SVM for prostate cancer using electrical impedance measurements.

Biopsies are currently the 'gold standard' method for identifying cancer of the prostate. While biopsies yield very accurate information regarding the area they sample, they are performed at discrete points and provide no information on the adjacent tissue. To enhance procedural accuracy, biopsies of a large number of sites are routinely carried out. Although more accurate, this method is both more complex and nevertheless remains discrete. In this paper, we evaluate the advantages of using bio-impedance information as the input for a support vector machines (SVMs) classifier to overcome these limitations. In this method, the biopsy probes are used as electrodes to obtain electrical impedance data during each biopsy sample. Using a computer model of the prostate, a SVM was trained and tested. Different tumor shapes and conductivity values, and the classifier's ability to generalize to these different properties, were examined. We demonstrate that by using this classifier the number of biopsies can be reduced and valuable information concerning the adjacent tissue which was not biopsied can be generated.