Skip to main content
NCBI home page
NIHPA Author Manuscripts logoLink to NIHPA Author Manuscripts
. Author manuscript; available in PMC: 2021 Sep 1.
Published in final edited form as: Ultrasound Med Biol. 2020 Jun 30;46(9):2560–2571. doi: 10.1016/j.ultrasmedbio.2020.05.022

3D intravascular characterization of blood flow velocity fields with a forward-viewing 2D array

Brooks D Lindsey 1,2,3,*, Bowen Jing 1, Saeyoung Kim 3,4, Graham C Collins 1, Muralidhar Padala 3,5,6
PMCID: PMC7429285  NIHMSID: NIHMS1601531  PMID: 32616428

Abstract

Risk stratification in coronary artery disease is an ongoing challenge for which few tools are available for quantifying physiology within coronary arteries. Recently, anatomy-driven computational fluid dynamic (CFD) modeling has enabled the mapping of local flow dynamics in coronary stenoses, with derived parameters such as wall shear stress demonstrating a strong capability for predicting adverse clinical events on a patient-specific basis. As cardiac catheterization is common in patients with coronary artery disease, minimally-invasive technologies capable of identifying pathological flow in situ in real time could have a significant impact on clinical decision-making. As a step towards in vivo quantification of slow flow near the arterial wall, proof-of-concept for 3D intravascular imaging of blood flow dynamics is demonstrated using a 118-element forward-viewing ring array transducer and a research ultrasound system. Blood flow velocity components are estimated in the direction of primary flow using an unfocused wave Doppler approach and in the lateral and elevation directions estimated using a transverse oscillation approach. This intravascular 3D vector velocity system is demonstrated by acquiring real-time 3D data sets in phantom experiments and in vivo in the femoral artery of a pig. The effect of the catheter on blood flow dynamics is also experimentally assessed in flow phantoms with both straight and stenotic vessels. Results indicate that 3D flow dynamics can be measured using a small form factor device, and that a hollow catheter design may provide minimal disturbance to flow measurements in a stenosis (peak velocity: 54.97 ± 2.13 cm/s without catheter vs. 51.37 ± 1.08 cm/s with hollow catheter, 6.5% error). In the future, such technologies could enable estimation of 3D flow dynamics near the wall in patients already undergoing catheterization.

Keywords: intravascular ultrasound, matrix array, ring array, forward viewing transducer, transverse oscillation, vector velocity, 3D ultrasound

Introduction

Eight million people in the U.S. suffer from stable coronary artery disease (CAD) (Benjamin, et al. 2019), a heterogeneous population with annual mortality rates varying by a factor of 6 between “low risk” and “high risk” stable CAD patients (Steg, et al. 2007). Diagnostic catheterization is common in these patients to perform catheter coronary angiography (CCA) and measure fractional flow reserve (FFR), however, 12% of patients with initial FFR > 0.80 will ultimately undergo percutaneous coronary intervention (PCI) and 6% will have a myocardial infarction (MI) within two years (Lee, et al. 2017). In a recent post-hoc analysis of a subset of 441 patients receiving medical therapy for stable CAD as part of the FAME 2 trial, increased wall shear stress (WSS) in the proximal segment of the atherosclerotic lesion was predictive of myocardial infarction (Kumar, et al. 2018). Such additional physiological insight could drive intervention in patients with FFR > 0.80 likely to experience MI, enabling risk stratification in the heterogeneous stable CAD population already undergoing catheterization. However, there is currently no method for measuring blood flow dynamics at the coronary arterial wall in order to estimate WSS.

The ability to estimate blood flow velocity fields near the wall of the artery and compute corresponding WSS maps in real-time during catheterization requires a catheter-based transducer capable of 3D + time measurements and an approach for 3D velocity estimation. The most common techniques for ultrasound-based blood flow velocity estimation are Doppler-based approaches, which quantify the component of blood flow along the ultrasound beam. In order to quantify blood flow components in directions orthogonal to the beam, Jensen et al., have presented the transverse oscillation method (Udesen and Jensen 2006). This method has been implemented in 3D using large, non-invasive matrix array transducers to estimate blood flow velocity fields in 3D (Pihl and Jensen 2014, Pihl, et al. 2014). More recently, Lenge et al. have proposed a frequency domain implementation of transverse oscillation (Lenge, et al. 2014). Alternatively, Correia et al. have performed real-time 3D ultrafast imaging at rates >4000 volumes per second using a non-invasive matrix array in the carotid arteries (Correia, et al. 2016). While these approaches allow non-invasive imaging of 3D blood flow velocity fields, spatial resolution at depth and physiological motion limit non-invasive ultrasound vector velocity imaging approaches to chambers of the heart and large, superficial vessels, as even assessment of 3D blood flow dynamics in the ascending aorta (3 to 7× larger than coronary arteries) requires a large inter-operative probe (Hansen, et al. 2017).

In recent years, several researchers have proposed new devices and systems for intravascular blood flow velocity estimation, including a catheter with one forward-viewing and one side-viewing transducer (Hong, et al. 2019), and a poly(vinylidene fluoride-trifluoroethylene) copolymer-based flexible implantable sensor (Cannata, et al. 2012). However, these techniques don’t allow measurement of vector velocity fields. In the catheterization lab, the capability for 3D measurement of vector velocity fields could be powerful due to the additional spatial information provided relative to the single pressure drop measurement provided by FFR. Performing minimally-invasive vector velocity measurements presents additional systems challenges, namely the need for a small form factor matrix array transducer and the corresponding decrease in signal-to-noise ratio (SNR) due to the high electrical impedance of the small elements in a piezoelectric matrix array transducer (Lindsey, et al. 2011).

While measurements of velocity and pressure are routinely made inside arteries using Doppler wires and pressure-sensitive FFR wires, the ability to quantify blood flow velocity fields in 3D could enable estimation of WSS relative to current measurements which only provide spatially-averaged hemodynamic information at a single discrete location. In this article, we present the initial proof-of-concept demonstration for 3D intravascular blood flow velocity estimation as a step towards development of minimally-invasive catheter-based systems for real-time 3D measurement of velocity vectors in coronary and peripheral arteries. Velocity fields are estimated using a 4 mm-diameter 2D array transducer in flow phantom experiments and in vivo by imaging through the wall of a femoral artery. The effect of the catheter on blood flow dynamics is also preliminarily investigated to drive future design of intravascular catheter-based measurement technology. To our knowledge, this approach utilizing a forward-viewing ring array transducer and transverse oscillation processing represents the first high frame rate 3D vector velocity imaging in an intravascular form factor.

Materials and Methods

Imaging system, transducer, and processing

A 4 mm-diameter, 118 element double ring array transducer (Fig. 1) described previously (Smith, et al. 2014) was connected to a research imaging system (Verasonics Vantage 256, Kirkland, WA, USA) using a custom printed circuit board. This transducer operates at 5 MHz and is used to acquire 7 unfocused, steered transmit events at a user-selected rate, resulting in the point spread function shown in Fig. 1C at a depth of z=12 mm from the face of the transducer. Briefly, all 118 elements are used to transmit a 1-cycle pulse, which is repeated seven times for steering angles from −3.67° to 3.67° in both the elevational and azimuth directions of the array. After the data was beamformed using a conventional delay-and-sum approach, the beamformed data are coherently compounded prior to Doppler processing (Montaldo, et al. 2009). While a higher transmit frequency is preferred for IVUS and will be pursued in future transducer designs, the increased penetration depth of the current 5 MHz device allows the transducer to be placed further from the region of interest to minimize the alteration of the flow dynamics in this region.

Figure 1.

Figure 1.

Open in a new tab

(A) En face photograph of dual ring array transducer (scale bar = 1 mm). (B) Side view of the same 4 mm-diameter transducer. (C) Simulated point spread function (PSF) for the transducer used in this work operating with the described pulse sequence with 7 steered, unfocused transmits at a depth of 12 mm from the face of the transducer. The −6 dB beam width is 0.57 mm. The peak side lobe level is −21.8 dB. All 7 angles were used to form a single frame as in (Montaldo, et al. 2009) and (Tierney, et al. 2017). Use of compounding improves spatial resolution relative to the single transmit case, in which −6 dB beam width is 0.71 mm at 12 mm.

As previously described, transverse oscillation can be implemented efficiently in the frequency domain by bandpass filtering acquired data in the lateral (and elevation) directions (Lenge, et al. 2014, Jensen, et al. 2015, Salles, et al. 2015). Velocity components orthogonal to the beam were estimated by transverse oscillation, implemented in both lateral and elevation directions in the frequency domain using a bandpass filter centered at 1.6 cycles/mm with −6 dB bandwidth of 93%. Filtering of beamformed RF data acquired by the array in the lateral direction to introduce transverse oscillation centered at 1.6 cycles/mm is shown in Fig. 2. The result of this filtering operation is a set of filtered, beamformed RF data containing transverse oscillations centered at this lateral spatial frequency without affecting the axial dimension. After these filtering operations in x and y (i.e. lateral and elevation directions), the result is complex, allowing flow direction to be determined in both directions. A standard fourth-order autocorrelation estimator was used to determine the lateral and elevation velocity components vx and vy (Jensen 2001). The axial velocity component vz is determined using a standard cross-correlation estimator after filtering stationary echoes with a high pass filter (cutoff frequencies described below) and an ensemble size of 50. In these studies, we are primarily interested in estimating slow velocities at the wall, thus the pulse repetition frequency (PRF) and cutoff frequency are selected accordingly, though a dealiasing approach (Wigen, et al. 2018) was used to rapidly identify and correct aliased pixels at the center of the vessel in order to display the correct orientation without searching all pixels.

Figure 2.

Figure 2.

Open in a new tab

Image showing two-dimensional frequency domain transverse oscillation applied to beamformed RF data at z=12.0 mm from the face of the transducer. (A) The 2D Fourier transform of beamformed RF data acquired by the transducer shows echoes centered at fz=5 MHz in the axial direction, as the axial spatial frequency corresponds to the center frequency of the transducer. In the lateral direction (fx), the majority of the energy is located close to 0 with no oscillation in this direction. By applying a filter in the lateral direction only (B), the axial dimension is unmodified, while the resulting RF image (C) contains transverse oscillations having a mean lateral frequency according to the frequency of the applied filter (Jensen, et al. 2017).

Phantom studies

Forward-viewing imaging: initial proof of concept

To test velocity estimation performance, the described transducer was used to acquire 7 steered transmit events at a rate of 3.5 kHz (i.e. 250 Hz for 14 transmit events) in a custom flow phantom under conditions of continuous flow of a blood-mimicking fluid (10 g/L corn starch) away from the transducer at either 35 mm/s or 3 mm/s. In the case of 35 mm/s flow, reverse flow was introduced by manually forcing retrograde flow for 0.5 s. The axial velocity component vz was determined using a cross- correlation estimator after filtering stationary echoes with a cutoff of 41.5 Hz and an ensemble size of 50. Velocity components orthogonal to the beam were estimated by transverse oscillation as described, implemented in both lateral and elevation directions in the frequency domain using a bandpass filter centered at 1.6 cycles/mm with −6 dB bandwidth of 93%.

Effect of catheter on blood flow measurement: preliminary studies

In order to study the effect of the catheter on blood flow dynamics in a femoral artery, custom flow phantoms were developed with and without stenoses. Briefly, tissue-mimicking phantoms (Madsen, et al. 1978) were constructed using an aluminum rod with a diameter of 6.33 mm as a negative to mimic the diameter of healthy human adult peripheral arteries (Benetos, et al. 1993, Sandgren, et al. 1999). After solidification, the negative removed after the phantom hardened. In order to form phantoms with stenoses, an aluminum rod was trimmed to a minimum diameter of 3.6 mm, creating a 43% diameter stenosis (%DS = [1- stenotic throat diameter / regular lumen diameter] × 100). The stenotic throat was positioned 87 mm from the inlet of the vessel to ensure the flow was fully developed before encountering the stenosis. Continuous flow was used in these studies, as previous studies indicate the error due to assuming steady state rather than pulsatile flow is small when estimating average velocity in arteries with small diameters, including femoral and carotid arteries (Eriksen 1992).

Particle imaging velocimetry: data acquisition and post-processing

Velocity measurements were acquired using a linear array transducer with 128 elements (L11–5, ATL, Bothell, WA) in phantoms with straight and stenotic vessels. The linear array was placed directly on top of the phantom and aligned with the longitudinal axis of the lumen. A syringe pump (PHD 2000, Harvard Apparatus, Holliston, MA) was used to flow microbubbles in degassed water through the lumen of the phantom. The flow rate of the pump was set to achieve an inlet velocity of 16 cm/s, similar to mean velocities measured in adult human femoral arteries (Fronek, et al. 1976). In order to investigate the effect of the catheter on flow dynamics for catheter-based intravascular transducer, measurements were acquired: 1) without a catheter, 2) with a solid catheter (4.18 mm outer diameter), and 3) with a hollow catheter (4.18 mm outer diameter). Measurements were repeated three times for each set of conditions. In the phantom with the stenosis, velocity measurements were acquired when the tip of the catheter was positioned 12 mm from the stenotic throat.

Ultrasound imaging data were acquired using a programmable ultrasound system (Verasonics Vantage 256). A transmit frequency of 10 MHz was used with unfocused transmit events at a rate of 5000 frames per second. The beamformed in-phase and quadrature (IQ) data were saved for post-processing in Matlab (The MathWorks, Natick, MA, USA). Clutter filtering was performed using a singular value decomposition (SVD) filter to discard first largest singular value to separate fast moving microbubbles from the stationary phantom background (Demené, et al. 2015). Next, the ultrasound frames were saved with a display dynamic range of 25 dB and velocity vectors were estimated using a particle imaging velocimetry approach (PIV Lab) (Thielicke and Stamhuis 2014). Centerline velocity magnitudes along the long axis of the vessel were plotted to compare cases with and without the catheter.

In vivo imaging study

The use of animals for this procedure was reviewed and approved by the Institutional Animal Care and Use Committee at Emory University, and the procedures were performed in an AAALAC accredited facility following the NIH guidelines for animal research. An 80 kg Yorkshire swine was sedated, intubated, mechanically ventilated and maintained on anesthesia per previously reported protocols (Sielicka, et al. 2018). Under general anesthesia, an arterial cutdown was performed in the right groin to expose the femoral artery, and ~5 cm length was dissected carefully from surrounding tissues with ligation of side branches. The distal and proximal ends of the exposed vessel were temporarily clamped, and the tip of the forward viewing IVUS catheter was positioned against the wall and angled to image flow towards the transducer through the wall of the femoral artery (i.e. not inside the artery). The distal clamp was removed to allow blood flow through the femoral artery towards the catheter, and imaging was performed. 3D imaging volumes were acquired with 7 compounding events at a rate of 700 Hz (100 Hz post-compounding). This rate was chosen to allow sampling of slow flow in an animal with a reduced heart rate. Because the animal was under heavy anesthesia, the heart rate was very low, approximately 40 beats per minute according to a heart monitor. Stationary echoes were filtered with a cutoff of 20 Hz and an ensemble size of 50.

Results

Phantom studies

Forward-viewing imaging: initial proof of concept

In Fig. 3A, the velocity at the center of the lumen in a phantom for both flow rates is presented over the duration of a 1 s acquisition. For the higher flow velocity case, the transient retrograde flow (0.25s) can be seen in this plot. Selected 2D representations (i.e. C-scan slices) of 3D vector velocity data acquired in real-time in flow phantom studies are shown in Fig. 3B–E to show sensitivity to slow flow (B) and to variation in primary flow direction (C-E).

Figure 3.

Figure 3.

Open in a new tab

Velocity in flow phantom experiments. (A) Velocity in the primary flow direction at the center of the lumen as a function of time during a one-second acquisition for constant flow away from the transducer at 3 mm/s (dashed black line) and at −35 mm/s (solid blue line). At 0.7 s, the flow direction is reversed for 0.25 s. (B) A single 2D cross-sectional representation of 3D vector velocity for the constant slow flow case (dashed line in A). Single 2D cross-sectional representations are shown at times before (C), during (D) and after (E) flow reversal.

Effect of catheter on blood flow measurement: preliminary studies

Centerline velocity profiles in a 6.33 mm diameter lumen in a phantom without a stenosis are shown in Figure 4. The average velocity magnitude for the case without a catheter is 28.94 cm/s ± 3.62 cm/s. When a hollow catheter with an inner diameter of 3.81 mm and an outer diameter of 4.18 mm was placed inside the lumen, the velocity magnitude at the tip of the catheter was 22.6 cm/s ± 1.64 cm/s, increasing with distance towards the true value. For example, at a depth of 20.8 mm from the end of the hollow catheter, the velocity magnitude is 31.50 ± 0.56 cm/s, compared to 32.23 ± 2.97 cm/s without a catheter and 26.82 ± 0.35 cm/s with a solid catheter. The velocity profile of the case with the hollow catheter beyond this distance is similar to the control case with an error of 1.40 cm/s. Figure 4A indicates that for a straight vessel in the absence of a stenosis, the presence of a solid catheter introduces some variation in velocity variation, with oscillation as a function of distance (blue line in Fig. 4A), however, the mean velocity magnitude remains within 10 cm/s of the velocity without the catheter (purple line in Fig. 4A) and in most cases much closer (2–5 cm/s). While it should be noted that this is only velocity magnitude in the primary flow direction, this represents initial quantification of the effect of the catheter on primary flow dynamics in a straight vessel without a stenosis.

Figure 4.

Figure 4.

Open in a new tab

Effect of catheter designs on hemodynamics. Centerline velocity profiles in a 6.33 mm diameter lumen in a flow phantom (A) without a stenosis and (B) with a stenosis for a mean inlet flow velocity of 16.0 cm/s. (A) In the absence of a stenosis, results indicate that the presence of solid catheter (blue) produces variations in mean steady-state velocity of approximately ±5 cm/s relative to the case without the catheter (purple) as function of distance from the catheter. In the case of a hollow catheter (green), the velocity becomes similar to the case without the catheter at a distance of approximately 20 mm from the end of the catheter. (B) In the presence of a 43% diameter stenosis, the insertion of a solid catheter results in similar spatial behavior but with decreased velocity at all distances, while the insertion of a hollow catheter causes only a minor decrease in velocity relative to the case without the catheter.

Alternatively, Figure 4B shows centerline velocity profiles in a 6.33 mm diameter lumen in a phantom with a 43% diameter stenosis when a hollow catheter with an inner diameter of 3.81 mm and an outer diameter of 4.18 mm was positioned 12 mm from the stenotic throat. The velocity profiles for the case without the catheter and with the hollow catheter are similar, however, the peak velocity magnitude without the catheter is 3.6 cm/s greater than the case with a catheter (54.97 ± 2.13 cm/s vs. 51.37 ± 1.08 cm/s). The peak velocity magnitude occurs at a distance of 11.82 mm for the control case and 11.11 mm for the case with the hollow catheter. After the stenosis, the velocity magnitude decreases more rapidly with increasing distance for the case with the hollow catheter than the control case. When a solid catheter is inserted, a decrease in velocity can be seen throughout the vessel and the peak velocity magnitude is 34.7 ± 2.05 cm/s, which is 20.27 cm/s less than the control case, and occurs at a distance of 11.47 mm.

In vivo porcine studies

The time course of flow velocity in vivo is shown in Fig. 5A for a location near the center of the lumen. In Fig. 5B and 5C, the time courses of velocity are shown for locations moving increasingly closer to the vessel wall. Colored vertical lines in Fig. 5A–C indicate time points of interest for which 2D slices of 3D vector velocity information are shown in Fig. 5D–F. The border outlining the velocity time course plots in Fig. 5A–C corresponds to a colored dot in Fig. 5D–F to indicate the spatial locations of these velocity measurements. Fig. 5D shows a local maximum in forward flow velocity at the center of the lumen occurring at 1.81 seconds (t1). In Fig. 5E, peak forward flow velocity at the center of the lumen is shown occurring at 2.27 seconds (t2). Finally, in Fig. 5F shows velocity fields at 2.97 seconds (t3), indicating a local maximum in backward velocity at the edge of the lumen with simultaneous near- zero or slightly forward flow in other spatial locations.

Figure 5.

Figure 5.

Open in a new tab

Corresponding temporal and spatial information from 3D data sets acquired at slow flow velocities in the porcine femoral artery. (A) Velocity in the primary flow direction as a function of time near the center of the lumen (at the location indicated by the green dot in D-F). (B) Velocity as a function of time between the center and the edge of the lumen (at the location indicated by the light blue dot in D-F). (C) Velocity as a function of time near the edge of the lumen (at the location indicated by the yellow dot in D-F). (D-F) C-scan representations of velocity vectors at a depth of 12 mm from the transducer during a local maximum in forward flow velocity at the center of the lumen occurring at t1 = 1.81 seconds (D), peak forward flow velocity at the center of the lumen occurring at t2 = 2.27 seconds (E), and during a local maximum in backward velocity at the edge of the lumen, with simultaneous near-zero or slightly forward flow in other spatial locations, at t3 = 2.97 seconds (F). Colored vertical bars in (A-C) indicate the time of the images with corresponding colored borders in (D-F).

After slight distal movement and angulation of the transducer, a different acquisition was captured to illustrate the 3D spatiotemporal imaging capabilities. Three cross-sectional slices of vector velocity data are reconstructed from a single volumetric acquisition inside the artery, shown in Figure 6. The separation between these three slices is 2 mm and indicate variation in velocity over this distance during an instance of complex hemodynamics as the blood flow transitions from forward to reverse flow. Note that in the direction of primary flow both forward and backward flow are present simultaneously at this time point.

Figure 6.

Figure 6.

Open in a new tab

Simultaneously-acquired 2D representations of 3D velocity in three different planes at varying distances from the transducer face. These figures separated by 2 mm provide an illustration of the complex dynamics occurring during the transition from peak forward flow to backward flow, as in the primary flow direction, both forward and backward flow are present simultaneously at closely-spaced locations.

In Fig. 7, only the directions of the two transverse velocity components of the data presented in Fig. 6 are shown in order to allow visualization of these velocity components (i.e. orthogonal to the direction of primary flow along the long axis of the vessel), which is difficult to appreciate in the 3D visualization of Fig. 6.

Figure 7.

Figure 7.

Open in a new tab

Out-of-plane (i.e. orthogonal to the beam) velocity components in three different planes at varying depths. The velocity components along the short axis of the vessel (i.e. in the lateral and elevation direction) are shown for the cases of Fig. 6.

Finally, in Fig. 8, the spatially-averaged velocity is shown in the primary flow direction over the entire lumen at a distance from the transducer of z=12 mm, illustrating the cyclical flow pattern over 4 cardiac cycles, with spatial regions of negative and positive flow averaging. This indicates that while a spatially-averaged measurement can encapsulate the primary flow velocity with respect to the cardiac cycle, the mean velocity contains many complex underlying flow dynamics, as each local minimum or maximum in velocity contains blood flow in both primary directions (Fig. 8B, E, F), while only the absolute peaks (Fig. 8C, D) show unidirectional flow across the vessel in the primary flow direction. These spatially-averaged flow dynamics are consistent with discrete measurements of femoral artery blood flow dynamics made in humans and in animals (Cole, et al. 2003, Hwang 2017), which indicate multiple phases including the acceleration phase (Cole, et al. 2003) (i.e. Fig. 8B), which immediately precedes the maximum flow phase (Fig. 8C). Maximum forward flow is then followed by the deceleration phase and period of reversed volumetric flow (Cole, et al. 2003), as in Fig. 8D, before positive flow resumes for the following cycle (Fig. 8E).

Figure 8.

Figure 8.

Open in a new tab

(A) Spatially-averaged velocity in the primary flow direction over the entire lumen at a depth of z=12 mm shows the cyclical flow pattern over 4 cardiac cycles, with spatial regions of negative and positive flow averaging. Over the final acquired cycle, the instantaneous velocity vector maps are shown for (B) increasing towards maximum forward flow, (C) maximum forward flow, (D) maximum backward flow, (E) local maximum (secondary peak) of forward flow despite the presence of flow in the opposite direction, and (F) beginning of increase towards maximum forward flow for the subsequent cycle.

Discussion

3D blood flow dynamics were measured using a forward-viewing intravascular 2D array for the first time. Initial testing was performed in phantom studies (Fig. 3–4), followed by in vivo imaging through the wall of a porcine femoral artery. Experimental studies in flow phantoms indicate that for a straight vessel, the catheter introduces spatial oscillation with respect to distance on the order of 3–5 cm/s, decreasing beyond ~12 mm from the end of the catheter (Fig. 4A). For a vessel with a stenosis, in the proximal region, the catheter produces significant decrease in velocity magnitude, although use of a hollow catheter allowing blood flow through the center produces minimal disturbance of the flow profile relative to the case without a catheter (Fig. 4B). The estimated mean velocity in the primary blood flow direction (Fig. 8) shows the expected cyclical behavior and agrees well with the period of the heart monitor, i.e. 1/(40 bpm) corresponding to a period of 1.5 seconds. In further analyzing segments of data acquired in vivo, temporal dynamics of blood flow appear similar but are not identical in different spatial locations (Fig. 5 A–C). Alternatively, by viewing multiple 3D slices of blood flow velocity data at the same time point, the spatial dynamics are correlated but the system is sensitive to 3D spatial variations in blood flow over distances of a few mm (Fig. 6). In Fig. 7, the cross-sectional 2D components show that velocities orthogonal to the beam vary slightly in both direction and magnitude across the artery. The ability to estimate 3D velocity could be useful for assessing blood flow dynamics at the vessel wall surrounding lesions, where high wall shear stress proximal to the lesion indicates increased likelihood of plaque rupture (Kumar, et al. 2018). For this task, the component along the vessel wall would need to be estimated rather than the cross-sectional components shown; these data are contained within the acquired sequence of volumes and can be interpolated to account for the angle of the vessel wall as needed.

Effect of catheter on blood flow dynamics

As demonstrated in laboratory testing in flow phantoms with a straight vessel and with a severe stenosis, inserting a catheter-based device into a vessel induces a change in the blood flow dynamics immediately downstream from the catheter, however, the change in variation in velocity magnitude is relatively small (though spatially-varying) for a straight vessel. While the transducer was positioned on the surface of the femoral artery to image through the wall without disrupting flow in the in vivo study in this work, in the future, smaller catheter-based transducers could allow direct intravascular imaging of blood flow dynamics. Although only limited preliminary studies have been performed to examine the effect of the catheter on blood flow dynamics, these early results suggest that a hollow catheter-based transducer allowing internal blood flow within the catheter may introduce minimal disruption to the flow dynamics, as the flow profile for a hollow catheter is very similar to the case without a catheter in the presence of a stenosis (Fig. 4B). Additional studies are required to determine the effect of the catheter on other flow components, at the vessel wall, and as a function of catheter design parameters such as diameter, strut thickness, etc. In order to enable imaging in coronary arteries (<5 mm diameter (Dodge, et al. 1992)), future designs will be smaller (1–2 mm), operate at higher frequency (>15 MHz), and utilize multiplexing via on-ASICs to reduce the number of electrical connections (Khuri-Yakub, et al. 2010).

Future improvements

Several additional steps are required to translate the benefits of a forward-viewing 2D array-based system to clinical diagnosis or risk stratification. Cross-validation between experimental studies and computational fluid dynamics—particularly in cases with stenoses—are important for confirming the accuracy of forward-viewing this technique relative to the more extensive studies that have been previously performed with CFD and could also aid in the catheter design. Additional animal studies, particularly in animals with stenoses and varying inlet velocities with corresponding data from external commercial ultrasound measurements (i.e. 1D pulsed spectral Doppler), would also increase the confidence in blood flow estimation accuracy using the 2D array-based system.

In addition, several technical improvements could increase the accuracy of 3D blood flow velocity estimates made by this system, including directional beamforming (Jensen, et al. 2017), which reduces bias and standard deviation. Additionally, in this work, processing was performed offline, however, implementation of online or rapidly reconstructed velocity maps would be needed to stratify risk in the cardiac catheterization lab. Improvements in filtering such as automatic selection of the clutter filter threshold or singular value decomposition (SVD) filtering may also improve the quality of the velocity estimates. Most importantly, in order to translate to carotid imaging, a smaller, higher resolution forward-viewing matrix array transducer needs to be developed.

In order to estimate wall shear stress, it is necessary to estimate slow velocities at the vessel wall with high accuracy, which means a high velocity resolution δν=νmaxN/2=cfPRF2f0cosθ1N is required (Cobbol2007). A high velocity resolution (i.e. small) can be achieved by a long ensemble (N) or a low pulse repetition frequency. However, a low value of as was used in this work may result in aliasing for fast flow velocities at the center of the vessel. Because the primary velocities of interest are not the maximum possible velocities in the center of a healthy vessel but the slower velocities adjacent to the wall to estimate WSS, a velocity resolution of just δν=νmaxN/2=cfPRF2f0cosθ1N=1.54 mm/s can be achieved for an ensemble length of N=500 at a fPRF= 5 kHz, which results in results in νmax=cfPRF4f0cosθ=385 mm/s (i.e. appropriate for the peak velocities encountered in coronary arteries (Fiorentini, et al. 2018)). In this case, sensitivity to slow flow is not lost, as sensitivity to slow flow is determined not by ensemble length or PRF but by ability to separate slow moving blood and tissue echoes, but velocity resolution may be insufficient for slow flow velocities near the wall. Alternatively, velocity resolution can be refined by decreasing fPRF. For example, for fPRF = 100 Hz, δν=0.031 mm/s, however, this results in = 7.7 mm/s and potential aliasing at the center of the artery. In the future, fPRF, ensemble length N, and number of compounding angles will need to be determined based on the application and may need to be adjusted for imaging the center of the vessel vs. fine velocity estimation near the wall.

Summary

This article presents the first demonstration of a 2D array for intravascular imaging of 3D blood flow velocity fields. Proof-of-concept for spatiotemporal analysis was demonstrated using a 118-element forward-viewing array operating at 5 MHz. Development of an intravascular 3D system for real-time 3D quantification of intravascular blood flow velocity fields near the wall of the artery could allow estimation of wall shear stress in order to assess likelihood of plaque rupture in patients with stable coronary artery disease. While additional development of a catheter-based imaging system with minimum flow disruption is required, direct 3D estimation of blood flow velocity fields in the cardiac catheterization lab could enable enhanced risk stratification for patients undergoing catheterization without the need for computation based on vessel geometry.

Acknowledgments

The authors thank Daisuke Onohara, M.D., Ph.D., for assistance during the animal study. This work was supported in part by the Department of Biomedical Engineering and the College of Engineering at Georgia Institute of Technology and by R01HL144714 from the National Institutes of Health.

Footnotes

Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.

References

  1. Benetos A, Laurent S, Hoeks AP, Boutouyrie PH, Safar ME. Arterial alterations with aging and high blood pressure. A noninvasive study of carotid and femoral arteries. Arterioscler Thromb 1993;13:90–7. [DOI] [PubMed] [Google Scholar]
  2. Benjamin EJ, Muntner P, Alonso A, Bittencourt MS, Callaway CW, Carson AP, Chamberlain AM, Chang AR, Cheng S, Das SR, Delling FN, Djousse L, Elkind MSV, Ferguson JF, Fornage M, Jordan LC, Khan SS, Kissela BM, Knutson KL, Kwan TW, Lackland DT, Lewis TT, Lichtman JH, Longenecker CT, Loop MS, Lutsey PL, Martin SS, Matsushita K, Moran AE, Mussolino ME, O’Flaherty M, Pandey A, Perak AM, Rosamond WD, Roth GA, Sampson UKA, Satou GM, Schroeder EB, Shah SH, Spartano NL, Stokes A, Tirschwell DL, Tsao CW, Turakhia MP, VanWagner LB, Wilkins JT, Wong SS, Virani SS, American Heart Association Council on Epidemiology, Prevention Statistics Committee, Stroke Statistics Subcommittee. Heart Disease and Stroke Statistics-2019 Update: A Report From the American Heart Association. Circulation 2019;139:e56–e528. [DOI] [PubMed] [Google Scholar]
  3. Cannata JM, Chilipka T, Yang HC, Han S, Ham SW, Rowe VL, Weaver FA, Shung KK, Vilkomerson D. Development of a flexible implantable sensor for postoperative monitoring of blood flow. Journal of ultrasound in medicine : official journal of the American Institute of Ultrasound in Medicine 2012;31:1795–802. [DOI] [PMC free article] [PubMed] [Google Scholar]
  4. Cobbold RSC. Foundations of Biomedical Ultrasound. New York: Oxford University Press, 2007. [Google Scholar]
  5. Cole JS, Watterson JK, O’Reilly MJG. Blood flow characteristics in a femoral artery bypass graft. Dev Chem Eng Minearl Process 2003;11:15–28. [Google Scholar]
  6. Correia M, Provost J, Tanter M, Pernot M. 4D ultrafast ultrasound flow imaging: in vivo quantification of arterial volumetric flow rate in a single heartbeat. Physics in medicine and biology 2016;61. [DOI] [PubMed] [Google Scholar]
  7. Demené C, Deffieux T, Pernot M, Osmanski B-F, Biran V, Gennisson J-L, Sieu L-A, Bergel A, Franqui S, Correas J-M. Spatiotemporal clutter filtering of ultrafast ultrasound data highly increases Doppler and fUltrasound sensitivity. IEEE transactions on medical imaging 2015;34:2271–85. [DOI] [PubMed] [Google Scholar]
  8. Dodge JT Jr., Brown BG, Bolson EL, Dodge HT. Lumen diameter of normal human coronary arteries. Influence of age, sex, anatomic variation, and left ventricular hypertrophy or dilation. Circulation 1992;86:232–46. [DOI] [PubMed] [Google Scholar]
  9. Eriksen M. Effect of pulsatile arterial diameter variations on blood flow estimated by Doppler ultrasound. Med Biol Eng Comput 1992;30:46–50. [DOI] [PubMed] [Google Scholar]
  10. Fiorentini S, Saxhaug LM, Bjastad TG, Holte E, Torp H, Avdal J. Maximum Velocity Estimation in Coronary Arteries Using 3-D Tracking Doppler. Ieee T Ultrason Ferr 2018;65:1102–10. [DOI] [PubMed] [Google Scholar]
  11. Fronek A, Coel M, Berstein EF. Quantitative ultrasonographic studies of lower extremity flow velocities in health and disease. Circulation 1976;53:957–60. [DOI] [PubMed] [Google Scholar]
  12. Hansen KL, Nielsen MB, Jensen JA. Vector velocity estimation of blood flow - A new application in medical ultrasound. Ultrasound 2017;25:189–99. [DOI] [PMC free article] [PubMed] [Google Scholar]
  13. Hong JH, Su M, Yu YY, Zhang ZQ, Liu R, Huang YC, Mu PT, Zheng HR, Qiu WB. A Dual-Mode Imaging Catheter for Intravascular Ultrasound Application. IEEE transactions on medical imaging 2019;38:657–63. [DOI] [PubMed] [Google Scholar]
  14. Hwang JY. Doppler ultrasonography of the lower extremity arteries: anatomy and scanning guidelines. Ultrasonography 2017;36:111–19. [DOI] [PMC free article] [PubMed] [Google Scholar]
  15. Jensen J, Hoyos CAV, Stuart MB, Ewertsen C, Nielsen MB, Jensen JA. Fast Plane Wave 2-D Vector Flow Imaging Using Transverse Oscillation and Directional Beamforming. Ieee T Ultrason Ferr 2017;64:1050–62. [DOI] [PubMed] [Google Scholar]
  16. Jensen J, Stuart MB, Jensen JA. High Frame Rate Vector Velocity Estimation using Plane Waves and Transverse Oscillation. 2015 Ieee International Ultrasonics Symposium (Ius) 2015. [Google Scholar]
  17. Jensen JA. A new estimator for vector velocity estimation. IEEE transactions on ultrasonics, ferroelectrics, and frequency control 2001;48:886–94. [DOI] [PubMed] [Google Scholar]
  18. Khuri-Yakub BT, Oralkan O, Nikoozadeh A, Wygant IO, Zhuang S, Gencel M, Choe JW, Stephens DN, de la Rama A, Chen P, Lin F, Dentinger A, Wildes D, Thomenius K, Shivkumar K, Mahajan A, Seo CH, O’Donnell M, Truong U, Sahn DJ. Miniaturized Ultrasound Imaging Probes Enabled by CMUT Arrays with Integrated Frontend Electronic Circuits. 2010 Annual International Conference of the Ieee Engineering in Medicine and Biology Society (Embc) 2010:5987–90. [DOI] [PMC free article] [PubMed] [Google Scholar]
  19. Kumar A, Thompson EW, Lefieux A, Molony DS, Davis EL, Chand N, Fournier S, Lee HS, Suh J, Sato K, Ko YA, Molloy D, Chandran K, Hosseini H, Gupta S, Milkas A, Gogas B, Chang HJ, Min JK, Fearon WF, Veneziani A, Giddens DP, King SB 3rd, De Bruyne B, Samady H. High Coronary Shear Stress in Patients With Coronary Artery Disease Predicts Myocardial Infarction. Journal of the American College of Cardiology 2018;72:1926–35. [DOI] [PubMed] [Google Scholar]
  20. Lee JM, Shin ES, Nam CW, Doh JH, Hwang D, Park J, Kim KJ, Zhang J, Ahn C, Koo BK. Clinical Outcomes According to Fractional Flow Reserve or Instantaneous Wave-Free Ratio in Deferred Lesions. JACC. Cardiovascular interventions 2017;10:2502–10. [DOI] [PubMed] [Google Scholar]
  21. Lenge M, Ramalli A, Boni E, Liebgott H, Cachard C, Tortoli P. High-Frame-Rate 2-D Vector Blood Flow Imaging in the Frequency Domain. Ieee T Ultrason Ferr 2014;61:1504–14. [DOI] [PubMed] [Google Scholar]
  22. Lindsey BD, Light ED, Nicoletto HA, Bennett ER, Laskowitz DT, Smith SW. The ultrasound brain helmet: new transducers and volume registration for in vivo simultaneous multi-transducer 3-D transcranial imaging. IEEE transactions on ultrasonics, ferroelectrics, and frequency control 2011;58:1189–202. [DOI] [PMC free article] [PubMed] [Google Scholar]
  23. Madsen EL, Zagzebski JA, Banjavie RA, Jutila RE. Tissue mimicking materials for ultrasound phantoms. Medical physics 1978;5:391–4. [DOI] [PubMed] [Google Scholar]
  24. Montaldo G, Tanter M, Bercoff J, Benech N, Fink M. Coherent Plane-Wave Compounding for Very High Frame Rate Ultrasonography and Transient Elastography. Ieee T Ultrason Ferr 2009;56:489–506. [DOI] [PubMed] [Google Scholar]
  25. Pihl MJ, Jensen JA. A transverse oscillation approach for estimation of three-dimensional velocity vectors, part I: concept and simulation study. IEEE transactions on ultrasonics, ferroelectrics, and frequency control 2014;61:1599–607. [DOI] [PubMed] [Google Scholar]
  26. Pihl MJ, Stuart MB, Tomov BG, Rasmussen MF, Jensen JA. A transverse oscillation approach for estimation of three-dimensional velocity vectors, part II: experimental validation. IEEE transactions on ultrasonics, ferroelectrics, and frequency control 2014;61:1608–18. [DOI] [PubMed] [Google Scholar]
  27. Salles S, Chee AJY, Garcia D, Yu ACH, Vray D, Liebgott H. 2-D Arterial Wall Motion Imaging Using Ultrafast Ultrasound and Transverse Oscillations. Ieee T Ultrason Ferr 2015;62:1047–58. [DOI] [PubMed] [Google Scholar]
  28. Sandgren T, Sonesson B, Ahlgren R, Lanne T. The diameter of the common femoral artery in healthy human: influence of sex, age, and body size. Journal of vascular surgery 1999;29:503–10. [DOI] [PubMed] [Google Scholar]
  29. Sielicka A, Sarin EL, Shi W, Sulejmani F, Corporan D, Kalra K, Thourani VH, Sun W, Guyton RA, Padala M. Pathological Remodeling of Mitral Valve Leaflets from Unphysiologic Leaflet Mechanics after Undersized Mitral Annuloplasty to Repair Ischemic Mitral Regurgitation. J Am Heart Assoc 2018;7:e009777. [DOI] [PMC free article] [PubMed] [Google Scholar]
  30. Smith SW, Gardea P, Patel V, Douglas SJ, Wolf PD. Double Ring Array Catheter for In Vivo Real-Time 3D Ultrasound. Ultrasonic imaging 2014;36:167–76. [DOI] [PubMed] [Google Scholar]
  31. Steg PG, Bhatt DL, Wilson PW, D’Agostino R Sr., Ohman EM, Rother J, Liau CS, Hirsch AT, Mas JL, Ikeda Y, Pencina MJ, Goto S, Investigators RR. One-year cardiovascular event rates in outpatients with atherothrombosis. JAMA : the journal of the American Medical Association 2007;297:1197–206. [DOI] [PubMed] [Google Scholar]
  32. Thielicke W, Stamhuis EJ. PIVlab – Towards User-Friendly, Affordable and Accurate Digital Particle Image Velocimetry in MATLAB. Journal of Open Research Software 2014;2. [Google Scholar]
  33. Tierney J, Coolbaugh C, Towse T, Byram B. Adaptive Clutter Demodulation for Non-Contrast Ultrasound Perfusion Imaging. IEEE transactions on medical imaging 2017;36:1979–91. [DOI] [PMC free article] [PubMed] [Google Scholar]
  34. Udesen J, Jensen JA. Investigation of transverse oscillation method. IEEE transactions on ultrasonics, ferroelectrics, and frequency control 2006;53:959–71. [DOI] [PubMed] [Google Scholar]
  35. Wigen MS, Fadnes S, Rodriguez-Molares A, Bjastad T, Eriksen M, Stensaeth KH, Stoylen A, Lovstakken L. 4-D Intracardiac Ultrasound Vector Flow Imaging-Feasibility and Comparison to Phase- Contrast MRI. IEEE transactions on medical imaging 2018;37:2619–29. [DOI] [PubMed] [Google Scholar]

RESOURCES