Learning Common Harmonic Waves on Stiefel Manifold - A New Mathematical Approach for Brain Network Analyses

Area of Science:

• Neuroscience
• Network Science
• Mathematical Modeling

Background:

• Brain alterations in disease follow large-scale brain networks, not random locations.
• Understanding neuropathological spread requires robust network analysis with a mathematical foundation.
• Brain network topology is governed by harmonic waves derived from the Laplacian matrix's Eigen-system.

Purpose of the Study:

• To propose a novel connectome harmonic analysis framework for detecting frequency-based brain alterations relevant to disorders.
• To develop a manifold algebra for inference across harmonic waves, overcoming Euclidean limitations.
• To establish common harmonic waves as new neurobiological bases for understanding disease progression.

Main Methods:

• Developed a novel manifold algebra for inference across harmonic waves.
• Utilized a manifold optimization scheme to learn common harmonic waves from individual Eigen-systems (Stiefel manifold).
• Measured individual harmonic differences using learned common harmonic waves.

Main Results:

• The framework successfully identifies frequency-based alterations in brain networks.
• Common harmonic waves reveal unique propagation patterns of neuropathological burden.
• Applied to Alzheimer's disease, the approach discovered more significant and reproducible network dysfunction than Euclidean methods.

Conclusions:

• Connectome harmonic analysis offers enhanced mathematical insights into brain disorders.
• The novel manifold approach provides a powerful tool for analyzing network dysfunction.
• This method advances our understanding of disease mechanisms and progression in the brain.