Tissue Discrimination from Impedance Spectroscopy as a Multi-objective Optimisation Problem with Weighted Naïve Bayes Classification

Electrical Impedance Spectroscopy with Multi-modal Tissue Discrimination for Ultrasound-Guided Targeted Biopsy

“…Using no original data in the validation set, DS1b, performs almost as well as DS1c and should also be considered. A further comparison can be made to [65] wherein only the original data is used, equivalent to DS1a. In that publication, the highest sensitivity obtained was 80.58%.…”

“…In earlier literature, there was an effort to model the medium under test using equivalent resistor-inductor-capacitor (RLC) circuits. They provide a means to model the electrical response of a known medium based on only a few parameters [65],…”

“…Table 2.1 summarizes the number of measurements for each of the tissue classes collected in each dataset. Dataset 1 shows the measurements taken using the probe presented in [65] and Dataset 2 shows the measurements taken using the probe designed for this thesis.…”

“…The other classifiers performed similarly with classification metrics all greater than 90%, with the exception of Naïve Bayes, which performed slightly worse overall with a macro-sensitivity of 77.13%, and the support vector machine, with a macro-sensitivity of 24.77%. As a comparison to the data collected using the previous sensorised probe from [65,1], the same classifiers used above were used on Dataset 1. To train the classifiers, 143 entries were used as the training set, and 100 were used as the test set, and 2 shows significant increases in classifier accuracy, even with such a small sub-set of measurements.…”