Interval Simulated Annealing applied to Electrical Impedance Tomography image reconstruction with fast objective function evaluation

Electrical Impedance Tomography (EIT) is a imaging technique that attempts to reconstruct the impedance distribution inside an object from the impedance between electrodes placed at the object’s surface. The EIT reconstruction problem can be approached as a non-linear non-convex optimization problem where one tries to maximize the matching between a simulated impedance problem and the observed data. This non-linear optimization problem is often ill-posed, and not very suited to methods that evaluate derivatives of the objective function. It may be approached by Simulated Annealing (SA), but at a large computational cost due to the expensive evaluation process of the objective function, that involves a full simulation of the impedance problem at each iteration. We propose here a variation of SA where the objective function is evaluated only partially, while ensuring boundaries on the behavior of the modified algorithm.