Statistical estimation of interface-boundary shapes from electrical impedance tomographic measurements

Brian A. Cattle, Robert M. West and Robert G. Aykroyd

Using a recently-developed boundary element approach for electrical impedance tomography, the identification of boundary shapes between materials of differing conductivity is considered. The examples considered are motivated by the desire to locate damages in filter and catalyst beds in industrial processes, but they are also applicable in many application areas where the unknown location of a boundary is sought. The approach to reconstruction of the boundary location is statistical and is based upon Markov chain Monte Carlo techniques. The degree of reliability (accuracy) of the reconstructed estimates is quantified through credibility intervals.

Keywords: electrical impedance tomography; boundary-element method; Bayesian statistics; prior elicitation; boundary estimation