The potential of terahertz imaging for improving the efficiency of Mohs's micrographic surgery in terms of tumor margin detection was previously studied. Thanks to high water content of human skin, its dielectric response to terahertz radiation can be described by the double Debye model which uses five parameters to fit experimental data. Skin tumors typically have a higher water content than normal tissues do, and this should be apparent in the parameters. The goal of this paper is to apply statistical methods to these parameters to test their power to differentiate skin cancer from normal tissue. Based on the prediction accuracy estimated using a cross-validation method, we found the best classifier was the static permittivity at low frequency (εs). By combining the most relevant parameters, we obtained a classification accuracy of 95.7%, confirming the classification capability of the parameters, thereby supporting their application to improve terahertz imaging for the purpose of skin cancer delineation.
|Number of pages||9|
|Journal||IEEE Transactions on Terahertz Science and Technology|
|Early online date||26 Oct 2015|
|Publication status||Published - 1 Nov 2015|