Constructing high-order functional networks based on hypergraph for diagnosis of autism spectrum disorders

Area of Science:

• Neuroscience
• Brain Connectivity Analysis
• Medical Imaging

Background:

• High-order functional connectivity networks (FCNs) are crucial for understanding brain function and disease mechanisms.
• Traditional FCN methods, while useful, suffer from noise, redundancy, and high computational complexity.
• Existing whole-brain connectivity analysis may be compromised by less significant brain regions.

Purpose of the Study:

• To introduce a novel hypergraph-based method for constructing high-order FCNs.
• To improve the accuracy and reliability of brain connectivity analysis.
• To enhance the diagnostic capabilities for brain diseases.

Main Methods:

• Constructed a low-order FCN from resting-state functional Magnetic Resonance Imaging (rs-fMRI) time series.
• Utilized a hypergraph approach to identify and average 'good friend' brain region time series.
• Generated a hypergraph high-order FCN by calculating correlations on averaged friend circle sequences.

Main Results:

• The proposed hypergraph-based method demonstrated superior performance compared to traditional FCN construction techniques.
• Feature fusion of hypergraph high-order FCN and low-order FCN improved diagnostic accuracy for brain diseases.
• The method showed effectiveness in assisting the diagnosis of autism spectrum disorder (ASD).

Conclusions:

• The hypergraph-based approach offers a more accurate and efficient way to construct high-order FCNs.
• Integrating low-order and high-order FCNs via feature fusion enhances diagnostic potential.
• Future work will explore extending this method to other brain connectivity patterns beyond ASD.