Comparison of linear and non-linear monotononicity-based shape reconstruction using exact matrix characterizations

TL;DR

This paper compares linear and non-linear monotonicity-based methods for shape reconstruction in electrical impedance tomography, demonstrating their similar effectiveness and speed in idealized conditions, especially within the unit disk domain.

Contribution

It provides a direct comparison of linear and non-linear monotonicity methods using exact matrix characterizations, highlighting their comparable performance and computational efficiency.

Findings

Both methods yield similar reconstructions in the unit disk.

Exact matrix characterizations improve computational speed.

Non-linear method matches linear method in accuracy and speed.

Abstract

Detecting inhomogeneities in the electrical conductivity is a special case of the inverse problem in electrical impedance tomography, that leads to fast direct reconstruction methods. One such method can, under reasonable assumptions, exactly characterize the inhomogeneities based on monotonicity properties of either the Neumann-to-Dirichlet map (non-linear) or its Fr\'echet derivative (linear). We give a comparison of the non-linear and linear approach in the presence of measurement noise, and show numerically that the two methods give essentially the same reconstruction in the unit disk domain. For a fair comparison, exact matrix characterizations are used when probing the monotonicity relations to avoid errors from numerical solution to PDEs and numerical integration. Using a special factorization of the Neumann-to-Dirichlet map also makes the non-linear method as fast as the linear method in the unit disk geometry.