- The paper presents DoVeR, a method that reconstructs two-dimensional LV blood flow by solving a Poisson equation for streamfunction and vorticity.
- It enforces mass conservation and accommodates complex flow rotations via iterative vorticity refinement and adaptive boundary conditions.
- DoVeR demonstrates superior accuracy with lower nRMSE and better noise robustness compared to traditional iVFM methods in both synthetic and in-vivo validations.
Streamfunction-Vorticity Reconstruction for Color Doppler Echocardiography: The DoVeR Method
Introduction
The reconstruction of blood flow velocity vector fields within the left ventricle (LV) using ultrasound has substantial implications for the quantitative analysis of cardiac hemodynamics. Despite decades of development, color Doppler ultrasound, while ubiquitous, presents only a one-dimensional projection of inherently three-dimensional cardiac flow, complicating the accurate characterization of intraventricular flow structures. Existing approaches such as echoPIV, blood speckle imaging (BSI), and various forms of intraventricular vector flow mapping (iVFM) have limitations relating to noise robustness, sensitivity to operator technique (especially probe placement), and inaccuracies in treating flow rotation and mass conservation.
This paper introduces Doppler Velocity Reconstruction (DoVeR), a methodology leveraging the streamfunction-vorticity ψ-ω formulation, which addresses these gaps by ensuring mass conservation and explicitly accounting for flow rotation.
Methodological Advances
DoVeR reformulates color Doppler velocity vector field reconstruction as the solution of a Poisson equation linking vorticity and streamfunction. The measured Doppler component (uy​) remains unaltered, while DoVeR solves for the unknown in-plane component (ux​) via iterative refinement of vorticity. Key mathematical underpinnings include the enforcement of divergence-free velocity fields (mass conservation) and the inherent accommodation of vorticity production, aligning the method with the physical principles of intraventricular flow.
Boundary conditions are set with a combination of color Doppler (CD) and pulsed-wave Doppler (PWD) data. PWD measurements at the mitral and aortic valves provide physically realistic in/outlet fluxes, while wall conditions are imposed via a piece-wise linear model. Unlike standard iVFM approaches, DoVeR does not rely on restrictive assumptions such as free-slip or axisymmetry at boundaries; instead, it generalizes BCs for arbitrary probe orientations and domain geometries.
The numerical solution utilizes LU decomposition of the discretized Poisson equation, with vorticity and velocity components repeatedly updated until a strict convergence criterion is satisfied. Additionally, DoVeR employs a tailored scheme for Doppler data de-aliasing and noise handling, employing thresholding and selective replacement for Doppler velocities subjected to machine-side high-pass filtering.
Comparative Analysis and Results
Synthetic Data Benchmarks
The DoVeR method was tested on artificial color Doppler data generated from CFD models of the LV. These benchmarks included systematic perturbations in probe placement, noise amplitude, and filtering strategies, providing a rigorous assessment of method robustness under clinically relevant variability.
Strong numerical results include:
- DoVeR produced normalized root mean square errors (nRMSE) for velocity magnitude between 3.81% and 6.67%, substantially lower and more consistent than iVFM1D (4.16%-24.17%) and iVFM2D (4.06%-400.21%). Notably, iVFM2D exhibited catastrophic failure (error > 400%) at some orientations.
- DoVeR was essentially agnostic to probe line density and signal filtering: improvements from filtering were negligible (<1% nRMSE change), and noise robustness remained high at all levels tested.
- Maximum direction errors for DoVeR did not exceed 53.15°, whereas for iVFM1D and iVFM2D, errors reached 96.53° and 81.06°, respectively.
- Error CDFs confirmed that DoVeR halves the reconstruction error compared to iVFM overall, maintaining consistent performance across noise and orientation perturbations.
In-Vivo Small Animal Experiments
Application to murine LV ultrasound demonstrated that DoVeR outperformed iVFM reconstructions in the resolution of characteristic diastolic vortex structures and in suppressing nonphysical vorticity bands that typically manifest with iVFM. The method proved robust despite the use of shallow imaging depths and transducers atypical for clinical adult studies. DoVeR captured expected shear layers and vortex pairs more accurately, while both iVFM variants yielded over-smoothed or fragmented vortex reconstructions—particularly when dominant velocity components deviated from the transducer’s scan lines.
Discussion
DoVeR’s main advantage arises from its direct enforcement of the planar flow assumption and mass conservation through the Poisson PDE framework, as opposed to the line-integration (iVFM1D) or regularized least-squares (iVFM2D) strategies. Unlike prior algorithms that are orientation-dependent—failing when dominant flow components are not normal to the transducer face—DoVeR’s treatment of BCs and its iterative vorticity updating confer robustness to sector rotation and noise.
Filtering, which is essential to stabilize iVFM reconstructions, was found in this study to increase error for both iVFM1D and iVFM2D once beyond moderate kernel size, due to over-smoothing and consequent under-resolution of flow features. DoVeR, in contrast, showed negligible impact from filtering, indicating that its regularization is intrinsic to its numerical formulation.
Unlike prior approaches that either ignore or inadequately model vorticity, DoVeR’s explicit vorticity refinement supports accurate recovery of flow rotation. The iterative refinement is self-consistent and inherently regularizing, negating the need for tuning external regularization weights as in iVFM2D.
DoVeR’s main limitations relate to its reliance on a heuristic, piece-wise linear balance for wall fluxes, which may not capture the full complexity of time-varying wall motion. The method’s validation was restricted to synthetic and small animal datasets; however, the underlying flow physics are characteristic of both mouse and human ventricles.
Implications and Future Directions
This work demonstrates that streamfunction-vorticity approaches can substantially improve the reliability and accuracy of cardiac vector flow quantification from color Doppler ultrasound, particularly under non-ideal conditions (e.g., noisy, non-standard probe locations). Clinically, robust vector field reconstructions may improve the noninvasive assessment of cardiac energetics, filling dynamics, and flow-mediated pathologies.
The extension of DoVeR to three-dimensional imaging, as well as to datasets acquired at clinical imaging depths and with standard phased array transducers, remains an open research direction. Adaptive boundary schemes that more accurately represent wall kinematics, perhaps integrating speckle-tracking-derived wall motion, could further improve accuracy.
Conclusion
Doppler Velocity Reconstruction (DoVeR) presents a viable, physically consistent alternative to existing iVFM-based algorithms for two-component flow field reconstruction from color Doppler echocardiography. Its robustness to typical sources of error—noise, probe angle, and filtering—along with its rigorous enforcement of mass conservation and ability to resolve vorticity, position DoVeR as a more reliable means for quantifying LV flow fields in both preclinical and eventual clinical contexts. The work underscores the importance of physically justified mathematical models in ultrasound flow imaging and offers multiple axes for future methodological developments.