Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Graphical Structural Learning of rs-fMRI data in Heavy Smokers

Yiru Gong, Qimin Zhang, Huili Zheng, Zheyan Liu, Shaohan Chen

TL;DR

This work addresses how heavy smoking alters brain network topology by applying Gaussian Undirected Graphs to resting-state fMRI data from smokers and controls. The authors employ Graphical Lasso to estimate sparse precision matrices (with $K = \Sigma^{-1}$ and the estimator $\hat{K}^{gl}$ defined by $\hat{K}^{gl} = \arg\min_K\{ -\log\det(K) + \mathrm{tr}(SK) + \lambda \|K\|_1 \}$) and use Rotation Information Criterion to select the regularization parameter. Stability is assessed via bootstrap across multiple resamples, and graph similarity is quantified with Sorensen-Dice and node-wise Jaccard scores, revealing 50 regions with significant connectivity differences, notably Temporal_Inf_L, Thalamus_R, and Cerebelum_Crus2. When compared to non-graphical methods (ICA, PCA, GLMNET), the graphical approach better captures smoking-related brain connectivity changes, supporting its utility for high-dimensional neuroimaging data and offering insights for clinical research and smoking cessation strategies.

Abstract

Recent studies revealed structural and functional brain changes in heavy smokers. However, the specific changes in topological brain connections are not well understood. We used Gaussian Undirected Graphs with the graphical lasso algorithm on rs-fMRI data from smokers and non-smokers to identify significant changes in brain connections. Our results indicate high stability in the estimated graphs and identify several brain regions significantly affected by smoking, providing valuable insights for future clinical research.

Graphical Structural Learning of rs-fMRI data in Heavy Smokers

TL;DR

This work addresses how heavy smoking alters brain network topology by applying Gaussian Undirected Graphs to resting-state fMRI data from smokers and controls. The authors employ Graphical Lasso to estimate sparse precision matrices (with and the estimator defined by ) and use Rotation Information Criterion to select the regularization parameter. Stability is assessed via bootstrap across multiple resamples, and graph similarity is quantified with Sorensen-Dice and node-wise Jaccard scores, revealing 50 regions with significant connectivity differences, notably Temporal_Inf_L, Thalamus_R, and Cerebelum_Crus2. When compared to non-graphical methods (ICA, PCA, GLMNET), the graphical approach better captures smoking-related brain connectivity changes, supporting its utility for high-dimensional neuroimaging data and offering insights for clinical research and smoking cessation strategies.

Abstract

Recent studies revealed structural and functional brain changes in heavy smokers. However, the specific changes in topological brain connections are not well understood. We used Gaussian Undirected Graphs with the graphical lasso algorithm on rs-fMRI data from smokers and non-smokers to identify significant changes in brain connections. Our results indicate high stability in the estimated graphs and identify several brain regions significantly affected by smoking, providing valuable insights for future clinical research.
Paper Structure (16 sections, 3 equations, 3 figures, 2 tables)

This paper contains 16 sections, 3 equations, 3 figures, 2 tables.

Table of Contents

  1. Introduction
  2. Data
  3. Method
  4. Gaussian Undirected Model
  5. Graphical Lasso Estimation
  6. Parameter tuning and RIC criteria
  7. Similarity comparison
  8. Non-structural Machine Learning Models
  9. Results
  10. Validation of Gaussian Undirected graphs
  11. Similarity Score Comparison based on stable edges
  12. Comparison between Graphical and Non-graphical Model
  13. Conclusion
  14. Discussion
  15. Alteration in Brain Region Connections
  16. ...and 1 more sections

Figures (3)

  • Figure 1: Estimated graph of different lambda values in non-smokers
  • Figure 2: Connectivity between brain regions - Pearson correlation matrices (top triangle) and undirected graphs (bottom triangle; direct connections as discovered by Glasso, correlation calculated from estimated precision matrix). Left: non-smokers; right: smokers.
  • Figure 3: Brain connection visualization in Top 3 largely changed brain regions - Red: node region, blue: directed connected regions; left: non-smokers, right: smokers.