Angle-domain null subtraction imaging from beamformed plane-wave data
Listen
Radio episode about this paper
Transcript
Introduction to the show: ident: Physics. The day in general physics: instruments, methods and results across the field.
Kai: We're looking at the paper: "Angle-domain null subtraction imaging from beamformed plane-wave data".
Mira: Here's the gist of the paper. Angle-NSI provides simple post-processing of per-angle beamformed images to narrow the apparent lateral response at low computational cost It narrows the apparent lateral response at low computational cost, at the expense of Bmode detectability The method was evaluated using a simulated point target, experimental resolution phantom targets, an open microbubble (MB) traces dataset and two in vivo carotid acquisitions Angle-NSI reduced the simulated −6 dB lateral width from 0.226 mm for DAS to 0.116 mm, compared with 0.009 mm for Receive-NSI, while the axial width was 0.720 mm for these three methods Angle-NSI had shorter reconstruction time than Receive-NSI and the lowest post-delay processing time among the nonlinear methods Angle-NSI operates on per-angle images during coherent plane-wave compounding The approach uses a matrix formulation to distinguish receive-domain and angle-domain null subtraction expressions and to identify their independent computational components For the angular null, the method is analogous to Receive-NSI that creates an antisymmetric transmit-angle null while Receive-NSI creates the null on receive The angular null field is expressed as Eθ(r) = N (U(r), Zθ(r); c) This construction allows for a simple post-processing technique for beamforming pipelines retaining per-angle images without RF data access Comparison with Other Beamforming Techniques Angle-NSI was compared to several nonlinear beamforming methods, including delayandsum (DAS), coherencefactorweighted DAS (CF-DAS), minimumvariance beamforming (MV), delaymultiplyandsum (DMAS), Receive-NSI and Angle-NSI The computational interpretation shows that Angle-NSI accumulates U and Zθ from the same per-angle beamformed images before angular compounding, leading to a low incremental cost compared to DAS Table 1 summarizes the resulting distinction between Receive-NSI and Angle-NSI regarding required data and additional operations Performance Metrics Across Imaging Modalities The performance of Angle-NSI is evaluated across various scenarios, including simulated point targets, experimental resolution phantoms, open microbubble (MB) traces dataset, and in vivo carotid acquisitions In the simulated point-target response, Angle-NSI reduced the lateral −6 dB width from 0.226 mm for DAS to 0.116 mm In experimental resolution phantom studies, Angle-NSI yielded a lateral width of 0.070 mm compared to DAS's 0.537 mm In power Doppler, microbubble trace widths were reduced by 30.9% for Angle-NSI compared to DAS However, both NSI variants reduced generalized contrast-to-noise ratio relative to the conventional methods on a phantom and in vivo Robustness and Trade-offs Robustness tests indicate that cancellation across angles is an intrinsic limitation, as width varied with angle sampling and increased under phase jitter The robustness to noise showed that Angle-NSI width was 0.122 ± 0.005 mm at 10 dB SNR, but increased to 0.194 ± 0.029 mm under a 10◦ phase-jitter standard deviation The study emphasizes the resolution–detectability trade-off, as both NSI methods decreased gCNR compared to DAS or other nonlinear beamforming methods The power Doppler study suggests that high-resolution Doppler imaging is the most promising use case for Angle-NSI due to its lower computational cost and pipeline complexity In this Note, Angle-NSI approximately halved the apparent lateral width in simulation A comprehensive comparison with other beamforming techniques using simulation, a resolution phantom, in vitro Doppler data and a carotid acquisition shows that Angle-NSI increases the lateral resolution at the expense of B-Mode detectability at a low computational cost Angle-NSI offers simple, low-cost post-processing for high-resolution power Doppler The use of Angle-NSI for contrast-enhanced and contrast-free Doppler ultrasound should be confirmed by further in vivo studies The central result is methodological: the NSI framework can use a null formed across angles after receive beamforming, i.e.
Kai: it can be adapted from receive apodization to angular apodization The timing study shows the low incremental cost of Angle-NSI under the tested conditions, using per-angle images after ordinary receive beamforming, making it faster than Receive-NSI or other nonlinear beamforming methods The experimental phantom and the carotid results emphasize the resolution–detectability trade-off, as both NSI methods decreased gCNR compared to DAS or other nonlinear beamforming methods The robustness tests identify cancellation across angles as an intrinsic limitation Eventually, the power Doppler study shows that high-resolution Doppler imaging is the most promising use case for Angle-NSI, as it effectively narrows the lateral response of the Doppler signal at a lower computational cost and pipeline complexity compared to Receive-NSI The use of Angle-NSI for contrast-enhanced and contrast-free Doppler ultrasound, like Receive-NSI (Kou et al., 2024), should be confirmed by further in vivo studies The use of Angle-NSI for contrast-enhanced and contrast-free Doppler ultrasound, like Receive-NSI (Kou and Oelze, 2025), could also be explored The paper's data are publicly available from their original repositories The analysis code and result records are available at https://github.
Title and authors: Kai: So, what's the big idea behind "Angle-domain null subtraction imaging from beamformed plane-wave data"? It’s about using matrix math to find that angular null subtraction without needing any raw radio frequency data access.
Mira: Right, so it’s not doing the whole receive domain null subtraction thing like some other methods do; it’s just focusing on separating the computational parts of those two subtractions so you can do it later with just the angle domain images <ref:2608.18252#pg2>.
Lev: That separation is key because if you’re trying to build a real hardware pipeline, you need to know exactly what data access you're requiring; this method seems designed to keep the RF processing fixed while making the post-processing simpler <ref:2608.18252#pg1>.
Mira: That’s a substantial improvement over the delay and sum method, which got you to zero point zero zero nine millimeter on the same simulation <ref:2608.18252#pg1>; but they have to be careful here because this resolution gain comes at the cost of B mode detectability; it means things might look less detailed in that specific mode.
Lev: I’ve seen that trade-off a lot when we push for higher sampling rates in error correction, but seeing it quantified this clearly makes the decision easier; it tells you exactly what your hardware will deliver versus what you need to run your algorithm on <ref:2608.18252#pg2>.
The paper's summary: Kai: So, looking at the summary of "Angle-domain null subtraction imaging from beamformed plane-wave data," they are focusing on how this technique can be used for real-world applications beyond just simulation results.
Mira: They are focusing on making the post processing simpler for high resolution power Doppler imaging by using those per angle images to narrow the lateral response while keeping the computational cost very low <ref:2608.18252#pg2>.
Lev: From an error correction view, this suggests a way to build ultrasound reconstruction pipelines that keep the RF processing fixed—meaning you don't have to completely overhaul your existing hardware or software stack—while leveraging Angle-NSI’s advantage in computational efficiency and accessibility <ref:2608.18252#pg1>.
Kai: So, they are aiming for a framework where you can retain per angle images while keeping the core radio frequency processing exactly as it is, which is huge for research software or even scanner compatibility <ref:2608.18252#pg2>.
Mira: And then there's this idea of creating an adaptive contrast enhancement framework by using Angle-NSI as a post filtering approach to manage that resolution versus image quality trade-off dynamically <ref:2608.18252#pg2>.
Lev: That sounds like a sophisticated control loop, which is something we’ve been working on for hardware, but applying it here suggests Angle-NSI isn't just a static filter; it could be part of a more intelligent processing chain <ref:2608.18252#pg1>.
Kai: They also have this idea about using Angle-NSI for contrast-enhanced and contrast-free Doppler ultrasound as potential applications, but they stress that we need further in vivo studies to confirm that the performance holds up when we move out of the lab and into patients <ref:2608.18252#pg1>.
Mira: Exactly, it’s a warning—don't take those results too far until you see them in a clinical setting; the paper is laying out the blueprint for how this technique *could* work in complex Doppler scenarios, but it’s not there yet for guaranteed clinical use <ref:2608.18252#pg2>.
Lev: So they are proposing a path from simulation to practical application through adaptive control and better pipeline integration, but they’re being very cautious about claiming it's ready for real-world deployment right now; it’s a roadmap rather than a finished product <ref:2608.18252#pg1>.
The paper's improvements: Kai: So we’re wrapping up this talk on "Angle-domain null subtraction imaging from beamformed plane-wave data" by summarizing how this method works and what it means for future work in the field.
Mira: Basically, they showed that you can use a matrix formulation to do angular null subtraction using only per angle images, which is much simpler than traditional receive domain methods because you don't need raw radio frequency data access <ref:2608.18252#pg2>.
Lev: And the main implication is that this method has very low incremental cost for post-processing because it’s leveraging information already there in the per angle images before angular compounding; that’s a big win for running fast on real hardware <ref:2608.18252#pg1>.
Kai: They proved it works across simulations, phantoms, and even in vivo carotid acquisitions, showing that the lateral width reduction is real—we saw that zero point two two six millimeter dropping to zero point one one six millimeters on a simulated target <ref:2608.18252#pg2>.
Mira: But we have to remember those numbers are conditional; they showed that under ten decibel signal to noise ratio, the width was around zero point one two two millimeters, but if you introduce a ten degree phase jitter standard deviation, it jumped up to zero point one nine four millimeters <ref:2608.18252#pg3>.
Lev: That sensitivity under phase jitter is the catch; it means you need very good control over your acquisition system if you want those numbers to stick in real hardware; it shows the method isn't totally robust when things get noisy and imperfect <ref:2608.18252#pg2>.
Kai: The authors stress that this resolution gain comes with a trade-off, because both NSI variants reduced the generalized contrast to noise ratio compared to other methods like delay and sum on phantoms and in vivo data <ref:2608.18252#pg2>.
Mira: That’s a fair warning; you’re trading some overall image quality for better spatial detail in that specific mode; they suggest that high resolution power Doppler imaging might be the most promising use case because of that lower computational cost you mentioned earlier <ref:2608.18252#pg2>.
Lev: If we focus on high resolution Doppler, it makes sense because the overhead lands much lower than other complex nonlinear methods; it’s a practical application where this technique could actually save you processing time on a real system <ref:2608.18252#pg1>.
Kai: Ultimately, the central result of "Angle-domain null subtraction imaging from beamformed plane-wave data" is that you can achieve substantial lateral resolution improvement by using simple post processing of per angle images at a low computational cost <ref:2608.18252#pg2>.
Lev: So, for real hardware implementation, it gives us a specific technique that's faster than Receive-NSI or other nonlinear methods under tested conditions, which is what we need for high speed acquisition <ref:2608.18252#pg1>.
Kai: We’ll have to keep an eye on those future in vivo studies they mentioned because confirming its use for contrast-enhanced and contrast-free Doppler ultrasound would be the next big step <ref:2608.18252#pg1>.
Mira: Agreed, focusing on high resolution power Doppler seems like the most immediate practical application based on what they've shown so far <ref:2608.18252#pg2>.
Lev: For now, we just see how well this framework translates from simulation to running on real acquisition hardware where phase jitter is present and where the actual error correction overhead lands <ref:2608.18252#pg1>.
Conclusion: Kai: So we’re wrapping up our look at "Angle-domain null subtraction imaging from beamformed plane-wave data." This paper shows you can use matrix math to simplify post processing for per angle images, cutting down on computational cost while narrowing the apparent lateral response.
Lev: From an error correction standpoint, that separation is key because if you’re trying to build a real hardware pipeline, you need to know exactly what data access you're requiring; this method seems designed to keep the RF processing fixed while making the post-processing simpler <ref:2608.18252#pg1>.
Lev: I've seen that trade-off a lot when we push for higher sampling rates in error correction, but seeing it quantified this clearly makes the decision easier; it tells you exactly what your hardware will deliver versus what you need to run your algorithm on <ref:2608.18252#pg2>.
Lev: For now, we just see how well this framework translates from simulation to running on real acquisition hardware where phase jitter is present and where the actual error correction overhead lands <ref:2608.18252#pg1>.
Laboratory of Wave Engineering, School of Engineering, Ecole Polytechnique Federale de Lausanne (EPFL)
physics.med-ph
Submitted: 2026-08-18
Updated: 2026-10-15
Code: https://github.com/hacheleroy/angle_
License: http://creativecommons.org/licenses/by-nc-sa/4.0/
Importance score: 86/100
The gist: Angle-NSI provides simple post-processing of per-angle beamformed images to narrow the apparent lateral response at low computational cost It narrows the apparent lateral response at low…
Key concepts
- Angle-NSI
- A post-processing technique that uses a matrix formulation on per-angle images during coherent plane-wave compounding to perform angular null subtraction. It aims to narrow the apparent lateral response of the signal without requiring access to raw RF data, making it computationally inexpensive.
- Lateral Response Narrowing
- The primary goal is to reduce the apparent width of a target signal in the lateral direction. Angle-NSI achieved this by effectively narrowing the simulated -6 dB lateral width significantly, improving resolution at a low computational expense.
- Resolution-Detectability Trade-off
- A key finding is that improving lateral resolution often comes at the cost of image quality, specifically reducing the generalized contrast-to-noise ratio (gCNR) compared to conventional methods like DAS. This trade-off must be considered when choosing an imaging technique.
- Per-Angle Images
- The method operates on images that are formed for each individual angle during coherent plane-wave compounding. Angle-NSI utilizes these pre-formed per-angle images for simple post-processing, avoiding the need for complex RF data access or extensive reconstruction.
Terminology
Summary
Angle-NSI provides simple post-processing of per-angle beamformed images to narrow the apparent lateral response at low computational cost It narrows the apparent lateral response at low computational cost, at the expense of Bmode detectability The method was evaluated using a simulated point target, experimental resolution phantom targets, an open microbubble (MB) traces dataset and two in vivo carotid acquisitions Angle-NSI reduced the simulated −6 dB lateral width from 0. p. 0 Angle-NSI provides simple post-processing of per-angle beamformed images. It narrows the apparent lateral response at low computational cost, at the expense of Bmode detectability: this makes it a well-suited alternative for high-resolution power Doppler
applications. Angle-NSI reduced the simulated −6 dB lateral width from 0. 226 mm for
DAS to 0. Angle-NSI reduced the simulated −6 dB lateral width from 0. 226 mm for
DAS to 0. 226 mm for
DAS to 0. 116 mm, compared with 0. 116 mm, compared with 0. 009 mm for Receive-NSI, while the axial width was
0.Where the paper says it
Where the paper says it
Where the paper says it
Where the paper says it
Where the paper says it
Methodology of Angle-Domain Null Subtraction
Angle-NSI operates on per-angle images during coherent plane-wave compounding The approach uses a matrix formulation to distinguish receive-domain and angle-domain null subtraction expressions and to identify their independent computational components For the angular null, the method is analogous to Receive-NSI that creates an antisymmetric transmit-angle null while Receive-NSI creates the null on receive The angular null field is expressed as Eθ(r) = N (U(r), Zθ(r); c) This construction allows for a simple post-processing technique for beamforming pipelines retaining per-angle images without RF data access
Comparison with Other Beamforming Techniques
Angle-NSI was compared to several nonlinear beamforming methods, including delayandsum (DAS), coherencefactorweighted DAS (CF-DAS), minimumvariance beamforming (MV), delaymultiplyandsum (DMAS), Receive-NSI and Angle-NSI The computational interpretation shows that Angle-NSI accumulates U and Zθ from the same per-angle beamformed images before angular compounding, leading to a low incremental cost compared to DAS Table 1 summarizes the resulting distinction between Receive-NSI and Angle-NSI regarding required data and additional operations
Performance Metrics Across Imaging Modalities
The performance of Angle-NSI is evaluated across various scenarios, including simulated point targets, experimental resolution phantoms, open microbubble (MB) traces dataset, and in vivo carotid acquisitions In the simulated point-target response, Angle-NSI reduced the lateral −6 dB width from 0.226 mm for DAS to 0.116 mm In experimental resolution phantom studies, Angle-NSI yielded a lateral width of 0. p. 0 The lateral narrowing was confirmed on an experimental
resolution phantom. Angle-NSI had shorter reconstruction time than Receive-NSI and
the lowest post-delay processing time among the nonlinear methods. Angle-NSI had shorter reconstruction time than Receive-NSI and
the lowest post-delay processing time among the nonlinear methods. In power Doppler,
microbubble trace widths were reduced by 50.Where the paper says it
Where the paper says it
Robustness and Trade-offs
Robustness tests indicate that cancellation across angles is an intrinsic limitation, as width varied with angle sampling and increased under phase jitter The robustness to noise showed that Angle-NSI width was 0.122 ± 0.005 mm at 10 dB SNR, but increased to 0. p. 8 At 10 dB SNR, Angle-NSI width was 0. 005 mm; a 10◦ phase-jitter standard
deviation increased it to 0.Where the paper says it
Conclusion
In this Note, Angle-NSI approximately halved the apparent lateral width in simulation A comprehensive comparison with other beamforming techniques using simulation, a resolution phantom, in vitro Doppler data and a carotid acquisition shows that Angle-NSI increases the lateral resolution at the expense of B-Mode detectability at a low computational cost Angle-NSI offers simple, low-cost post-processing for high-resolution power Doppler The use of Angle-NSI for contrast-enhanced and contrast-free Doppler ultrasound should be confirmed by further in vivo studies The central result is methodological: the NSI framework can use a null formed across angles after receive beamforming, i.e. it can be adapted from receive apodization to angular apodization The timing study shows the low incremental cost of Angle-NSI under the tested conditions, using per-angle images after ordinary receive beamforming, making it faster than Receive-NSI or other nonlinear beamforming methods The experimental phantom and the carotid results emphasize the resolution–detectability trade-off, as both NSI methods decreased gCNR compared to DAS or other nonlinear beamforming methods The robustness tests identify cancellation across angles as an intrinsic limitation Eventually, the power Doppler study shows that high-resolution Doppler imaging is the most promising use case for Angle-NSI, as it effectively narrows the lateral response of the Doppler signal at a lower computational cost and pipeline complexity compared to Receive-NSI The use of Angle-NSI for contrast-enhanced and contrast-free Doppler ultrasound, like Receive-NSI (Kou et al. p. 9 Their different PSFs, microbubble traces and carotid images confirm that the methods are
distinct. The timing study shows the low incremental cost of Angle-NSI under the tested conditions,
using per-angle images after ordinary receive beamforming, making it faster than Receive-NSI or
other nonlinear beamforming methods.Where the paper says it
Green — the paper says it, and the page is printed. Red — no passage carries it, so treat that sentence as unverified. The same check runs on your own text: check a summary against its source, $5.
Improvements for AI systems
-
Improve high resolution power Doppler imaging by using Angle-NSI for simple post-processing of per-angle beamformed images to narrow lateral response at a low computational cost. This allows for a well suited alternative for high-resolution power Doppler applications.
-
Develop ultrasound reconstruction pipelines that retain per-angle images while keeping RF processing fixed, leveraging Angle-NSI's advantage over Receive-NSI in terms of computational efficiency and accessibility for research software or scanners.
-
Enhance microbubble trace analysis by applying Angle-NSI to reduce trace widths, achieving a 30.9% reduction compared to DAS in power Doppler settings, which improves the detection of microvessel structures.
-
Create a robust beamforming comparison tool that benchmarks six methods (DAS, CF-DAS, MV, DMAS, Receive-NSI, and Angle-NSI) across simulation and experimental data to identify the most computationally light technique once DAS is excluded.
-
Implement an adaptive contrast enhancement framework for ultrasound by utilizing Angle-NSI as a post-filtering approach to manage the trade-off between lateral resolution and image quality, acknowledging that
Angle-NSI provides simple post-processing of per-angle beamformed images.
Abstract
Null subtraction imaging (NSI) narrows the apparent lateral response by combining fields produced with a zero-mean apodization and two DC-offset variants. Usually, NSI requires channel data and flexible beamformers (Receive-NSI). We investigate Angle-NSI, which operates on per-angle images during coherent plane-wave compounding. A matrix formulation was used to distinguish receive-domain and angle-domain null subtraction expressions. The method was evaluated using a simulated point target, experimental resolution phantom targets, an open microbubble (MB) traces dataset and two in vivo carotid acquisitions. Angle sampling, noise, phase jitter and target position were swept in simulation. A synchronized GPU benchmark compared the complete reconstructions. Six beamforming methods were compared: delay-and-sum (DAS), coherence-factor-weighted DAS (CF-DAS), minimum variance beamforming (MV), delay-multiply-and-sum (DMAS), Receive-NSI and Angle-NSI. Angle-NSI reduced the simulated-6 dB lateral width from 0.226 mm for DAS to 0.116 mm, compared with 0.009 mm for Receive-NSI, while the axial width was 0.720 mm for these three methods. The lateral narrowing was confirmed on an experimental resolution phantom. Angle-NSI had shorter reconstruction time than Receive-NSI and the lowest post-delay processing time among the nonlinear methods. In power Doppler, microbubble trace widths were reduced by 50.2% for Receive-NSI and 30.9% for Angle-NSI compared to DAS. On a phantom and in vivo, both NSI variants reduced generalized contrast-to-noise ratio relative to the conventional methods. Angle-NSI provides simple post-processing of per-angle beamformed images. It narrows the apparent lateral response at low computational cost, at the expense of B-mode detectability: this makes it a well-suited alternative for high-resolution power Doppler applications.
Related papers
- BART Online Open-Source Sequence Toolbox for Computational MRI
- Constrained estimation of rotational invariants of the cumulant expansion (RICE) for rapid tensor-valued diffusion MRI
- Improved Myelin Water Fraction Estimation and Uncertainty Mapping in Multicomponent T2 Relaxometry via NNLS Bootstrap Aggregation
- Radiation damage to normal mammalian tissue in vivo with laser-driven protons at ultra-high instantaneous dose rate
- BEAM3R: Beam's-eye-view architecture with Mamba-3 for implicit dose reconstruction
- MORSE-PI -- Flexible and artefact-free image reconstruction for structural and functional QSM and other phase-critical imaging applications