The Effect of Spatial Velocity Gradients on Block-Matching Accuracy for Ultrasound Velocimetry

OBJECTIVE

Block matching serves as the foundation for ultrasound velocimetry techniques such as blood speckle tracking and echo-particle image velocimetry. Any spatial velocity gradients (SVGs) inside a block-matching pair will result in tracking error, due to both the finite block size and the ultrasound point-spread-function. We assess, using an in silico sinusoidal flow phantom, the effect of SVG magnitude and beam-to-flow angle on block-matching bias and precision. Secondarily we assess the effect that SVGs have on velocimetry bias when using angled plane-wave compounding.

METHODS

The magnitude and angle of SVGs were varied by adjusting the wavelength and direction of a sinusoidal flow profile. Scatterers displaced by this flow profile were used for simulating ultrasound radio frequency data at discrete time points. After beamforming, the 2-D flow field was estimated using block matching. Two imaging sequences were tested, a single plane-wave and a three-angled plane-wave.

RESULTS

Smaller sinusoidal flow wavelengths resulted in increased bias and reduced precision, revealing an inverse relationship between sinusoidal flow wavelength and tracking error, with median errors ranging from 69%-90% for the smallest flow wavelengths (highest SVGs) down to 3%-5% for the largest (lowest SVGs). The SVG angle was also important, in which lateral SVGs (with axially oriented flows) resulted in significant speckle decorrelation and high tracking errors in regions with high SVGs. Conversely, axial SVGs (laterally oriented flow) experienced higher bias in the peak velocity regions of the flow profile. Coherent compounding resulted in higher velocity errors than using a single transmission for lateral SVGs but not for axial SVGs.

CONCLUSION

The highest SVGs that could be measured with ≤10% error was when the sinusoidal flow wavelength was less than 20 times the ultrasound pulse wavelength. The clinical significance is that the high SVGs present in high kinetic energy flows, such as severe carotid stenosis and aortic regurgitation, will limit the ability to accurately quantify the velocities in these flow structures.

Contrast Agent-Free Assessment of Blood Flow and Wall Shear Stress in the Rabbit Aorta using Ultrasound Image Velocimetry

Blood flow velocity and wall shear stress (WSS) influence and are influenced by vascular disease. Their measurement is consequently useful in the laboratory and clinic. Contrast-enhanced ultrasound image velocimetry (UIV) can estimate them accurately but the need to inject contrast agents limits utility. Singular value decomposition and high-frame-rate imaging may render contrast agents dispensable. Here we determined whether contrast agent-free UIV can measure flow and WSS. In simulation, accurate measurements were achieved with a signal-to-noise ratio of 13.5 dB or higher. Signal intensity in the rabbit aorta increased monotonically with mechanical index; it was lowest during stagnant flow and uneven across the vessel. In vivo measurements with contrast-free and contrast-enhanced UIV differed by 4.4% and 1.9% for velocity magnitude and angle and by 9.47% for WSS. Bland-Altman analysis of waveforms revealed good agreement between contrast-free and contrast-enhanced UIV. In five rabbits, the root-mean-square errors were as low as 0.022 m/s (0.81%) and 0.11 Pa (1.7%). This study indicates that with an optimised protocol, UIV can assess flow and WSS without contrast agents. Unlike contrast-enhanced UIV, contrast-free UIV could be routinely employed.

4-D Echo-Particle Image Velocimetry in a Left Ventricular Phantom

Left ventricular (LV) blood flow is an inherently complex time-varying 3-D phenomenon, where 2-D quantification often ignores the effect of out-of-plane motion. In this study, we describe high frame rate 4-D echocardiographic particle image velocimetry (echo-PIV) using a prototype matrix transesophageal transducer and a dynamic LV phantom for testing the accuracy of echo-PIV in the presence of complex flow patterns. Optical time-resolved tomographic PIV (tomo-PIV) was used as a reference standard for comparison. Echo-PIV and tomo-PIV agreed on the general profile of the LV flow patterns, but echo-PIV smoothed out the smaller flow structures. Echo-PIV also underestimated the flow rates at greater imaging depths, where the PIV kernel size and transducer point spread function were large relative to the velocity gradients. We demonstrate that 4-D echo-PIV could be performed in just four heart cycles, which would require only a short breath-hold, providing promising results. However, methods for resolving high velocity gradients in regions of poor spatial resolution are required before clinical translation.