Temporal averaging of two-dimensional correlation functions for velocity vector imaging of cardiac blood flow

Ultrasonic imaging of blood flow in the cardiac lumen is a very useful tool to evaluate the pumping function of the human heart. The speckle tracking technique makes it possible to estimate the blood velocity vector. However, a stable estimation of the velocity vector of blood flow is difficult because signal-to-noise ratios of echoes from tiny blood particles are low. In this study, the speckle tracking technique with averaging of multiple two-dimensional correlation functions was employed for stable estimation of the blood velocity vector.

Multiple two-dimensional correlation functions can be averaged during a very short period by using the echo data acquired by high-frame-rate echocardiography with diverging beam transmission. A steady flow experiment using blood-mimicking fluid (mean fluid velocity 0.2 m/s, flow angle 56° from the transducer surface) was implemented to investigate the effect of the averaging of two-dimensional correlation functions at a frame rate of 6024 Hz.

First, to examine the averaging duration required for stable estimation of the flow velocity vector, the accuracies of vector estimates were evaluated at different durations for averaging of two-dimensional correlation functions in the steady flow measurement. It was found that the proposed averaging process with an averaging duration of over 8 frames could reduce the directional error in vector estimation to almost half that of the conventional speckle tracking technique. In subsequent experiments, the averaging duration was set at 12 frames corresponding to 2 ms. Measurements of steady flow at higher velocities were further implemented. The steady flow measurements with higher flow velocities of 0.4 and 0.6 m/s were simulated by changing the frame interval of the echo data at a flow velocity of 0.2 m/s. Although the averaging duration was a mere 2 ms, directional errors at mean flow velocities of 0.2, 0.4, and 0.6 m/s were reduced significantly. In an in vivo experiment of the healthy human heart, to produce a fine B-mode image, the diverging wave transmissions with different steered angles for compounding were interleaved in the transmission sequence. From the in vivo experimental result, the blood velocity vector of the left ventricular cavity showed the flow getting into/out of the cavity in ejection and early diastolic phases. Furthermore, estimated flow directions revealed rotating flow in the cavity in mid-diastole.

Our proposed method has the feasibility to visualize the vortex flow by velocity vector mapping without a contrast agent.