PAC-KENReg: An interpretable hypergraph framework capturing nonlinear and dynamic functional connectivity for...

Area of Science:

• Neuroscience
• Computational Psychiatry
• Network Science

Background:

• Hypergraph analysis is key for modeling brain connectivity in psychiatric disorders.
• Existing methods struggle with nonlinear interactions and lack neurophysiological meaning in hyperedge weights.

Purpose of the Study:

• To develop an enhanced hypergraph framework integrating nonlinear modeling and neurophysiological information.
• To analyze both static and dynamic functional brain networks for psychiatric disorder research.

Main Methods:

• Introduced "Phase-Amplitude Coupling-weighted Kernelized Elastic Net Regularization" (PAC-KENReg).
• Utilized kernelized Elastic Net for high-order nonlinear interactions and hypergraph topology.
• Weighted hyperedges with phase-amplitude coupling (PAC) strength for neurophysiological insight.
• Extended to dynamic analysis using temporal stability matrices.

Main Results:

• Achieved 78.39% classification accuracy for Major Depressive Disorder (MDD) on EEG data.
• Demonstrated superior discriminative power for Attention Deficit Hyperactivity Disorder (ADHD).
• Identified key pathology-associated brain regions and abnormal dynamic connectivity profiles.

Conclusions:

• PAC-KENReg offers a robust and interpretable tool for brain connectivity analysis.
• Provides an effective computational approach for objective diagnosis and mechanistic investigation of psychiatric disorders.