Hyper-Ordinal Pattern: Measuring High-Order Connection Relationship in Brain Disease Networks.

Area of Science:

• Neuroscience
• Graph Theory
• Machine Learning

Background:

• Brain hyper-networks, represented as hypergraphs, are crucial for analyzing high-order interactions in brain regions.
• Existing hypergraph methods often overlook weighted information within hyperedges, limiting their effectiveness in brain network analysis.
• Mild cognitive impairment (MCI) and Alzheimer's disease (AD) research benefits from advanced brain hyper-network analysis.

Purpose of the Study:

• To propose a novel approach, hyper-ordinal pattern (HOP), for representing and analyzing brain hyper-networks.
• To address the limitation of existing methods by incorporating weighted information on hyperedges.
• To develop a new kernel, ordinal pattern based hyper-network (OPHN), for brain hyper-network similarity calculation and disease classification.

Main Methods:

• Constructed hyper-ordinal patterns (HOPs) by utilizing ordinal pattern relationships on weighted hyperedges.
• Developed a node HOP (NHOP) kernel for measuring node similarity within brain hyper-networks.
• Introduced the ordinal pattern based hyper-network (OPHN) kernel to compute brain hyper-network similarity.

Main Results:

• The OPHN kernel demonstrated superior performance compared to state-of-the-art methods in classifying brain diseases (MCI and AD).
• The NHOP kernel successfully identified altered hyper-ordinal patterns in brain hyper-networks of patients with neurological disorders.
• Experimental results validated the effectiveness of the proposed OPHN kernel in brain disease classification tasks.

Conclusions:

• The proposed OPHN kernel offers a significant advancement in brain hyper-network analysis for disease classification.
• HOPs and the NHOP kernel provide novel tools for understanding complex interactions in brain networks.
• This approach holds promise for improving diagnostic capabilities for neurodegenerative diseases like MCI and AD.