
Photoacoustic tomography turns absorbed light into sound: a short optical pulse produces pressure waves that ultrasound detectors record and software reconstructs into an image. But electrical interference can enter before reconstruction, leaving stripes or drifting baselines in the detector signals. A study published on 2 October 2026 tested a two-stage, learning-free denoising method for this problem. The work used simulated data and in vivo finger and zebrafish datasets; it is a technical imaging study, not a clinical validation. [1]
Light in, sound out
In photoacoustic imaging, absorbed optical energy causes a small, rapid temperature rise and thermoelastic expansion. That expansion launches an acoustic wave; detectors measure the wave, and reconstruction methods infer where the initial absorption occurred. The image therefore combines optical absorption contrast with acoustic detection and propagation. [2]
This chain contains several distinct places where quality can be lost. Tissue affects light delivery and sound propagation; the detector and electronics add their own response; reconstruction translates measured waveforms into a spatial estimate. A denoising method that improves one stage does not automatically correct errors in the others.
The noise had structure
The 2026 study focused on electromagnetic interference in a ring-based system. Its authors describe interference that appears across channels at nearly the same times as stripe-like patterns in raw data, along with baseline drift and residual random noise within individual channels. A ring array records signals at multiple detector positions, so a shared disturbance can look different from a local fluctuation even before an image is reconstructed. [1]
That distinction suggests a practical sequence: first identify disturbances that are synchronized across channels, then address slower temporal changes within each channel. Treating every sample as independent could miss the shared pattern; smoothing everything equally could erase genuine acoustic detail.
A two-stage, learning-free approach
The proposed Structured-Sequential Denoising framework first applies robust order-statistic processing to suppress quasi-synchronous stripe-like interference. It uses an interquartile-range-based procedure to identify affected time indices. A second, single-channel temporal restoration stage targets baseline drift and remaining random noise. The method is described as physics-motivated and learning-free, so it does not require a trained neural network to map noisy traces to cleaner ones. [1]
“Learning-free” does not mean assumption-free. The method still relies on patterns in the data being separable: interference shared across channels should differ from the acoustic signal structure the system aims to preserve. If a true wave arrives in a pattern that resembles a common-mode artifact, indiscriminate removal could damage the measurement. The relevant test is thus not simply whether the trace looks smoother, but whether useful acoustic information survives.
What the experiments establish
The authors evaluated simulated signals and in vivo finger and zebrafish datasets. Against selected conventional and filtering baselines, they report improved image-quality measures on their simulation and higher mean A-line signal-to-noise ratios over noisy input for the two in vivo datasets. An A-line is a one-dimensional depth profile; its signal-to-noise ratio describes signal strength relative to noise under the study’s calculation. These measurements support the claim that the method improved data quality in the tested setting. [1]
The boundaries matter. A simulated image-quality score depends on the chosen reference and noise model. An increased A-line signal-to-noise ratio is not itself proof of better tissue classification, more accurate physiology, or a useful diagnosis. Finger and zebrafish experiments demonstrate application to those datasets, not clinical benefit for patients. A practical comparison would also ask whether important waveform features and reconstructed structures remain faithful, not only whether noise falls.
Why denoising is not the whole inverse problem
Photoacoustic reconstruction is an inverse problem: software tries to infer an internal source from acoustic measurements made around or over an object. Detector coverage, sampling, acoustic speed variation, and reconstruction assumptions can all shape the result. A tutorial review of the field emphasizes the linked roles of imaging physics, instrumentation, standardization, and interpretation. Cleaning raw traces can help, but it cannot create information that a limited detector view never captured or fully repair an incorrect propagation model. [2]
A useful mental model is a microphone recording in a reverberant room. Removing a shared electrical buzz can make the recording easier to analyze, but it cannot tell you where a sound source was if the microphones missed the relevant directions. In tomography, signal cleanup and image reconstruction are connected stages with different failure modes.
The next checks
Further evaluation should test more hardware configurations and acquisition conditions, include repeat measurements, and compare preserved signal features against independent references. Researchers would also need to establish how parameter choices behave when the interference pattern changes, and whether denoising affects downstream quantitative measurements. The tutorial literature points to system design and standardization as part of practical translation, not optional details after an algorithm is developed. [2]
The October study offers a focused engineering contribution: exploit the difference between cross-channel interference and within-channel drift, and test the sequence on simulated and in vivo data. Its evidence supports a promising cleanup method for the studied ring-array setup. Whether it generalizes across devices or improves a health decision remains an open question.
Sources & further reading
- Structured-sequential denoising for ring-based photoacoustic tomography npj Imaging (2026)
- Photoacoustic imaging on its way toward clinical utility: a tutorial review focusing on practical application in medicine Journal of Biomedical Optics (2023)
Sources checked: 14 November 2026. The calculations illustrate physical models; they are not clinical settings.
Prepared for the independent PSI Center reading library. About edition dates and source checking.
This explainer describes scientific principles. It does not assess individual treatment choices.


