Abstract:Investigating blood flow in the cardiovascular system is crucial for assessing cardiovascular health. Computational approaches offer some non-invasive alternatives to measure blood flow dynamics. Numerical simulations based on traditional methods such as finite-element and other numerical discretizations have been extensively studied and have yielded excellent results. However, adapting these methods to real-life simulations remains a complex task. In this paper, we propose a method that offers flexibility and can efficiently handle real-life simulations. We suggest utilizing the physics-informed neural network (PINN) to solve the Navier-Stokes equation in a deformable domain, specifically addressing the simulation of blood flow in elastic vessels. Our approach models blood flow using an incompressible, viscous Navier-Stokes equation in an Arbitrary Lagrangian-Eulerian form. The mechanical model for the vessel wall structure is formulated by an equation of Newton's second law of momentum and linear elasticity to the force exerted by the fluid flow. Our method is a mesh-free approach that eliminates the need for discretization and meshing of the computational domain. This makes it highly efficient in solving simulations involving complex geometries. Additionally, with the availability of well-developed open-source machine learning framework packages and parallel modules, our method can easily be accelerated through GPU computing and parallel computing. To evaluate our approach, we conducted experiments on regular cylinder vessels as well as vessels with plaque on their walls. We compared our results to a solution calculated by Finite Element Methods using a dense grid and small time steps, which we considered as the ground truth solution. We report the relative error and the time consumed to solve the problem, highlighting the advantages of our method.

What problem does this paper attempt to address?

The main problem that this paper attempts to solve is: **How to simulate the hemodynamic behavior in elastic blood vessels efficiently and accurately**. Specifically, the author proposes a mesh - free solver based on the Physics - Informed Neural Network (PINN) to solve the Navier - Stokes equations in deformable domains, thereby achieving the simulation of blood flow. ### Problem Background Traditional numerical methods such as the Finite Element Method (FEM) have achieved remarkable results in simulating blood flow, but still face challenges when dealing with complex geometries and real - life scenarios. For example, complex meshing and time - step settings can lead to high computational costs and may require frequent manual adjustments. In addition, traditional methods are prone to convergence problems when dealing with large - displacement structures. ### Solutions Proposed in the Paper To solve these problems, the paper proposes a PINN - based method, whose main features include: 1. **Mesh - free method**: By using PINN, the complex meshing and discretization steps in traditional methods are avoided, making the simulation more flexible and able to handle complex geometries efficiently. 2. **Efficient parallel computing**: By using existing open - source machine - learning frameworks and parallel modules, the computational efficiency can be significantly improved through GPU acceleration and parallel computing. 3. **Arbitrary Lagrangian - Eulerian (ALE) form**: The ALE form is adopted to describe the interaction between the flow and the structure, ensuring that the coupling conditions are met at the fluid - structure interface. 4. **Multi - objective optimization training strategy**: A sequential alternating training method is designed, which decomposes the problem into two sub - problems of fluid dynamics and deformation, and optimizes the corresponding loss functions respectively to achieve stable training and better convergence. ### Main Contributions 1. **Combining the Navier - Stokes equations with the ALE form into PINN**, which solves the fluid - structure interaction problem. 2. **Proposing the sequential alternating training method**, which achieves stable training and better convergence through a multi - objective optimization strategy. 3. **Introducing a new network architecture** to predict velocity and pressure separately, in order to better capture the characteristics of different variables in the Navier - Stokes equations. 4. **Improving the selection of activation functions**, by alternately using ReLU and Sigmoid activation functions, which improves the training speed and convergence performance. ### Experimental Verification To evaluate the effectiveness of the proposed method, the author conducted multiple experiments, including: - Conducting blood flow simulations in regular cylindrical blood vessels and comparing the results with those of the Finite Element Method under high - resolution meshes. - Conducting blood flow simulations in complex blood vessels with plaques to verify the robustness of the method on complex geometries. - Conducting self - ablation experiments to study the influence of different network depths and widths on the results. - Evaluating the effect of multi - GPU acceleration. - Analyzing the influence of plaques of different sizes on hemodynamics. In summary, this paper aims to provide an efficient, flexible, and suitable for complex geometries blood flow simulation method through PINN to overcome the limitations of traditional numerical methods in practical applications.

A Meshless Solver for Blood Flow Simulations in Elastic Vessels Using Physics-Informed Neural Network

A Novel Physics-Based Model for Fast Computation of Blood Flow in Coronary Arteries

Physics-constrained coupled neural differential equations for one dimensional blood flow modeling

Enhancing Arterial Blood Flow Simulations through Physics-Informed Neural Networks

Physics-informed neural networks (PINNs) for 4D hemodynamics prediction: An investigation of optimal framework based on vascular morphology

Physics-Informed Graph Neural Networks to solve 1-D equations of blood flow

Uncovering near-wall blood flow from sparse data with physics-informed neural networks

Machine learning in cardiovascular flows modeling: Predicting arterial blood pressure from non-invasive 4D flow MRI data using physics-informed neural networks

Mesh Neural Networks for SE(3)-Equivariant Hemodynamics Estimation on the Artery Wall

Physics driven reduced order model for real time blood flow simulations

Physics-informed neural networks with domain decomposition for the incompressible Navier–Stokes equations

226 Advancing Neurovascular Diagnostics for Abnormal Hemodynamic Conditions Through AI-Driven Physics-informed Neural Networks

Physics-informed neural networks (PINNs) for fluid mechanics: A review

Physics-informed neural networks for blood flow inverse problems

A one-dimensional computational model for blood flow in an elastic blood vessel with a rigid catheter

Mesh Convolutional Neural Networks for Wall Shear Stress Estimation in 3D Artery Models

On Physics-Informed Deep Learning for Solving Navier-Stokes Equations

Applying Physics-Informed Neural Networks to Solve Navier–Stokes Equations for Laminar Flow around a Particle

Enhancing physics informed neural networks for solving Navier–Stokes equations