Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

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

Minheng Chen, Xiaowei Yu, Jing Zhang, Tong Chen, Chao Cao, Yan Zhuang, Yanjun Lyu, Lu Zhang, Tianming Liu, Dajiang Zhu

TL;DR

The paper tackles efficient functional connectome classification by addressing Transformer quadratic complexity with a Core-Periphery State-Space Model (CP-SSM). CP-SSM combines Mamba, a linear-complexity state-space block, with a Core-Periphery guided Mixture-of-Experts (CP-MoE) to model long-range dependencies and CP organization in brain networks. Across ABIDE and ADNI, CP-SSM achieves superior classification performance over Transformer-based baselines while reducing computational cost; ablation and neuroscientific analyses further support the method’s effectiveness and interpretability. The work offers a scalable, biologically informed framework for neuroimaging-based disease diagnosis and provides insight into discriminative brain regions associated with ASD and MCI.

Abstract

Understanding the organization of human brain networks has become a central focus in neuroscience, particularly in the study of functional connectivity, which plays a crucial role in diagnosing neurological disorders. Advances in functional magnetic resonance imaging and machine learning techniques have significantly improved brain network analysis. However, traditional machine learning approaches struggle to capture the complex relationships between brain regions, while deep learning methods, particularly Transformer-based models, face computational challenges due to their quadratic complexity in long-sequence modeling. To address these limitations, we propose a Core-Periphery State-Space Model (CP-SSM), an innovative framework for functional connectome classification. Specifically, we introduce Mamba, a selective state-space model with linear complexity, to effectively capture long-range dependencies in functional brain networks. Furthermore, inspired by the core-periphery (CP) organization, a fundamental characteristic of brain networks that enhances efficient information transmission, we design CP-MoE, a CP-guided Mixture-of-Experts that improves the representation learning of brain connectivity patterns. We evaluate CP-SSM on two benchmark fMRI datasets: ABIDE and ADNI. Experimental results demonstrate that CP-SSM surpasses Transformer-based models in classification performance while significantly reducing computational complexity. These findings highlight the effectiveness and efficiency of CP-SSM in modeling brain functional connectivity, offering a promising direction for neuroimaging-based neurological disease diagnosis.

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

TL;DR

The paper tackles efficient functional connectome classification by addressing Transformer quadratic complexity with a Core-Periphery State-Space Model (CP-SSM). CP-SSM combines Mamba, a linear-complexity state-space block, with a Core-Periphery guided Mixture-of-Experts (CP-MoE) to model long-range dependencies and CP organization in brain networks. Across ABIDE and ADNI, CP-SSM achieves superior classification performance over Transformer-based baselines while reducing computational cost; ablation and neuroscientific analyses further support the method’s effectiveness and interpretability. The work offers a scalable, biologically informed framework for neuroimaging-based disease diagnosis and provides insight into discriminative brain regions associated with ASD and MCI.

Abstract

Understanding the organization of human brain networks has become a central focus in neuroscience, particularly in the study of functional connectivity, which plays a crucial role in diagnosing neurological disorders. Advances in functional magnetic resonance imaging and machine learning techniques have significantly improved brain network analysis. However, traditional machine learning approaches struggle to capture the complex relationships between brain regions, while deep learning methods, particularly Transformer-based models, face computational challenges due to their quadratic complexity in long-sequence modeling. To address these limitations, we propose a Core-Periphery State-Space Model (CP-SSM), an innovative framework for functional connectome classification. Specifically, we introduce Mamba, a selective state-space model with linear complexity, to effectively capture long-range dependencies in functional brain networks. Furthermore, inspired by the core-periphery (CP) organization, a fundamental characteristic of brain networks that enhances efficient information transmission, we design CP-MoE, a CP-guided Mixture-of-Experts that improves the representation learning of brain connectivity patterns. We evaluate CP-SSM on two benchmark fMRI datasets: ABIDE and ADNI. Experimental results demonstrate that CP-SSM surpasses Transformer-based models in classification performance while significantly reducing computational complexity. These findings highlight the effectiveness and efficiency of CP-SSM in modeling brain functional connectivity, offering a promising direction for neuroimaging-based neurological disease diagnosis.

Paper Structure

This paper contains 13 sections, 6 equations, 4 figures, 3 tables.

Table of Contents

  1. Introduction
  2. Material and Method
  3. Data Acquisition and Preprocessing
  4. State Space Model for Functional Connectome Classification
  5. Core-periphery Principle Guided Mixture-of-Experts
  6. Experiments
  7. Experiment Settings
  8. Performance Comparison with Baseline Methods
  9. Ablation Study
  10. Neuroscientific Analysis
  11. Conclusion
  12. Subject Demographics and Preprocessing details of ADNI
  13. Implementation Details of the Baselines

Figures (4)

  • Figure 1: Overall architecture of the proposed method. Our approach is founded on a state-space model, with its key components comprising a Mamba block and a Core-Periphery principle guided MoE.
  • Figure 2: Sensitivity analysis of CP-SSM on ABIDE. The hyperparameters include core node rate $r$ and top-$k$ expert selection.
  • Figure 3: Top 5 discriminative brain regions derived from the learnable weight in the last CP-SSM layer on (a) ABIDE and (b) ADNI, with different colormap intensities reflecting relative significance.
  • Figure 4: Visualization of functional connectivity for a randomly selected subject from the ABIDE and ADNI datasets respectively.