often analyze functional connectivity (FC) within single frequency bands, neglecting cross-frequency interactions and nonlinear dynamics.
• Conventional approaches using Pearson's correlation coefficient (PCC) fail to capture complex relationships within and across frequency bands in resting-state functional magnetic resonance imaging (rs-fMRI) data.

Purpose of the Study:

• To develop and validate a novel multiple networks fusion method for schizophrenia diagnosis.
• To integrate information from different frequency bands and capture nonlinear relationships in brain activity.
• To enhance the accuracy of computer-aided diagnosis for schizophrenia.

Main Methods:

• Constructed functional connectivity (FC) networks within and across four frequency bands (slow-5, slow-4, slow-3, slow-2) using the extended maximal information coefficient (eMIC) on rs-fMRI data.
• Employed a nonlinear network fusion technique to combine these multi-frequency networks into a unified network for each subject.
• Extracted features from the unified network for classification to diagnose schizophrenia.

Main Results:

• The proposed multiple networks fusion method significantly improved classification performance for schizophrenia diagnosis.
• Interactions between brain regions across different frequency bands were found to be crucial for enhanced diagnostic accuracy.
• The fusion approach outperformed conventional FC analyses based on single or combined low-frequency bands.

Conclusions:

• Integrating cross-frequency band interactions via network fusion offers a more comprehensive analysis of brain activity for schizophrenia.
• The developed framework shows promise as a valuable tool for computer-aided diagnosis of schizophrenia.
• This approach highlights the importance of considering nonlinear dynamics and multi-frequency information in neuroimaging studies of psychiatric disorders.