Large-scale ensemble simulations of biomathematical brain arteriovenous malformation models using graphics processing...

Area of Science:

• Biomedical Engineering
• Computational Fluid Dynamics
• Medical Imaging

Background:

• Current theoretical models for cerebral arteriovenous malformation (AVM) hemodynamics lack clinical realism due to oversimplification.
• Advanced computational power is needed to model the complex angioarchitecture of AVM nidus.

Purpose of the Study:

• To develop a more realistic and clinically representative computational model of cerebral AVM hemodynamics.
• To utilize graphics processing unit (GPU) computing for simulating complex AVM nidus architectures with thousands of vessels.

Main Methods:

• A theoretical electrical circuit AVM model was constructed using a stochastic block model (SBM) to represent the nidus with approximately 1000 vessels.
• Ensemble simulations of 10,000 distinct nidus morphologies were performed, assigning biophysical values and calculating hemodynamic parameters like mean pressure and flow rate.
• Network analysis was employed to derive an expression for nidus rupture risk, followed by a sensitivity analysis of model parameters.

Main Results:

• Simulations showed total intranidal volumetric blood flow averaging 463 mL/min (range: 268–535 mL/min).
• The maximum vessel rupture risk within the nidus averaged 29% (range: 0%–60%).

Conclusions:

• The developed biomathematical AVM model, enabled by GPU computing, offers a practical tool for theoretical research.
• This model facilitates investigations into AVM therapies and their hemodynamic consequences, enhancing clinical understanding and treatment planning.