Error Study of EIT Inverse Problem Solution Using Neural Networks

Abstract

Electrical Impedance Tomography (EIT) is a visualization of the internal electric conductivity of an object using measurements performed on its surfaces. As an Inverse problem, the solution can be approximated by means of Artificial Neural Networks. In this paper, an Artificial Neural Network solution to this Inverse Problem is presented. Based on the electrical voltage and current measurements on the boundary of the object, the conductivity distribution has been found and the resulting error is calculated. The error is compared for different Neural Network architectures to detect and minimize the errors caused by the solution method. Also, different Neural Networks were tested in the noisy and noiseless conditions to reach the suitable architecture for each case and investigate the measurement error and noise effects. Other than overall error of the whole circuit, distribution of error in different areas of the object is analyzed.