🔬 applied physics

Revisiting XDoppler estimator for high spatiotemporal resolution volumetric axial velocity measurement using row-column arrays

This paper presents a novel XDoppler velocity estimator for row-column addressed ultrasound arrays that leverages cross-correlation of orthogonal apertures to achieve high spatiotemporal resolution volumetric axial velocity measurements with improved accuracy, reduced aliasing, and enhanced sensitivity to slow flows compared to traditional methods.

Original authors: Henri Leroy, Adrien Bertolo, Guillaume Goudot, Mickael Tanter, Thomas Deffieux, Mathieu Pernot

Published 2026-05-01

📖 5 min read🧠 Deep dive

CC BY 4.0

Original authors: Henri Leroy, Adrien Bertolo, Guillaume Goudot, Mickael Tanter, Thomas Deffieux, Mathieu Pernot

Original paper licensed under CC BY 4.0 (http://creativecommons.org/licenses/by/4.0/). ✨ This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer

The Big Picture: Seeing Blood Flow in 3D

Imagine trying to take a video of a busy highway. If you only have a camera that can take a flat, 2D photo, you can see cars passing by, but you can't easily see how traffic flows in three dimensions (up, down, left, right, and forward).

In medical ultrasound, doctors want to see blood flowing through veins and arteries in 3D to diagnose problems. However, building a camera (ultrasound probe) that can do this 3D video is very hard. It usually requires thousands of tiny wires and massive computers, making the machines too big and expensive for regular clinics.

The Solution: The researchers used a special type of probe called a Row-Column Addressed (RCA) array. Think of this like a grid of streetlights. Instead of needing a separate wire for every single light (which is expensive), you only need wires for the "rows" and wires for the "columns." You turn on a whole row, then a whole column. This makes the hardware much cheaper and smaller, but it creates a "blurry" picture with some unwanted noise (like static on an old TV).

The Problem: The "Blurry" Picture

Because this cheaper probe creates a blurrier image, the standard way of measuring blood speed (called the Kasai estimator) gets confused. It's like trying to count how fast cars are driving on a highway while looking through a foggy window. The standard method often:

Underestimates the speed of fast cars (aliasing).
Overestimates the speed of slow cars near the edges.
Gets the "average" speed wrong because the blur mixes fast and slow areas together.

The New Trick: The "XDoppler" Estimator

The researchers took an existing idea called XDoppler (which was originally just for seeing where blood is, not how fast it's going) and upgraded it to measure speed accurately.

Here is how their new method works, using an analogy:

The "Two-Person Relay" Analogy
Imagine two people, Row and Column, trying to measure how fast a runner is moving down a track.

Row takes a photo of the runner.
Column takes a photo of the runner a split-second later.
Because they are looking from slightly different angles (orthogonal), their photos have different "noise" or blurs.

The Old Way (Kasai):
The old method would just take one person's photos, compare them to the next photo, and guess the speed. If the photo was blurry, the guess would be wrong.

The New Way (XDoppler):
The new method is smarter. It takes Row's photo and compares it to Column's photo, and then does the reverse (Column's vs. Row's).

By comparing these two different "views" of the same moment, the method cancels out the "noise" and the blur.
It effectively creates a much sharper, clearer picture of the blood flow.

What Did They Find? (The Results)

1. Sharper Vision (Better Resolution)
Just like switching from a blurry lens to a high-definition camera, the new method sees the blood flow much more clearly. In their tests with a fake blood vessel (a flow phantom), the old method guessed the speed was too slow in the middle and too fast at the edges. The new method got the speed right across the whole pipe, matching the perfect "parabolic" shape of flowing water.

2. Seeing Faster Speeds (No Aliasing)
In video games, if a car moves too fast, it sometimes looks like it's driving backward (this is called "aliasing").

The old method had a "speed limit" (Nyquist limit). If the blood moved faster than this limit, the measurement would flip and become wrong.
The new method doubled this speed limit. It can handle blood moving twice as fast before it gets confused.
Real-life test: When they tested this on a human volunteer's neck (carotid artery), the old method got confused and showed "aliased" (wrong) speeds in the veins. The new method kept the speed measurement correct, even when the blood was moving quickly.

3. Catching the Pulse
The new method was able to track the heartbeat. It could see the blood speed go up and down as the heart pumped, showing the natural rhythm of the artery without getting lost in the noise.

The Trade-off

There is one small catch. To get this super-clear, high-speed measurement, the computer needs to "think" for a tiny bit longer (about 0.05 to 0.2 seconds) to gather enough data.

Analogy: It's like taking a long-exposure photo to get a sharp image in low light. You have to hold the camera still for a split second.
The paper notes that while this is a slight delay, it is fast enough to see the heartbeat and flow changes in real-time for most clinical needs.

Summary

The researchers took a cheaper, 3D ultrasound probe that usually produces blurry images and invented a new math trick (the XDoppler velocity estimator) to fix the blur. This trick allows the probe to measure blood speed much more accurately, see faster blood flow without getting confused, and provide a clearer 3D map of how blood moves through the body. This brings high-quality 3D blood flow imaging closer to being a standard tool in hospitals.

1. Problem Statement

Accurate 3D volumetric velocity estimation is critical for diagnostic and therapeutic ultrasound applications (e.g., tissue elasticity, functional imaging, and blood flow dynamics). However, traditional fully populated matrix arrays face significant hardware and computational barriers (cost, size, power) that hinder their clinical adoption.

Row-Column Addressed (RCA) arrays offer a solution by reducing hardware complexity. However, they introduce specific imaging challenges:

Poor Contrast and Resolution: RCA arrays suffer from strong sidelobes and elongated Point Spread Functions (PSF) due to the separate transmission of rows and columns.
Velocity Estimation Limitations: While methods like Orthogonal Plane Wave (OPW) compounding improve image quality, they often result in inaccurate motion estimation due to spatial averaging caused by sidelobes.
Aliasing: Conventional phase-shift autocorrelators (Kasai estimator) applied to compounded RCA data have a limited Nyquist velocity, leading to aliasing in high-velocity flows (e.g., carotid arteries).
Gap in Literature: Previous improvements to RCA imaging (e.g., XDoppler, RC-FMAS) focused primarily on Power Doppler (contrast) and resolution, leaving velocity estimation largely unaddressed.

2. Methodology

The authors propose a novel XDoppler velocity estimator that extends the cross-correlation concept of the XDoppler power Doppler method to axial velocity estimation.

A. Acquisition Strategy

Hardware: A 128+128 element RCA probe (6 MHz center frequency) connected to a Verasonics Vantage 256 scanner.
Sequence: The system acquires two sub-volumes sequentially:
1. RC Sub-volume: Plane waves transmitted by Rows, received by Columns.
2. CR Sub-volume: Plane waves transmitted by Columns, received by Rows.
Temporal Lag: There is a specific time delay ( $\Delta T$ ) between the acquisition of the RC and CR sub-volumes.

B. Mathematical Derivation

The core innovation lies in deriving the velocity from the phase of the cross-correlation between the orthogonal sub-volumes ( $s_{RC}$ and $s_{CR}$ ).

The Bias Problem: A simple cross-correlation $s_{RC}(i)s_{CR}^*(i+\Delta T)$ yields a biased phase estimate because the static phase terms ( $\phi_{RC}$ and $\phi_{CR}$ ) specific to each aperture do not cancel out.
The Solution: The authors propose a combined estimator that sums two complementary cross-correlations to isolate the Doppler phase shift ( $\Delta \phi_{Doppler}$ ):
$v_z = -\frac{c}{4\pi f_0 T} \left[ \arg \left( \sum s_{RC}(i)s_{CR}^*(i+0.5) \right) + \arg \left( \sum s_{CR}(i+0.5)s_{RC}^*(i+1) \right) \right]$
By summing the arguments of these two terms, the static phase biases cancel out, leaving only the Doppler shift.

C. Theoretical Advantages

Nyquist Velocity: By splitting the total observation time $T$ into two shorter intervals ( $\Delta T_1 = \Delta T_2 = T/2$ ), the effective displacement time is halved. This doubles the unambiguous velocity range (Nyquist limit) compared to the conventional Kasai estimator applied to OPW compounding:
$v_{N,X} = 2 \times v_{N,comp}$
Spatial Resolution: The cross-correlation operation inherently suppresses sidelobes, leading to a sharper PSF and better contrast compared to coherent OPW summation.

D. Experimental Validation

In Vitro: Flow phantoms (4mm and 8mm diameters) with blood-mimicking fluid were used to test accuracy, convergence time, and aliasing robustness against a Poiseuille flow model.
In Vivo: Acquisitions were performed on the carotid artery and jugular vein of a healthy volunteer to assess performance in pulsatile flow and aliasing scenarios.
Metrics: Root Mean Squared Error (RMSE), flow rate bias, standard deviation, and convergence time ( $t_{90\%}$ ).

3. Key Contributions

Novel Velocity Estimator: The first derivation of a phase-based axial velocity estimator specifically for RCA arrays using the XDoppler cross-correlation framework.
Bias Correction: A mathematical proof and implementation of a method to cancel static phase biases inherent to orthogonal aperture cross-correlations.
Doubled Nyquist Limit: Demonstration that the method theoretically and experimentally achieves a Nyquist velocity twice as high as the standard Kasai estimator on OPW data.
Enhanced Contrast & Resolution: The method leverages the sidelobe suppression of XDoppler to provide more accurate velocity profiles, particularly near vessel walls where conventional methods fail due to spatial averaging.

4. Results

In Vitro Performance

Accuracy: The XDoppler estimator closely matched the theoretical Poiseuille parabolic velocity profile. In contrast, the conventional Kasai/OPW estimator underestimated central velocities by ~10% and overestimated velocities near the walls.
Flow Rate Estimation: XDoppler showed a mean bias of -0.17 mL/min and a standard deviation of 3.43 mL/min, significantly outperforming the Kasai estimator (Bias: -9.33 mL/min; Std Dev: 6.19 mL/min).
Convergence: The estimator stabilized within 0.05s to 0.2s, reaching 90% of the final value quickly.
Aliasing Resistance: In tests where flow velocities exceeded the Nyquist limit of the Kasai estimator (7.7 cm/s) but remained below the XDoppler limit (15.4 cm/s), the Kasai estimator showed severe aliasing, while XDoppler provided correct measurements.

In Vivo Performance (Carotid Artery)

Dynamic Flow: The method successfully tracked pulsatile flow dynamics over multiple cardiac cycles (approx. 44 bpm).
Aliasing Reduction: In the jugular vein, the conventional estimator showed aliasing during diastole due to vessel geometry, whereas XDoppler did not.
Velocity Magnitude: During systole, the XDoppler estimator measured higher peak velocities (1.54 cm/s) compared to OPW (1.32 cm/s), attributed to the reduced aliasing and better preservation of high-velocity signals.

5. Significance and Conclusion

This study successfully bridges the gap between high-quality 3D imaging and accurate velocity quantification for RCA arrays.

Clinical Impact: The method enables more reliable 3D color Doppler imaging in complex flow environments (e.g., stenosis, valvular regurgitation) where high velocities and aliasing are common.
Technical Advancement: It proves that the cross-correlation of orthogonal apertures is not just a tool for Power Doppler contrast but a robust mechanism for velocity estimation, offering a 2x improvement in Nyquist velocity without sacrificing spatial resolution.
Future Directions: The authors suggest integrating this estimator with Vector Flow Imaging (VFI) to resolve full 3D velocity vectors and applying de-aliasing techniques for even higher velocity applications (e.g., cardiac imaging).

In summary, the XDoppler velocity estimator represents a significant step forward in making 4D (volumetric + time) ultrasound imaging clinically viable for quantitative hemodynamic analysis.