Classification of Alzheimer's Disease by Modeling Brain Networks as Signed Networks Under Deep Learning Frameworks

Area of Science:

• Neuroscience
• Artificial Intelligence
• Medical Diagnostics

Background:

• Alzheimer's disease (AD) is a progressive neurodegenerative disorder with complex pathology.
• Current diagnostic tools for AD lack definitive accuracy.
• Understanding brain region interactions is crucial for AD diagnosis.

Purpose of the Study:

• To develop an innovative approach for predicting and analyzing Alzheimer's disease.
• To leverage signed graph neural network technologies for enhanced AD diagnosis.
• To validate the role of negative correlations in neural interactions for AD prediction.

Main Methods:

• Constructing signed brain network models representing positive and negative correlations.
• Utilizing graph convolutional networks (GCNs) and variants to process signed brain networks.
• Deriving positive and negative attention matrices to identify brain region biomarkers.

Main Results:

• The signed graph model significantly improved Alzheimer's disease prediction accuracy.
• Diagnostic precision increased by at least 19% compared to unsigned models.
• Negative edge information proved vital for enhanced diagnostic capabilities.

Conclusions:

• Signed brain network models effectively capture nuanced neural interactions.
• Incorporating negative correlations is crucial for improving AD diagnostic accuracy.
• This approach shows promise for early Alzheimer's disease diagnosis and clinical applications.