Coherent Pattern in Multi-Layer Brain Networks: Application to Epilepsy Identification.

Area of Science:

• Neuroscience
• Computational Biology
• Medical Imaging

Background:

• Identifying brain diseases using brain network analysis is a significant challenge.
• Existing methods often fuse structural connectivity (SC) and functional connectivity (FC) at the decision level, neglecting their inherent relationship.
• A more integrated approach is needed to leverage the complementary information from SC and FC.

Purpose of the Study:

• To propose a novel method for jointly fusing SC and FC in brain network analysis for disease identification.
• To model the brain as a multi-layer network incorporating both SC and FC.
• To develop a coherent pattern that captures the relationship between SC and FC for enhanced classification.

Main Methods:

• Modeled the brain network as a multi-layer network using SC and FC.
• Proposed a 'coherent pattern' representing paired-subgraphs from SC and FC within the same node-set.
• Developed a framework involving multi-layer network construction, coherent pattern mining, discriminative pattern selection, and Support Vector Machine (SVM) classification.

Main Results:

• The proposed coherent pattern effectively captures both individual layer connectivity and the inter-layer relationship between SC and FC.
• The developed framework demonstrated superior performance in brain disease classification compared to existing methods.
• Experimental results on epilepsy datasets validated the efficacy of the novel approach.

Conclusions:

• The multi-layer network approach with coherent pattern fusion offers a powerful new strategy for brain disease identification.
• This method enhances classification accuracy by considering the intrinsic relationship between structural and functional brain connectivity.
• The findings suggest significant potential for clinical applications in diagnosing neurological disorders.