Talk by Changyou Li, Technical University of Denmark

A computational approach for acousto-electric tomography based on Levenberg-Marquardt algorithm

Abstract

Acousto-electric tomography (AET) is a hybrid conductivity imaging technique. It is essentially the electrical impedance tomography (EIT) enriched by the modulation effects of focused ultrasound wave on the localized conductivity. It reconstructs conductivity map of physical bodies from measured internal power density and boundary potentials or currents. Most existing computational methods and mathematical analysis of AET were developed by using the ideal continuum model with either Dirichlet or Neumann boundary conditions. This talk will introduce a more practical and accurate computational approach for AET. It will first build the computational reconstruction approach based on Levenberg-Marquardt iteration and complete electrode model (CEM) which is the most accurate forward model for EIT. Numerical experiments shows that this pure CEM-based reconstruction is sensitive to input current pattern and regularization parameters for Levenberg-Marquardt iteration. To overcome this problem, CEM and the ideal continuum model with Dirichlet boundary conditions are mixed in the Levenberg-Marquardt iterations.