🔬 applied physics

An angular distortion matrix approach for joint wave-speed tomography and aberration correction in scattering media

This paper introduces an iterative angular distortion matrix approach that leverages time-reversal analysis to simultaneously estimate wave-speed distributions and correct phase aberrations in scattering media, thereby significantly enhancing reflection imaging quality for clinical ultrasound applications.

Original authors: Chenyu Cui, Flavien Bureau, Foucauld Chamming's, Alexandre Aubry, Naiara Korta Martiartu

Published 2026-09-10

📖 6 min read🧠 Deep dive

CC BY 4.0

Original authors: Chenyu Cui, Flavien Bureau, Foucauld Chamming's, Alexandre Aubry, Naiara Korta Martiartu

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

Sound waves are nature's way of seeing what lies hidden. By sending vibrations into a material and listening for the echoes that bounce back, scientists and doctors can reconstruct a picture of the inside of objects that cannot be opened, from the Earth's crust deep beneath our feet to the soft tissues inside a human body. This technique, known as reflection imaging, is the foundation of ultrasound scans used in hospitals every day. However, the clarity of these pictures depends entirely on knowing exactly how fast sound travels through the material being scanned. In a perfect, uniform world, this speed is constant, but real bodies are complex. They are made of layers of fat, muscle, and organs, each with a slightly different speed of sound. When the imaging machine assumes a single, average speed for the whole body, the waves get distorted as they pass through these different layers, much like light bending when it passes through a warped piece of glass. These distortions, called aberrations, blur the final image and can hide important details or make healthy tissue look sick.

For decades, researchers have faced a difficult catch-22: to fix the blurry image, they need to know the true speed of sound in every part of the body, but to measure that speed, they need a clear image to begin with. The distortions that ruin the picture also carry the clues needed to fix it, but untangling those clues from the noise of the tissue itself has been a major hurdle. A new study by a team of researchers from Switzerland and France has found a way to break this cycle. They developed a method that simultaneously maps the speed of sound and corrects the image distortions, turning a blurry, unreliable scan into a sharp, quantitative map of tissue properties.

The researchers tested their approach using a gelatin block designed to mimic human tissue, complete with layers of fat and muscle that have different sound speeds. They also applied the method to real patients, scanning the livers of healthy volunteers and the breasts of patients with confirmed cancer. In the lab, the gelatin block contained distinct regions where sound traveled at speeds ranging from 1420 to 1550 meters per second. When the team initially imaged this block using a standard assumption that sound travels at a uniform 1540 meters per second, the resulting picture was fuzzy, and the layers were hard to distinguish. The waves had been scrambled by the speed differences, causing the image to lose its focus.

To solve this, the team used a technique called matrix imaging. Instead of just sending a single beam of sound and listening for an echo, they recorded how the sound behaved when sent from every possible angle and received by every sensor in the probe. This created a massive amount of data describing how the waves interacted with the tissue. The key innovation was a new mathematical tool they called an "angular distortion matrix." This tool allowed them to look at the data in a specific way that isolated the distortions caused by the tissue's speed variations from the actual texture of the tissue itself. By analyzing how the waves correlated with each other across different angles, they could extract the specific phase errors—the timing delays—that the waves accumulated as they traveled through the body.

These timing errors were then used to build a new, more accurate map of the sound speed inside the body. With this new map, the researchers could update their model of how sound travels and re-process the original data. This process was repeated in a loop: the improved speed map led to a sharper image, which in turn allowed for an even more precise speed map. After just a few rounds of this refinement, the blurry gelatin block transformed into a crisp image where the different layers of fat and muscle were clearly separated. The final map of sound speeds was accurate to within one percent of the known values, proving that the method could recover the true physical properties of the tissue.

The true test came when the team applied this technique to living humans. In a breast imaging case involving a patient with invasive carcinoma, the standard ultrasound image showed a lesion, but the surrounding tissue was obscured by the complex mix of fat and glandular tissue. The new method corrected the distortions caused by these layers, revealing a sharp, clear view of the tumor. More importantly, the speed map showed that the cancerous tissue had a higher sound speed than the surrounding healthy tissue, a known characteristic of malignant growths. The corrected image also revealed that the lesion appeared in a slightly different location than the standard scan suggested, proving that the uncorrected distortions had been misleading the eye.

In the liver scans of healthy volunteers, the method proved equally powerful. The abdominal wall is a thick stack of skin, fat, and muscle that often ruins the quality of liver images. The new approach successfully mapped the speed of sound through each of these layers, allowing the researchers to see deep into the liver with unprecedented clarity. The resulting images showed fine details and blood vessels that were previously invisible, and the map of sound speeds matched the expected values for healthy liver tissue. The technique worked even in the presence of complex, multi-layered structures, demonstrating that it could handle the messy reality of human anatomy.

This work does more than just sharpen pictures; it provides a way to measure the physical state of tissue without cutting into it. Sound speed is a biomarker, a measurable sign of what is happening inside the body. In the breast, a higher speed can signal cancer, while in the liver, a lower speed can indicate the buildup of fat. By correcting the image distortions at the same time as measuring these speeds, the method offers doctors two powerful tools at once: a clearer view of anatomy and a quantitative measure of tissue health. The researchers showed that this approach works not just in controlled lab settings but in the complex, variable environment of the human body, opening the door to more reliable diagnoses and a deeper understanding of how sound interacts with life.

Technical Summary: Angular Distortion Matrix Approach for Joint Wave-Speed Tomography and Aberration Correction

Problem Statement
In reflection imaging (ultrasound, optics, seismology), accurate reconstruction of medium reflectivity relies on a precise wave-propagation model governed by the wave-speed distribution. However, in complex scattering media, the wave speed is rarely known a priori. Mismatches between the assumed and actual wave speeds introduce phase aberrations that degrade focusing, broadening the point-spread function and distorting the apparent location of scatterers. While these aberrations encode information about the underlying wave-speed heterogeneities, disentangling them from medium reflectivity is difficult, particularly in speckle-dominated media. Existing methods face a fundamental coupling problem: correcting aberrations requires an accurate wave-speed model, yet retrieving this model requires observables that are themselves unbiased by aberrations. Furthermore, traditional distortion matrix approaches assume aberrations arise from either the incident or reflected path, whereas in reality, both contributions are inherently mixed, leading to depth shifts that cannot be corrected without updating the velocity model.

Methodology
The authors propose a unified framework that iteratively estimates wave-speed distribution and corrects aberrations using a novel angular distortion matrix ( $D_{\theta\theta}$ ). The methodology proceeds as follows:

Matrix Imaging and Frequency-Wavenumber Representation:
The process begins by acquiring a reflection matrix $R_{uu}$ (transducer-to-transducer responses). This is projected into a focused basis to form a focused reflection matrix $R_{xx}$ . To isolate distortions, the authors introduce a frequency-wavenumber representation of wave distortions, $D_{kk}$ . This operator is constructed by removing the expected geometrical phase components from the responses as if scattering originated from an arbitrary point $x$ . This allows the synthesis of distortions for any virtual experiment in post-processing.
Angular Distortion Matrix Construction:
The $D_{kk}$ operator is projected onto a Gaussian-beam basis to form the angular distortion matrix $D_{\theta\theta}$ . This basis utilizes steered Gaussian beams defined by incident ( $\theta_{in}$ ) and reflected ( $\theta_{out}$ ) angles. This projection leverages the angular memory effect, where waves exhibit strong correlations along specific angular directions despite disorder. By analyzing the antidiagonals of $D_{\theta\theta}$ (constant mid-angle $\theta_m = (\theta_{in} + \theta_{out})/2$ ), the method isolates phase aberrations associated with well-defined propagation paths.
Local Aberration Estimation:
The field of view (FOV) is divided into overlapping "isoplanatic patches." Within each patch, a local correlation matrix is computed from the angular distortion matrix elements. Under the assumption that wave distortions are invariant over small regions, the first eigenvector of this correlation matrix provides an estimate of the local phase aberration law ( $u_m$ ). To avoid phase aliasing, the method utilizes relative phase differences between successive half-difference angles ( $\theta_d$ ).
Iterative Inversion for Wave-Speed Tomography:
The estimated phase aberrations are used to reconstruct the sound-speed distribution. The problem is formulated as an optimization task to minimize the misfit between the measured phase differences and those predicted by a propagation model. The inversion is performed iteratively:
- An initial homogeneous model (e.g., $c_0 = 1540$ m/s) is assumed.
- Aberrations are estimated and corrected.
- A linearized inverse problem (using a straight-ray approximation for the Jacobian and Tikhonov regularization) solves for slowness perturbations ( $\Delta s$ ).
- The wave-propagation model is updated, and the process repeats until residual aberrations become negligible.
Handling Multiple Scattering:
In highly heterogeneous media (e.g., breast tissue) where multiple scattering degrades the memory effect, an adaptive confocal filter is applied to the reflection matrix before computing the distortion matrix. This suppresses incoherent background noise while preserving the diagonal energy associated with single scattering.

Key Results
The approach was validated using ultrasound experiments in three scenarios:

Tissue-Mimicking Phantom: A phantom with known layered structures (fat, muscle, liver) and varying sound speeds (1420–1550 m/s) was used. The iterative method successfully recovered the sound-speed map with mean errors within 1% of the ground truth. The Strehl ratio (a measure of focusing quality) improved from 0.34 to 0.96, and reflectivity images showed sharp interfaces and enhanced speckle contrast.
In Vivo Breast Imaging: Applied to a patient with invasive carcinoma. The method identified a localized region of elevated sound speed (~1574 m/s) consistent with malignant lesions, distinguishing it from surrounding adipose and fibroglandular tissue. Aberration correction revealed that uncorrected images had biased the apparent depth of anatomical features.
In Vivo Liver Imaging: Applied to a healthy volunteer. The method resolved the layered structure of the abdominal wall (skin, fat, muscle) and the liver parenchyma with high axial resolution. The retrieved sound speeds matched literature values for healthy tissues, and the Strehl ratio increased from 0.61 to 0.93, significantly sharpening the reflectivity image and revealing previously obscured structures like anechoic veins.

Significance and Claims
The paper claims that the angular distortion matrix approach provides a robust solution to the coupled problem of wave-speed tomography and aberration correction. Its primary significance lies in:

Unification: It integrates distortion matrix concepts (typically used for aberration correction) with computed ultrasound tomography in echo mode (CUTE) within a single, iterative matrix-imaging framework.
Model Independence: The formulation of the angular distortion matrix is independent of the specific wave-propagation model used for focusing, allowing the method to progressively refine the model itself.
High Resolution: By projecting onto a Gaussian-beam basis, the method can estimate aberrations over smaller isoplanatic patches (approx. 1 mm) compared to classical distortion matrix approaches, enabling higher spatial resolution in tomography.
Clinical Utility: The method demonstrates the ability to provide quantitative sound-speed maps (potential biomarkers for pathology) and simultaneously improve structural image quality for clinical interpretation in complex, heterogeneous tissues.
Generality: While demonstrated with ultrasound, the framework is applicable to any wave modality (optics, seismology) where reflection-matrix imaging can be implemented.

Technical Summary: Angular Distortion Matrix Approach for Joint Wave-Speed Tomography and Aberration Correction

More like this