Hard and soft tissue characterization with microwave dielectric spectroscopy

Abstract

The dielectric property discrepancy between hard and soft tissues at microwave frequencies can potentially be utilized for the separation of those tissues from each other. Microwave dielectric properties of biological tissues are traditionally measured with the open-ended coaxial probe technique. However, the technique suffers from high error rates thanks to tissue heterogeneity, user errors, mathematical approach, and calibration degradation. It is known that datasets with different values will be classified with high accuracy when a machine learning algorithm is applied. Therefore, choosing a classification parameter that might be least plagued by inherent errors is critical for increasing the accuracy of tissue categorization. Empirically, dielectric properties at microwave frequencies abide by the power law. Supported this fact, one unexplored parameter is the power parameter which might be derived from the dielectric properties. This work presents investigations on the potential use of the power parameter to separate different tissues, spesifıcally hard and soft tissues, supported by the datasets within the literature. Additionally, to research the effectiveness of the power parameter, classification was performed with machine learning algorithms using the power parameters obtaıned from dielectric property measurements of healthy and malignant liver tissues. Through the appliance of the technique 82% accuracy was obtained. Towards this goal, it's predicted that the power parameter might be used as a feature containing different information additionally to dielectric properties in tissue classification. Alternatively, in some cases dielectric properties do not provide enough information, one example is that the separation of hard and soft tissues, under such conditions the power parameter might be employed for classification purposes. This approach might be used as an alternative method for rapid diagnostic to high-cost imaging or mutation screening tests. The frequency-dependent dielectric properties of the biological tissues are crucial to developing diagnostic technologies.