Core-Periphery Principle Guided State Space Model for Functional Connectome Classification.

Area of Science:

• Neuroscience
• Computational Neuroscience
• Medical Imaging

Background:

• Human brain network organization is key to understanding brain function and diagnosing neurological disorders.
• Functional connectivity analysis using fMRI and machine learning is advancing, but faces limitations.
• Traditional machine learning struggles with complex relationships, while deep learning models like Transformers have high computational costs.

Purpose of the Study:

• To develop an efficient and effective framework for functional connectome classification.
• To address the limitations of existing machine learning and deep learning models in brain network analysis.
• To improve the diagnosis of neurological disorders through advanced neuroimaging analysis.

Main Methods:

• Proposed a Core-Periphery State-Space Model (CP-SSM) for functional connectome classification.
• Integrated Mamba, a selective state-space model with linear complexity, to capture long-range dependencies in brain networks.
• Developed CP-MoE, a core-periphery guided Mixture-of-Experts, to enhance representation learning of connectivity patterns.

Main Results:

• CP-SSM demonstrated superior classification performance compared to Transformer-based models on ABIDE and ADNI fMRI datasets.
• The proposed model significantly reduced computational complexity.
• Effectively captured long-range dependencies and improved representation learning in functional brain networks.

Conclusions:

• CP-SSM offers an effective and computationally efficient solution for modeling brain functional connectivity.
• The framework shows significant promise for neuroimaging-based diagnosis of neurological diseases.
• The study provides a novel approach to analyzing complex brain network data.