Accelerating massively parallel hemodynamic models of coarctation of the aorta using neural networks

Abstract

Comorbidities such as anemia or hypertension and physiological factors related to exertion can influence a patient’s hemodynamics and increase the severity of many cardiovascular diseases. Observing and quantifying associations between these factors and hemodynamics can be difficult due to the multitude of co-existing conditions and blood flow parameters in real patient data. Machine learning-driven, physics-based simulations provide a means to understand how potentially correlated conditions may affect a particular patient. Here, we use a combination of machine learning and massively parallel computing to predict the effects of physiological factors on hemodynamics in patients with coarctation of the aorta. We first validated blood flow simulations against in vitro measurements in 3D-printed phantoms representing the patient’s vasculature. We then investigated the effects of varying the degree of stenosis, blood flow rate, and viscosity on two diagnostic metrics – pressure gradient across the stenosis (ΔP) and wall shear stress (WSS) - by performing the largest simulation study to date of coarctation of the aorta (over 70 million compute hours). Using machine learning models trained on data from the simulations and validated on two independent datasets, we developed a framework to identify the minimal training set required to build a predictive model on a per-patient basis. We then used this model to accurately predict ΔP (mean absolute error within 1.18 mmHg) and WSS (mean absolute error within 0.99 Pa) for patients with this disease.

Clinically-Driven Virtual Patient Cohorts Generation: An Application to Aorta

“…Even with low resolution models, this process can require a computation time in the range of minutes to hours on a modern workstation. Machine learning and deep neural networks are already being used to accelerate different processes related to simulation of hemodynamics in the aorta or perform diagnosis (Xiao et al, 2016 ; Liang et al, 2018 , 2020 ; Feiger et al, 2020 ). We show that, in the generation of virtual patients cohorts, machine learning can replace the evaluation of acceptance functions with high accuracy.…”

The Critical Role of Lumped Parameter Models in Patient-Specific Cardiovascular Simulations

“…With the exciting advances in artificial intelligence (AI) being applied to cardiology, this marks one of the next chapters for cardiovascular simulation tools. The integration of these advanced methods has the potential to render LPM and multiscale frameworks even more personalized, reduce the needed computation time and provide clinicians and scientists with previously inaccessible data and trends [190][191][192][193][194].…”

An efficient, localised approach for the simulation of elastic blood vessels using the lattice Boltzmann method

“…Implementation of this can be time consuming in terms of both computational development and implementation as well as execution time. For this reason, a number of studies will fall back on a rigid wall assumption due to its simplicity [4,5,6]. In Table 1 we compare the average time taken to complete 1000 iterations for our elastic wall implementation and the same domain and core configuration using the rigid wall assumption.…”

“…Depending on the fluid solver used, the changing wall position may demand the fluid domain to be modified in response. Both of these procedures can be complex and computationally costly to conduct in 3D and may be a reason as to why a number of 3D models utilise a rigid wall assumption [4,5,6].…”