Machine learning-augmented finite element modeling for transient hemodynamics in human arteries.

Area of Science:

• Biomedical Engineering
• Computational Fluid Dynamics
• Nanotechnology

Background:

• Nanoparticles show promise for targeted drug delivery and cancer therapy.
• Understanding blood flow dynamics with nanoparticles is crucial for biomedical applications.
• Magnetohydrodynamics offers potential for targeted treatment mechanisms.

Purpose of the Study:

• To develop a machine learning-enhanced finite element model for blood flow in stretching arteries with sulfonated polystyrene nanoparticles (NSPS).
• To investigate the influence of key parameters on NSPS behavior for clinical applications.
• To create a computationally efficient framework for exploring parameter spaces in drug delivery simulations.

Main Methods:

• Governing equations solved using the finite element method (FEM).
• Numerical simulations performed in Python to determine velocity, temperature, and concentration fields.
• A multilayer perceptron regression surrogate model trained on FEM data for predictive analysis.

Main Results:

• The hybrid FEM-ML model achieved 97.3% predictive accuracy.
• Increased magnetic field strength reduced axial velocity by approximately 18%.
• Improved porosity increased drug penetration depth by approximately 11%.

Conclusions:

• The hybrid FEM-ML framework offers a fast computational tool for parameter space exploration, reducing reliance on expensive FEM simulations.
• NSPS show potential for localized drug release triggered by hyperthermia or pH changes in tumor environments.
• This approach enhances targeted drug action while minimizing side effects on healthy tissues.