Establishing group-level brain structural connectivity incorporating anatomical knowledge under latent space modeling

Area of Science:

• Neuroimaging
• Computational Neuroscience
• Network Science

Background:

• Brain structural connectivity, derived from diffusion MRI, characterizes anatomical organization.
• Current group-level connectivity analyses often oversimplify by ignoring network topology and connection associations.
• A need exists for advanced methods to model group brain connectivity that incorporate biological and anatomical information.

Purpose of the Study:

• To develop a novel latent space-based generative network model for estimating group-level brain connectivity.
• To integrate anatomical attributes of brain regions into the connectivity model for enhanced plausibility and biological interpretation.
• To introduce the attributes-informed brain connectivity (ABC) model for superior group-level connectivity estimation.

Main Methods:

• Proposed a latent space-based generative network model incorporating anatomical node attributes.
• Developed a Bayesian Markov Chain Monte Carlo (MCMC) algorithm for model estimation.
• Evaluated the ABC model through extensive simulations and application to Alzheimer's Disease (AD) data.

Main Results:

• The ABC model provides an interpretable latent space for group-level connectivity.
• Incorporation of anatomical knowledge improved connectivity estimation and revealed co-varying relationships.
• The model demonstrated superior predictive power for out-of-sample structural connectivity.
• Identified meaningful sex-specific network neuromarkers in AD subjects.

Conclusions:

• The attributes-informed brain connectivity (ABC) model offers a significant advancement over traditional methods for group-level brain connectivity analysis.
• The ABC model enhances biological interpretability by integrating anatomical data.
• This approach successfully identified sex-specific neuroimaging biomarkers relevant to Alzheimer's Disease.