Learning reduced-order models for cardiovascular simulations with graph neural networks

Luca Pegolotti; Martin R Pfaller; Natalia L Rubio; Ke Ding; Rita Brugarolas Brufau; Eric Darve; Alison L Marsden

doi:10.1016/j.compbiomed.2023.107676

Learning reduced-order models for cardiovascular simulations with graph neural networks

Comput Biol Med. 2024 Jan:168:107676. doi: 10.1016/j.compbiomed.2023.107676. Epub 2023 Nov 19.

Authors

Luca Pegolotti¹, Martin R Pfaller², Natalia L Rubio³, Ke Ding⁴, Rita Brugarolas Brufau⁴, Eric Darve⁵, Alison L Marsden⁶

Affiliations

¹ Department of Pediatrics, Stanford University, United States of America; Institute for Computational and Mathematical Engineering, Stanford University, United States of America. Electronic address: lpego@stanford.edu.
² Department of Pediatrics, Stanford University, United States of America; Institute for Computational and Mathematical Engineering, Stanford University, United States of America.
³ Department of Mechanical Engineering, Stanford University, United States of America.
⁴ Intel Corporation, United States of America.
⁵ Institute for Computational and Mathematical Engineering, Stanford University, United States of America; Department of Mechanical Engineering, Stanford University, United States of America.
⁶ Department of Pediatrics, Stanford University, United States of America; Institute for Computational and Mathematical Engineering, Stanford University, United States of America; Department of Mechanical Engineering, Stanford University, United States of America; Department of Bioengineering, Stanford University, United States of America.

PMID: 38039892
PMCID: PMC10886437 (available on 2025-01-01)
DOI: 10.1016/j.compbiomed.2023.107676

Abstract

Reduced-order models based on physics are a popular choice in cardiovascular modeling due to their efficiency, but they may experience loss in accuracy when working with anatomies that contain numerous junctions or pathological conditions. We develop one-dimensional reduced-order models that simulate blood flow dynamics using a graph neural network trained on three-dimensional hemodynamic simulation data. Given the initial condition of the system, the network iteratively predicts the pressure and flow rate at the vessel centerline nodes. Our numerical results demonstrate the accuracy and generalizability of our method in physiological geometries comprising a variety of anatomies and boundary conditions. Our findings demonstrate that our approach can achieve errors below 3% for pressure and flow rate, provided there is adequate training data. As a result, our method exhibits superior performance compared to physics-based one-dimensional models while maintaining high efficiency at inference time.

Keywords: Cardiovascular modeling; Graph neural networks; Reduced-order models.

MeSH terms

Cardiovascular System*
Computer Simulation
Hemodynamics
Models, Cardiovascular
Neural Networks, Computer*

Abstract

MeSH terms

Grants and funding