A Neural Network Finite Element Trileaflet Heart Valve Model Incorporating Multi-Body Contact

Area of Science:

• Computational mechanics
• Biomedical engineering
• Machine learning applications

Background:

• Patient-specific computational modeling is vital for cardiovascular disease care.
• Current finite element method (FEM) simulations are too slow for clinical use.
• Accelerated simulations are needed for optimal therapeutic approach determination.

Purpose of the Study:

• Develop and validate a Neural Network Finite Element (NNFE) approach for rapid soft tissue organ simulation.
• Extend NNFE to simulate trileaflet heart valve closure, including multi-body contact.
• Enable patient-specific computational modeling for improved cardiovascular disease treatment.

Main Methods:

• Utilized NNFE with conventional FEM meshes and GPU-based software for hyperelasticity PDEs.
• Extended NNFE to simulate 3D solid heart valve leaflets with multi-body contact.
• Verified NNFE against tIGAr for single leaflet closure, analyzing nodal displacement error and collagen fiber effects.

Main Results:

• NNFE achieved a ~100-fold speedup compared to FEM (0.28s vs. 61s for single leaflet closure).
• Full trileaflet valve simulations with contact took ~5s (vs. hours for FEM).
• Average nodal displacement error was 0.020mm (0.47%), with physiologically accurate deformation patterns.

Conclusions:

• NNFE successfully simulates complex 3D soft organ systems like trileaflet heart valves with large deformations and multi-body contact.
• The method offers rapid post-training simulations, promising for patient-specific predictive models.
• NNFE advances machine learning in computational mechanics for clinical applications.