Classification of Multi-Layer Tissue-Mimicking Dielectric Stacks From 2 to 20 GHz

Abstract

Determination of the thickness, permittivity, and conductivity of tissue layers in the microwave region of the electromagnetic spectrum is relevant to a number of applications, such as breast-cancer imaging and non-invasive subcutaneous tissue thermometry. Many current characterization approaches are limited to one or two layers, often required to be aqueous. This paper presents simplified modeling of a stack of tissue layers as a series of complex impedance transmission lines in the 2–20 GHz decade. A near-field, broadband interrogation antenna designed for this frequency range and placed on the skin is validated with complex reflection coefficient measurements on seventeen different stacks of materials. Initial measurements are used to build a lookup table of features that are then used to classify three independent sets of follow-up measurements on the same stacks. After processing and consideration of very thin and very low loss materials, the error rates for classification are found to be between 5.9% and 14.7%. This confirms that features extracted from a simple, calibrated one-port broadband reflection coefficient measurement provide sufficient information to identify the composition of a layered stack, modeling tissue layers.