Physics-informed neural networks for parameter estimation in blood flow models

• Published in: Computers in biology and medicine 2024-08, Vol.178, p.108706, Article 108706
• Main Authors: Garay, Jeremías, Dunstan, Jocelyn, Uribe, Sergio, Sahli Costabal, Francisco
• Format: Article
• Language: English
• Subjects: Aorta; Aorta - physiology; Blood flow; Blood Flow Velocity - physiology; Complexity; Computer Simulation; Flow distribution; Hemodynamics; Hemodynamics - physiology; Humans; Information systems; Inverse problems; Kalman filters; Models, Cardiovascular; Neural networks; Neural Networks, Computer; Parameter estimation; Parameter robustness; Patient-specific model; Physics; Physics-informed neural networks; Reduced order models; Reduced-order modeling; Scattering; Two dimensional flow; Velocity; Velocity distribution
• ISSN: 0010-4825; 1879-0534; 1879-0534
• DOI: 10.1016/j.compbiomed.2024.108706

Physics-informed neural networks (PINNs) have emerged as a powerful tool for solving inverse problems, especially in cases where no complete information about the system is known and scatter measurements are available. This is especially useful in hemodynamics since the boundary information is often difficult to model, and high-quality blood flow measurements are generally hard to obtain. In this work, we use the PINNs methodology for estimating reduced-order model parameters and the full velocity field from scatter 2D noisy measurements in the aorta. Two different flow regimes, stationary and transient were studied. We show robust and relatively accurate parameter estimations when using the method with simulated data, while the velocity reconstruction accuracy shows dependence on the measurement quality and the flow pattern complexity. Comparison with a Kalman filter approach shows similar results when the number of parameters to be estimated is low to medium. For a higher number of parameters, only PINNs were capable of achieving good results. The method opens a door to deep-learning-driven methods in the simulations of complex coupled physical systems. •Physics-informed neural networks can be used for solving forward and inverse problems.•Clinical-relevant inverse problems are extremely difficult due to the lack of models and reliable data.•Estimating reduced-order parameters can be done using PINNs.•Data quality strongly affects the results.•The results can be enhanced by using a finite element solver a post-process layer.