A neural network reconstruction algorithm for Intraventricular Impedance Imaging

Abstract

An Intraventricular Impedance Imaging (III) system, that will be used for assessing electrical and mechanical cardiac properties via an intraventricular catheter, is presently under development. One of the major problems to be solved is the determination of the intraventricular catheter position within the ventricle. Existing methods for determining catheter position within a cardiac ventricle, including X-ray, fluoroscopy and computer tomography, are accurate but cumbersome, expensive, and unable to ascertain the continuous real-time intraventricular catheter position. The purpose of this work was to develop a reconstruction algorithm, based on Artificial Neural Networks (ANN), which will be used to process the electrical information from the catheter to ascertain the continuous, real-time intraventricular catheter position. A back-propagation neural network was trained using results from computer simulations of the III system. The neural network predicted the desired output variables with errors ranging from 0.05% to 1.8% and correlation coefficients (r) ranging from 83% to 99% The RMS error of the output variables was 4.9% These results indicate that ANN have great potential as a tool in determining the continuous real-time intraventricular catheter position.