Detection of arterial wall abnormalities via Bayesian model selection.

Area of Science:

• Computational fluid dynamics
• Biomedical engineering
• Bayesian inference

Background:

• Patient-specific hemodynamic modeling is crucial for understanding arterial diseases.
• Current models often rely on simplified or computationally expensive approaches.
• Accurate parameter estimation and model selection are key challenges.

Purpose of the Study:

• To develop a Bayesian uncertainty quantification framework for complex hemodynamic models.
• To enable efficient parameter estimation in large-scale arterial network simulations.
• To provide a system for evidence-based selection between different physical models.

Main Methods:

• Implemented an efficient parallel Bayesian framework for parameter estimation.
• Developed a practical model selection system for comparing distinct physical models.
• Utilized simulated noisy flow velocity data from a branching arterial tree model.

Main Results:

• The methodology accurately located a structural defect in a simulated arterial tree.
• Physical properties of the abnormality were estimated reliably despite significant observational and systemic errors.
• The framework demonstrated efficient parameter estimation for complex forward models.

Conclusions:

• The proposed Bayesian framework enhances patient-specific hemodynamic modeling capabilities.
• It offers a robust approach for identifying and characterizing abnormalities in arterial networks.
• The method shows significant potential for real-world clinical applications in hemodynamic flow analysis.