How locating sensors in thermo-acoustic tomography?

TL;DR

This paper develops a mathematical and numerical framework for optimally locating sensors in thermo-acoustic tomography to improve the accuracy of internal tissue reconstruction from acoustic measurements.

Contribution

It introduces a novel two-step approach combining worst-case data reconstruction and sensor placement optimization, with mathematical analysis and numerical simulations.

Findings

Optimal sensor locations improve reconstruction quality.

The model ensures existence of solutions and provides a practical algorithm.

Numerical simulations validate the effectiveness of the proposed method.

Abstract

Thermo-acoustic tomography is a non-invasive medical imaging technique, constituting a precise and cheap alternative to X-imaging. The principle is to excite a body to reconstruct with a pulse inducing an inhomogeneous heating and therefore expansion of tissues. This creates an acoustic wave pressure which is measured with sensors. The reconstruction of heterogeneities inside the body can be then performed by solving an inverse problem, knowing measurements of the acoustic waves outside the body. As the intensity of the measured pressure is expected to be small, a challenging problem consists in locating the sensors in a adequate way. This paper is devoted to the determination of an optimal sensors location to achieve this reconstruction. We first introduce a model involving a least square functional standing for an observation of the pressure for a first series of measures by sensors, and an observability-like constant functional describing for the quality of reconstruction. Then, we determine an appropriate location of sensors for two series of measures, in two steps: first, we reconstruct possible initial data by solving a worst-case design like problem. Second, we determine from the knowledge of these initial conditions the optimal location of sensors for observing in the best way the corresponding solution of the wave equation. Far from providing an intrinsic solution to the general issue of locating sensors, solving this problem allows to determine a new sensors location improving the quality of reconstruction before getting a new series of measures. We perform a mathematical analysis of this model: in particular we investigate existence issues and introduce a numerical algorithm to solve it. Eventually, several numerical 2D simulations illustrate our approach.