Tissue characterization with an electrical spectroscopy SVM classifier.

This feasibility study introduces the use of a classifier based on electrical spectroscopy measurements for breast cancer tissue characterization. The classifier is of the support vector machine type, and the vector of data is made of electrical voltage measurements at 12 discrete electrical excitation frequencies over the beta dispersion range of the analyzed tissue and at discrete locations selected from information produced by conventional medical imaging. The database was generated through a mathematical simulation model. The performance of the classifier was evaluated through a test of its ability to distinguish between simulations of malignant and benign tissues in the breast. The results demonstrate the feasibility of the concept and illustrate the tissue characterization ability of this classifier.