Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Exploring Subnetwork Interactions in Heterogeneous Brain Network via Prior-Informed Graph Learning

Siyu Liu, Guangqi Wen, Peng Cao, Jinzhu Yang, Xiaoli Liu, Fei Wang, Osmar R. Zaiane

Abstract

Modeling the complex interactions among functional subnetworks is crucial for the diagnosis of mental disorders and the identification of functional pathways. However, learning the interactions of the underlying subnetworks remains a significant challenge for existing Transformer-based methods due to the limited number of training samples. To address these challenges, we propose KD-Brain, a Prior-Informed Graph Learning framework for explicitly encoding prior knowledge to guide the learning process. Specifically, we design a Semantic-Conditioned Interaction mechanism that injects semantic priors into the attention query, explicitly navigating the subnetwork interactions based on their functional identities. Furthermore, we introduce a Pathology-Consistent Constraint, which regularizes the model optimization by aligning the learned interaction distributions with clinical priors. Additionally, KD-Brain leads to state-of-the-art performance on a wide range of disorder diagnosis tasks and identifies interpretable biomarkers consistent with psychiatric pathophysiology. Our code is available at https://anonymous.4open.science/r/KDBrain.

Exploring Subnetwork Interactions in Heterogeneous Brain Network via Prior-Informed Graph Learning

Abstract

Modeling the complex interactions among functional subnetworks is crucial for the diagnosis of mental disorders and the identification of functional pathways. However, learning the interactions of the underlying subnetworks remains a significant challenge for existing Transformer-based methods due to the limited number of training samples. To address these challenges, we propose KD-Brain, a Prior-Informed Graph Learning framework for explicitly encoding prior knowledge to guide the learning process. Specifically, we design a Semantic-Conditioned Interaction mechanism that injects semantic priors into the attention query, explicitly navigating the subnetwork interactions based on their functional identities. Furthermore, we introduce a Pathology-Consistent Constraint, which regularizes the model optimization by aligning the learned interaction distributions with clinical priors. Additionally, KD-Brain leads to state-of-the-art performance on a wide range of disorder diagnosis tasks and identifies interpretable biomarkers consistent with psychiatric pathophysiology. Our code is available at https://anonymous.4open.science/r/KDBrain.
Paper Structure (18 sections, 6 equations, 3 figures, 1 table)

This paper contains 18 sections, 6 equations, 3 figures, 1 table.

Table of Contents

  1. Introduction
  2. Method
  3. Preliminaries
  4. Subnetwork Topology Representation Learning
  5. Subnetwork Semantic Interaction Learning (SSIL)
  6. Disorder-specfic Semantic Prior Embedding Generation
  7. Semantic-Conditioned Interaction
  8. The Subnetwork Interaction Prior
  9. Experiments and Interpretation
  10. Dataset and Experimental Details
  11. Results and Discussion
  12. Comparison with SOTAs
  13. Ablation Analysis
  14. Effectiveness of Semantic-Conditioned Interaction
  15. Interpretation
  16. ...and 3 more sections

Figures (3)

  • Figure 1: Illustration of the KD-Brain.
  • Figure 2: (a) The correlation between prior disorder-agnostic subnetwork embeddings, where red lines represent significant correlations ($P < 0.05$). (b) The effectiveness of Semantic-Conditioned Interaction.
  • Figure 3: The visualization of multi-level biomarkers identified by KD-Brain. (a) Top-2 critical subnetwork interaction strengths, quantified by the normalized attention coefficients $\alpha_{k,j}^{(q)}$ in $\mathcal{P}_{sni}^{(q)}$. (b) Top-5 critical brain regions. (c) Top-2 functional pathways, where the pathway score is calculated as the joint probability of sequential interactions across orders: $Score = \prod_{l=1}^{q} \alpha_{k, j}^{(l)}$, where $q=1,2,3$, respectively.