💻 computer science

A Full-waveform Approximation of Finite-Sized Acoustic Apertures: Forward and Adjoint Wavefields

This paper establishes an equivalence between analytic integral formulations and full-waveform approximations for finite-sized acoustic apertures to derive reception operators and their adjoints, thereby enhancing accurate amplitude modeling for applications such as therapeutic ultrasound optimization and photoacoustic tomography.

Original authors: Ashkan Javaherian, Seyed Kamaledin Setarehdan

Published 2026-04-07

📖 5 min read🧠 Deep dive

CC BY 4.0

Original authors: Ashkan Javaherian, Seyed Kamaledin Setarehdan

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

Imagine you are trying to understand how sound travels through a room, but instead of just listening to the echo, you want to predict exactly how a specific speaker will sound to a specific listener, accounting for every tiny detail of the speaker's size and shape.

This paper is about building a super-accurate mathematical model for that scenario, specifically for medical ultrasound and acoustic imaging. Here is the breakdown using simple analogies:

1. The Problem: The "Point Source" vs. The "Real Speaker"

In many physics textbooks, sound sources are treated as mathematical points—infinitely small dots that pop with sound. It's like saying a speaker is just a single pixel on a screen. This is easy to calculate, but it's not realistic.

Real speakers (or medical ultrasound transducers) have size. They are like actual drumheads or flat panels.

The Old Way: If you treat a drumhead as a single point, you miss out on how the sound waves interfere with each other across the surface of the drum. This leads to errors, especially when you need high precision (like focusing ultrasound to burn a tumor without hurting the surrounding tissue).
The Paper's Goal: The authors wanted to create a model that treats these sound sources as finite-sized objects (like real drums) but still uses the powerful, fast computer methods (Full-Waveform) that usually only work for simple points.

2. The Two Languages of Sound: "Monopoles" and "Dipoles"

The paper translates between two different ways of describing sound sources:

The "Monopole" (The Push): Imagine a piston pushing air straight out. This is like a monopole. In the math, this is described by how fast the air is moving (velocity) at the surface.
The "Dipole" (The Squeeze): Imagine a surface that vibrates but doesn't push air out the sides, only the front. This is like a dipole. In the math, this is described by the pressure itself on the surface.

The authors proved that you can switch between these two descriptions perfectly. They showed that if you know the pressure on a surface, you can mathematically "translate" it into a force that pushes the air, and vice versa.

3. The Magic Trick: Smearing the Source

How do you put a flat, 2D surface (like a drumhead) into a 3D computer simulation that only understands points?

The Analogy: Imagine trying to paint a sharp line on a canvas using a thick brush. You can't get a perfect line, but if you use a "smart brush" that spreads the paint just a tiny bit, you can get very close.
The Math: The authors use a technique called regularization. Instead of putting the sound source exactly on a 2D line (which causes computer errors), they "smear" it out slightly into a thin 3D layer. This allows the computer to solve the equations smoothly without crashing, while still keeping the physics accurate.

4. The "Time-Travel" Camera (Adjoint Operators)

This is the coolest part of the paper. In medical imaging, we often try to solve the Inverse Problem: "We heard these echoes; where did the sound come from?"
To solve this, computers use a trick called Time Reversal (or the Adjoint Operator).

The Analogy: Imagine recording a video of a glass shattering. If you play the video backwards, the shards fly back together and the glass reforms.
The Application: In ultrasound, if you record the sound waves hitting a sensor, you can mathematically "play them backwards" to see where they originated.
The Breakthrough: The authors proved that when you do this "time reversal" for a finite-sized sensor (not just a point), the math looks exactly like a specific type of sound source called a Dipole.
- Why this matters: It means the computer doesn't need a complex, slow algorithm to reverse the sound. It can use a standard "time-reversed" simulation, provided it treats the sensor as a "Dipole" (a pressure-based source) rather than a simple point.

5. Why This Matters for You

This isn't just abstract math; it has real-world impacts:

Better Medical Imaging: It allows for clearer photos of the inside of the body (Photoacoustic Tomography).
Safer Treatments: It helps doctors focus ultrasound energy precisely on tumors (Therapeutic Ultrasound) without burning healthy tissue.
Accuracy: By accounting for the actual size of the sensors and speakers, the models stop making mistakes that happen when you assume everything is a tiny dot.

Summary

Think of this paper as a translator and a guide.

It translates the complex physics of real-sized speakers into a language that fast computers can understand.
It shows that if you want to reverse-engineer sound (to find out where it came from), you should treat your sensors as pressure-sensitive dipoles, not just simple points.
It proves that this new, more accurate way of thinking matches the old, slow, perfect math, giving us the best of both worlds: speed and precision.

1. Problem Statement

The paper addresses the challenge of accurately modeling acoustic wave propagation induced by finite-sized acoustic apertures (transducers) in biomedical applications such as ultrasound tomography, photoacoustic tomography, and therapeutic ultrasound optimization.

The Limitation of Current Models: Traditional full-waveform inversion (FWI) and time-reversal methods often approximate acoustic transducers as omnidirectional point sources or receivers. While computationally efficient, this assumption fails to capture critical physical phenomena at high frequencies, specifically:
- Obliquity effects: The angular dependence of the source/receiver sensitivity.
- Near-field effects: Complex wave interactions close to the aperture surface.
- Finite-size effects: The spatial extent of the transducer, which significantly impacts amplitude and phase accuracy.
The Mathematical Gap: Analytic solutions for finite apertures exist (Kirchhoff–Helmholtz and Rayleigh–Sommerfeld integrals), but they are often difficult to integrate directly into iterative optimization frameworks (like FWI) that require efficient forward and adjoint operators. Conversely, standard full-waveform solvers (e.g., k-Wave) typically lack rigorous formulations for surface-restricted sources that match these analytic integrals exactly.

2. Methodology

The authors propose a rigorous mathematical framework to bridge the gap between analytic integral formulas and full-waveform numerical simulations.

A. Theoretical Foundation

Equivalence of Formulations: The study establishes an equivalence between:
- Monopole Formulation: Derived from the Rayleigh–Sommerfeld integral for rigid baffles (Neumann boundary condition), representing the source as the normal derivative of pressure ( $-\partial p/\partial n$ ).
- Dipole Formulation: Derived for soft baffles (Dirichlet boundary condition), representing the source as the pressure itself ( $p$ ) acting with a normal derivative of the Green's function.
Regularized Source Construction: To implement these surface integrals in a volumetric full-waveform solver, the authors define regularized source terms using a smoothed Dirac delta distribution ( $\delta_b$ $δ_{b}$ ).
- Monopole Source: Mapped to a scalar mass source ( $S_m$ ) added to the continuity equation.
- Dipole Source: Mapped to a vector-valued force source ( $S_f$ ) added to the equation of motion.
- Key Insight: The dipole formulation requires a force source to correctly capture the obliquity factor ( $n \cdot x_d / x_d$ ) and near-field terms, whereas the monopole formulation uses a mass source.

B. Operator Derivation

The paper defines a complete set of operators for ultrasound tomography:

Extension Operator: Maps surface-restricted data (pressure or normal velocity) on a boundary to a volumetric source term in free space.
Restriction Operator: Maps the volumetric wavefield back to boundary data (pressure or normal derivative) on receiver surfaces.
Forward Operator ( $A$ ): Combines the extension, wave propagation (via a coupled first-order system), and restriction operators.
Adjoint Operator ( $A^*$ ): The authors derive the adjoint of the forward operator. They prove that under Dirichlet-type boundary data (pressure measurements), the adjoint operator coincides (up to a constant) with a time-reversed interior-field dipole integral formula. This means the adjoint simulation naturally accounts for the finite size and angular sensitivity of the receivers without needing explicit analytic integration.

C. Discretization

The continuous equations are discretized on a regular grid using a k-space pseudo-spectral method (Algorithm 1).

The surface sources are triangulated, and the regularized delta functions are applied to grid points near the surface.
The force source formulation (Eq. 87) is specifically designed to avoid the "omnidirectional" assumption inherent in simpler mass-source approximations.

3. Key Contributions

Rigorous Full-Waveform Approximation: The paper provides a mathematically proven method to approximate the Rayleigh–Sommerfeld dipole integral formula using a standard full-waveform solver by introducing a vector-valued force source.
Adjoint Operator Equivalence: It demonstrates that the adjoint of the forward operator (used in inverse problems) is effectively a time-reversed dipole integral. This validates the use of time-reversal back-projection in iterative inversion schemes for finite-sized receivers.
Correction of Common Approximations: The authors explicitly show that approximating a dipole source (finite aperture) as a scalar mass source (omnidirectional) introduces significant errors, particularly in the near-field and at oblique angles.
Unified Framework: The work unifies analytic integral methods and numerical PDE solvers, allowing the benefits of both (analytic accuracy and handling of complex heterogeneities) to be utilized in inverse problems.

4. Numerical Results

The authors validated their method using the k-Wave toolbox and compared results against analytic benchmarks (Field II toolbox and primary Green's function solutions).

Point Source Validation: The regularized mass source accurately reproduced the causal Green's function action on a point source.
Monopole Source (Rigid Baffle): A disk-shaped monopole source modeled via a discretized mass source (Eq. 85) showed excellent agreement with the analytic monopole integral formula.
Dipole Source (Soft Baffle):
- Scenario A (Mass Source Approximation): Modeling the dipole source as a scalar mass source (Eq. 86) resulted in large discrepancies compared to the analytic solution. This confirmed that the "omnidirectional" assumption fails to capture obliquity and near-field effects.
- Scenario B (Force Source Approximation): Modeling the dipole source using the proposed vector-valued force source (Eq. 87) yielded near-perfect agreement with the analytic dipole integral formula (Eq. 24) across various receiver positions, including near-field and off-axis locations.
Error Analysis: Relative error plots demonstrated that the force-source approach maintains high accuracy, while the mass-source approach degrades significantly as the receiver moves into the near-field or off-axis regions.

5. Significance and Impact

Improved Inverse Problems: By providing an accurate adjoint operator that accounts for finite receiver sizes, this method significantly improves the accuracy of Quantitative Ultrasound Tomography (QUT) and Photoacoustic Tomography (PAT). It allows for better reconstruction of attenuation and speed of sound maps.
Therapeutic Ultrasound: Accurate amplitude modeling is critical for safety in High-Intensity Focused Ultrasound (HIFU). This method ensures that the energy deposition patterns predicted by simulations match the physical reality of finite transducers.
Algorithmic Efficiency: The approach allows researchers to use efficient, grid-based full-waveform solvers (like k-Wave) to solve problems that previously required complex analytic integrals or boundary element methods, while retaining high fidelity for finite apertures.
Theoretical Clarity: The paper clarifies the relationship between "single-layer" (monopole) and "double-layer" (dipole) potentials and their corresponding source terms in the time-domain wave equation, resolving ambiguities in how surface sources are implemented in numerical codes.

In summary, this paper provides a critical advancement in computational acoustics by rigorously defining how to simulate finite-sized acoustic apertures in full-waveform frameworks, ensuring that both forward and adjoint operators accurately reflect the physics of real-world transducers.