Brain Network Analysis of Schizophrenia Patients Based on Hypergraph Signal Processing

Area of Science:

• Neuroscience
• Network Science
• Medical Imaging

Background:

• High-order relationships among brain regions-of-interest (ROIs) are crucial for understanding neurological diseases.
• Existing brain hypernetworks often treat hyperedges equally, limiting their analytical power.
• A weighted approach is needed to better represent complex brain connectivity.

Purpose of the Study:

• To propose a framework for constructing a truly weighted brain hypernetwork using an adjacency tensor.
• To develop a novel hyperedge weight estimation method for brain hypernetworks.
• To apply hypergraph signal processing for analyzing brain networks and classifying neurological conditions.

Main Methods:

• Developed a weighted brain hypernetwork framework using an adjacency tensor.
• Implemented a novel hyperedge weight estimation method.
• Applied hypergraph Fourier transform and tensor decomposition for spectral analysis.
• Utilized hypergraph spectrum and spectral signals as classification features.

Main Results:

• Identified significantly more high-frequency components in the brain spectrum of schizophrenia patients compared to controls.
• Observed a greater average amplitude in the brain spectrum of patients.
• Achieved effective classification of schizophrenia using spectral features on two public datasets.

Conclusions:

• The proposed weighted brain hypernetwork and spectral analysis offer a powerful tool for neurological disease research.
• Hypergraph spectral features provide effective classification metrics for conditions like schizophrenia.
• This approach advances the analysis of complex brain connectivity patterns.