Evaluation of a Dielectric Inclusion Using Inductive RF Antennas and Artificial Neural Networks for Tissue Diagnosis

In this paper, the relevance of the non contact RF evaluation of the complex permittivity of organic material is addressed by means of a computational approach. The authors consider a simple configuration constituted of a single loop RF antenna interacting with a dielectric material mimicking a typical organic tissue (the electrical conductivity is 0.6 S/m and the dielectric constant is 80) which includes a buried inclusion (e.g. a tumor featuring a conductivity of 0.2 to 1.6 S/m and a dielectric constant ranging from 20 to 160). First a three dimensional semi analytical model (DPSM) is implemented in order to evaluate the sensitivity of such an antenna to the complex permittivity of the buried inclusion. Then, the inverse problem which consists in evaluating the complex permittivity, the size and the location of the inclusion is addressed by means of an artificial neural network (ANN) approach. For the considered configuration (5 mm radius antenna, 40 mm radius spherical inclusion buried at a 5 to 20 mm depth within the tissue, antenna operated at 135 MHz), the main conclusions are that the complex permittivity and the depth of the inclusion can be fairly estimated (estimation error smaller than 5%), even in the case of antenna positioning uncertainties, providing the ANN is adequately trained. Also, a double antenna configuration significantly enhances the estimation of the location and size of the inclusion.