Integrating multi-fidelity blood flow data with reduced-order data assimilation.

Area of Science:

• Biomedical Engineering
• Computational Fluid Dynamics
• Medical Imaging

Background:

• Accurate patient-specific modeling of cardiovascular flows and hemodynamics is crucial but challenging.
• In-vivo (4D flow MRI) and in-vitro methods face limitations like low resolution, noise, and parameter uncertainty.
• Quantifying near-wall hemodynamics, such as wall shear stress, presents additional experimental difficulties.

Purpose of the Study:

• To develop a computationally efficient data assimilation method for high-fidelity cardiovascular flow modeling.
• To overcome the limitations of low resolution, noise, and uncertainty in existing modeling approaches.
• To improve the quantification of near-wall hemodynamics in patient-specific cardiovascular models.

Main Methods:

• Proposed a reduced-order modeling Kalman filter (ROM-KF) combining a sequential Kalman filter with reduced-order modeling.
• Utilized dynamic mode decomposition (DMD) to provide a linear model for the reduced-order modeling component.
• Assessed accuracy using 1D Womersley flow, 2D idealized, and 3D patient-specific cerebral aneurysm models with synthetic data.

Main Results:

• The ROM-KF method demonstrated superior accuracy compared to both computational fluid dynamics (CFD) and synthetic experimental datasets.
• Successfully reconstructed near-wall hemodynamics, even when such data were absent in the experimental input.
• Significantly improved the quantification of near-wall hemodynamic parameters.

Conclusions:

• The ROM-KF method offers a computationally efficient and accurate approach for patient-specific cardiovascular flow modeling.
• This technique effectively addresses challenges associated with data resolution, noise, and uncertainty.
• ROM-KF shows significant potential for advancing the analysis and clinical application of cardiovascular hemodynamics.