Key result
A machine learning framework trained on a minimal set of simulations accurately predicted the pressure gradient (mean absolute error within 1.18 mmHg, R = 0.99) and wall shear stress (mean absolute error within 0.99 Pa) in patients with coarctation of the aorta.
Why the study?
Observing and quantifying associations between comorbidities, physiological factors, and hemodynamics is difficult in real patient data due to multiple co-existing conditions and blood flow parameters.
Can machine learning models trained on massively parallel hemodynamic simulations accurately predict pressure gradient and wall shear stress in coarctation of the aorta?
Population
3D-printed phantoms representing patient vasculature and simulation models of coarctation of the aorta
Comparison
Varying degrees of stenosis, blood flow rate, and viscosity
Design
In vitro validation and machine learning-driven physics-based simulation study
Authors
Loading...
May enable rapid patient-specific hemodynamic prediction in coarctation; leaves open prospective human validation before clinical use.
Can machine learning models trained on massively parallel hemodynamic simulations accurately predict pressure gradient and wall shear stress in coarctation of the aorta?
Effect estimate: R = 0.99
Machine learning models trained on a minimal set of high-fidelity hemodynamic simulations can accurately and rapidly predict patient-specific pressure gradients and wall shear stress in coarctation of the aorta.
Feiger et al. (2020) studied Coarctation of the aorta (n=4). Machine learning models trained on minimal simulation sets vs. Full computational fluid dynamics simulations was evaluated on Prediction accuracy of pressure gradient (ΔP) (R = 0.99). A machine learning framework trained on a minimal set of simulations accurately predicted the pressure gradient (mean absolute error within 1.18 mmHg, R = 0.99) and wall shear stress (mean absolute error within 0.99 Pa) in patients with coarctation of the aorta.
Synapse has enriched 5 closely related papers on similar clinical questions. Consider them for comparative context: