Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Balanced Graph Structure Information for Brain Disease Detection

Falih Gozi Febrinanto, Mujie Liu, Feng Xia

TL;DR

This paper tackles brain disease detection from resting-state fMRI by addressing the volatility of graph structures used to model brain networks. It introduces Bargrain, a framework that balances two graph representations: a filtered correlation graph capturing domain knowledge and an optimal learnable sampling graph learned via a differentiable Gumbel reparameterization method. Each graph is embedded with a two-layer GCN, and their graph-level representations are concatenated for final binary disease classification, yielding state-of-the-art performance across three datasets with an average F1 of $0.7329$. The approach enhances robustness to noise and overfitting while preserving interpretability by comparing two complementary graph structures and their role in disease prediction, with potential for data-efficient scaling in dense brain networks.

Abstract

Analyzing connections between brain regions of interest (ROI) is vital to detect neurological disorders such as autism or schizophrenia. Recent advancements employ graph neural networks (GNNs) to utilize graph structures in brains, improving detection performances. Current methods use correlation measures between ROI's blood-oxygen-level-dependent (BOLD) signals to generate the graph structure. Other methods use the training samples to learn the optimal graph structure through end-to-end learning. However, implementing those methods independently leads to some issues with noisy data for the correlation graphs and overfitting problems for the optimal graph. In this work, we proposed Bargrain (balanced graph structure for brains), which models two graph structures: filtered correlation matrix and optimal sample graph using graph convolution networks (GCNs). This approach aims to get advantages from both graphs and address the limitations of only relying on a single type of structure. Based on our extensive experiment, Bargrain outperforms state-of-the-art methods in classification tasks on brain disease datasets, as measured by average F1 scores.

Balanced Graph Structure Information for Brain Disease Detection

TL;DR

This paper tackles brain disease detection from resting-state fMRI by addressing the volatility of graph structures used to model brain networks. It introduces Bargrain, a framework that balances two graph representations: a filtered correlation graph capturing domain knowledge and an optimal learnable sampling graph learned via a differentiable Gumbel reparameterization method. Each graph is embedded with a two-layer GCN, and their graph-level representations are concatenated for final binary disease classification, yielding state-of-the-art performance across three datasets with an average F1 of . The approach enhances robustness to noise and overfitting while preserving interpretability by comparing two complementary graph structures and their role in disease prediction, with potential for data-efficient scaling in dense brain networks.

Abstract

Analyzing connections between brain regions of interest (ROI) is vital to detect neurological disorders such as autism or schizophrenia. Recent advancements employ graph neural networks (GNNs) to utilize graph structures in brains, improving detection performances. Current methods use correlation measures between ROI's blood-oxygen-level-dependent (BOLD) signals to generate the graph structure. Other methods use the training samples to learn the optimal graph structure through end-to-end learning. However, implementing those methods independently leads to some issues with noisy data for the correlation graphs and overfitting problems for the optimal graph. In this work, we proposed Bargrain (balanced graph structure for brains), which models two graph structures: filtered correlation matrix and optimal sample graph using graph convolution networks (GCNs). This approach aims to get advantages from both graphs and address the limitations of only relying on a single type of structure. Based on our extensive experiment, Bargrain outperforms state-of-the-art methods in classification tasks on brain disease datasets, as measured by average F1 scores.
Paper Structure (17 sections, 3 equations, 2 figures, 2 tables)

This paper contains 17 sections, 3 equations, 2 figures, 2 tables.

Table of Contents

  1. Introduction
  2. Related Work
  3. GNNs for Brain Disease Detection
  4. Learnable Graph Generations
  5. Framework Overview
  6. Proposed Framework
  7. Brain Signal Preprocessing
  8. Graph Modeling
  9. Classifier
  10. Experiment
  11. Datasets
  12. Baseline Approaches and Reproducibility
  13. Model Selection and Experiment Setup
  14. RQ1. Performance Comparison
  15. RQ2. Ablation Studies
  16. ...and 2 more sections

Figures (2)

  • Figure 1: The proposed framework. Bargrain starts with Brain Signal Preprocessing to develop a correlation matrix based on ROI signals. Graph Modeling processes two graph structures to balance information from domain knowledge and optimal structures. Classifier maps the combined knowledge for brain disease detection.
  • Figure 2: Balanced graph structure interpretation for a person in the Cobre dataset.