Distinguishability in EIT using a hypothesis-testing model

Andy Adler1, Pascal Gaggero2 and Yasheng Maimaitijiang1 1Carleton University, Ottawa, Canada 2Centre Suisse d’Electronique et de Microtechnique, Landquart, Switzerland Abstract: In this paper we propose a novel formulation for the distinguishability of conductivity targets in electrical impedance tomography (EIT). It is formulated in terms of a classic hypothesis test to make it directly applicable to experimental configurations. We test to distinguish conductivity distributions σ2 from σ1, from which EIT measurements are obtained with added white Gaussian noise with covariance Σn. In order to distinguish the distributions, we must reject the null hypothesis H0: x̂ = 0, which has a probability based on the z-score: z = x̄ σx . This result shows that distinguishability is a product of the impedance change amplitude, the measurement strategy and the inverse of the noise amplitude. This approach is used to explore different current stimulation strategies.