[Paper Review] Mathematics of thermoacoustic tomography

Peter Kuchment, Leonid Kunyansky|ArXiv.org|Apr 2, 2007

Photoacoustic and Ultrasonic Imaging82 references4 citations

TL;DR

This paper provides a comprehensive mathematical survey of thermoacoustic tomography (TAT), formulating the imaging problem as an inverse problem for the wave equation or spherical mean operator. It establishes uniqueness, inversion formulas, and stability conditions, with key results on injectivity sets and range conditions for various geometries and function spaces.

ABSTRACT

The paper presents a survey of mathematical problems, techniques, and challenges arising in the Thermoacoustic and Photoacoustic Tomography.

Motivation & Objective

• To systematically analyze the mathematical challenges in thermoacoustic tomography (TAT), a hybrid medical imaging modality combining electromagnetic and acoustic signals.
• To address the inverse problem of reconstructing the initial energy absorption distribution from measured acoustic pressure data.
• To investigate the theoretical conditions under which the reconstruction is unique, stable, and invertible.
• To provide a detailed treatment of mathematical models, including the spherical mean transform and wave equation formulations.
• To identify open problems and unresolved issues in TAT reconstruction, particularly concerning non-compact support and non-spherical observation surfaces.

Proposed method

• Formulates TAT as an inverse problem for the wave equation with initial condition given by the energy absorption function $ f(x) $.
• Reduces the problem to the spherical mean transform (spherical Radon transform) when sound speed is constant, enabling analysis via integral geometry.
• Applies tools from integral geometry and microlocal analysis, including the $ \kappa $-operator and Fourier analysis, to study inversion and range conditions.
• Uses the method of Fourier transforms in time and spatial transforms to derive reconstruction formulas and stability estimates.
• Analyzes the role of observation surface geometry (plane, cylinder, sphere, general closed surfaces) in determining the feasibility and structure of inversion formulas.
• Considers the impact of partial data, non-compact support, and non-constant sound speed on the solvability and stability of the inverse problem.

Experimental results

Research questions

• RQ1 Under what conditions is the initial energy absorption function $ f(x) $ uniquely recoverable from acoustic pressure measurements on a surface $ S $?
• RQ2 Can explicit inversion formulas be derived for non-spherical observation surfaces when the sound speed is constant?
• RQ3 What are the necessary and sufficient conditions on the data (i.e., the range) for the spherical mean transform in general geometries?
• RQ4 How does the stability of reconstruction depend on the geometry of the observation surface and the location of the support of $ f(x) $?
• RQ5 To what extent do singularities of $ f(x) $ outside the observation surface remain stably recoverable?

Key findings

• Uniqueness of reconstruction holds for compactly supported $ f \in L^p(\mathbb{R}^d) $ with $ p \leq 2d/(d-1) $, and the set $ S $ is an injectivity set for such functions.
• For $ p > 2d/(d-1) $, uniqueness fails, indicating a sharp threshold in function space integrability for injectivity.
• Closed-form inversion formulas exist for $ S $ being a plane, cylinder, or sphere, but their validity breaks down when the support of $ f $ extends beyond $ S $.
• Stability of reconstruction is lost when parts of $ f $'s support lie outside the observation surface $ S $, as shown by Theorem 9.
• Range conditions for the spherical mean transform are necessary but not expected to be sufficient for general closed surfaces $ S $, especially when $ f $ has support outside $ S $.
• The existence of explicit inversion formulas for spherical $ S $ contradicts expectations from the $ \kappa $-operator theory, leaving a theoretical paradox unresolved.