Investigating BMI, Brain Structure, and Decision-Making
Authors:
W. Farzana [1] , L. Ma [2] , F. Gerry Moeller [2] , and K. M. Iftekharuddin [1,3]
Affiliations:
  1. Vision Lab, Department of Electrical and Computer Engineering, Old Dominion University, Norfolk, VA 23529
  1. Institute of Drug and Alcohol Studies, Department of Psychiatry, Virginia Commonwealth University, Richmond, VA, USA 23284-2520
  1. Data Science Institute, Old Dominion University, Virginia Beach, VA 23462
Research Questions
  • Does BMI directly influence delay discounting behavior?
  • Can brain structures explain this potential effect?
  • How do we integrate Delay Discounting (DD) , BMI, and brain measures?
  • How can we control for age, gender, and education?
Main Objective
We investigate BMI's relationship with DD. We explore whether brain structures, including cortical, subcortical volume, thickness, and CbVD, mediate this relationship. We also control for key demographic variables.
Directed Acyclic Graph (DAG)
Data Generating Process Hypothesis
This graphical model hypothesizes that BMI impacts brain gray matter. It affects cortical thickness, volume, sub-cortical volume, and cerebrovascular disease (CbVD). CbVD also influences gray matter measures. Ultimately, all these measures affect cognition (DD_AUC).
A path is a sequence of edges connecting nodes.
  • BMI → Delay Discounting (DD_AUC)
  • BMI → Subcortical Volume (Sub-Cort Vol) → DD_AUC (Mediator: Chain Structure)
Reference: Association Between Midlife Obesity and Its Metabolic Consequences, Cerebrovascular Disease, and Cognitive Decline (Clinical Research Article, 2021).
Methodology Overview
Our investigation integrates advanced modeling techniques with robust neuroimaging data. This approach allows for a comprehensive analysis of the proposed relationships.
Modeling
We employ Bayesian Structural Equation Modeling (BSEM). This allows for complex relationships. It also accounts for measurement error.
HCP Data
The Human Connectome Project (HCP) provides high-quality neuroimaging data. This public dataset supports our detailed brain structure analysis.
Detailed Brain Structure Impacts
Further examination highlights the specific contributions of various brain components. Both cortical volume and cerebrovascular disease play critical roles.
Cortical Volume
Cortical volume includes both gray and white matter. Changes in this volume can affect overall brain function. This directly influences cognitive flexibility.
Cerebrovascular Disease (CbVD)
CbVD, characterized by white matter hyperintensities, can impair cognitive processing. It disrupts neural pathways essential for optimal decision-making.
Impact of Demographics on Delay Discounting
We assessed the influence of age, gender, and education on delay discounting. Education emerged as a significant predictor of this cognitive trait.
Covariate Effects
Education showed a statistically significant positive effect. Gender and age did not significantly influence DD_AUC in this analysis.
Data Description
  • Male: 176 participants
  • Female: 344 participants
This dataset includes a diverse sample size. The gender distribution is approximately 1:2 male to female.
Data and Graph Node Description
Data Overview
Our study utilizes publicly available data from the Human Connectome Project (HCP). This dataset provides extensive neuroimaging and behavioral information.
1
BMI Categories
  • BMI < 25: Healthy weight (control)
  • BMI ≥ 30: Obese
Total 525 cases, 4 missing DD values.
2
Sample Sizes
  • Total: 520 participants (1 missing education data)
  • BMI < 25: 350 participants
  • BMI ≥ 30: 170 participants
1
Cortical Volume
Includes white matter plus gray matter.
2
Sub-cortical & Cortical Thickness
Derived from DKT atlas using FreeSurfer.
3
Sub-cortical Structures
Thalamus, caudate, putamen, pallidum, hippocampus, amygdala, nucleus accumbens.
4
Cerebrovascular Disease (CbVD)
Volume of white matter hyperintensities on MRI.
Figure 1 illustrates the schematic representation of the Structural Equation Model. Latent variables are represented by circular nodes. DD_AUC uses monetary values ($200 and $40,000).
Data Preparation: Image Phenotypes
Data Processing Overview
Before analysis, neuroimaging data undergo crucial transformations. This ensures accuracy and comparability across different brain metrics.
Unnormalized Data
Raw data for age, gender, education, and BMI are not normalized. This preserves their original distribution and relationships.
Normalized Volumetric Measurements
Only volumetric measurements are normalized. This prevents parameter estimate bias. It ensures robust statistical inferences.
Figure 2 shows the distributions of transformed variables. These include subcortical volume, CbVD, cortical volume, and cortical thickness. This visual representation aids in understanding data characteristics.
Reference: Association Between Midlife Obesity and Its Metabolic Consequences, Cerebrovascular Disease, and Cognitive Decline (Clinical Research Article, 2021).
Key Takeaway Points
Our validated hypothesis, based on current data samples, reveals several crucial findings. These insights shed light on BMI's influence on brain structure and decision-making.
No Mediation by Average Brain Structures
Brain structure average measures do not significantly mediate the relationship between BMI and DD.
Zero Effects from Averaging Volumes
Averaging sub-cortical regions yields zero effects. Specific region effects might be nullified by others.
Inverse Direct Effect of BMI on DD_AUC
Higher BMI leads to lower DD_AUC, indicating increased impulsivity.
Education's Positive Impact
Higher education years correlate with higher DD_AUC, suggesting less impulsive behavior.
BMI's Effect on Cortical Thickness
BMI significantly affects cortical thickness (posterior probability = 0.956).
Made with