Ultrasound Engineering
Ultrasound Receive Apodization: Judge the Trade-off
A short source-led review of ultrasound receive apodization, resolution–sidelobe trade-offs, recent learned and two-way window designs, and proposed reproducible OEM acceptance tests.
By ALEVA | California, USA
Literature analysis and a proposed evaluation protocol.
The engineering question
When should an ultrasound OEM change receive-aperture weights instead of adding channels or buying a faster processor? Start by asking which image property needs to improve, and what must remain unchanged.
Weighting is part of image formation
In delay-and-sum reconstruction, delayed channel signals are multiplied by receive weights before summation. Conventional tapering can suppress sidelobes at the cost of a wider main lobe. Thus, a cleaner-looking background does not automatically establish better separation of nearby targets. The 2025 null-subtraction imaging analysis also illustrates why nonlinear combinations of differently weighted beams need their own resolution model; a conventional window comparison cannot describe every beamformer. [1]
Recent work changes the design space
OATS learns depth-dependent receive weights using differentiable beamforming and simulated training data. Its authors evaluated focused and unfocused imaging using simulated and real-world data. This is a peer-reviewed journal study, published online in 2025 and assigned to the March 2026 issue of Ultrasonics—not evidence that a particular commercial system will reproduce its improvements. [2]
A separate study in the January 2026 issue of Signal Processing jointly designs transmit and receive windows. Under its array model, the effective two-way aperture depends on both windows. Its simulation results motivate evaluating the transmit–receive pair, rather than treating a receive window as an isolated accessory. [3]
A proposed OEM acceptance test
Replay the same channel dataset through the baseline and candidate implementations. Hold the transmit sequence, aperture geometry, delay model, gain, envelope processing and display mapping constant wherever the comparison permits. Document every deliberate exception.
Evaluate point-target width and sidelobes alongside background texture, weak-target preservation and low-echo-region fill-in, across depth and steering. Record runtime and coefficient storage on the intended hardware. If weights are learned, separate training and evaluation cases and include conditions outside the training set. These are proposed engineering checks, not reported experimental results.
The OEM/ODM decision
Specify the weighting method and coefficient version in the imaging baseline, and agree on acceptance criteria before optimization. A candidate should justify its image-quality trade-off and integration cost. A darker image alone is not an acceptance criterion, and these papers do not establish clinical superiority or ALEVA implementation of their methods.
References
Han C., Näsholm S. P., Austeng A., Arnestad H. K. Taylor-Series-Based Derivation of the Resolution of Null Subtraction Imaging for a Uniform Linear Array. IEEE OJUFFC 5, 19–22, 10 March 2025. DOI: 10.1109/OJUFFC.2025.3550096. University of Oslo summary.
Xiao D., Nahas H., Hiroshima M., Kishimoto A., Yu A. C. H. Optimized Apodizations with Training Simulations (OATS): Learned depth-dependent apodizations via differentiable beamforming for reduced operator tuning. Ultrasonics 159, 107827, March 2026; PubMed article date 24 September 2025. DOI: 10.1016/j.ultras.2025.107827. PubMed record.
Arnestad H. K., Austeng A., Näsholm S. P. Joint design of transmit and receive array window functions via polynomial factorization. Signal Processing 238, 110124, January 2026. DOI: 10.1016/j.sigpro.2025.110124.