Terahertz reflection imaging (at frequencies ~0.1–10 THz/1012 Hz) is non-ionizing and has potential as a medical imaging technique; however, there is currently no consensus on the optimum imaging parameters to use and the procedure for data analysis. This may be holding back the progress of the technique. This article describes the use of various intelligent analysis methods to choose relevant imaging parameters and optimize the processing of terahertz data in the diagnosis of ex vivo colon cancer samples. Decision trees were used to find important parameters, and neural networks and support vector machines were used to classify the terahertz data as indicating normal or abnormal samples. This work reanalyzes the data described in Reid et al. (2011) (Physics in Medicine and Biology, 56, 4333–4353), and improves on their reported diagnostic accuracy, finding sensitivities of 90–100% and specificities of 86–90%. This optimization of the analysis of terahertz data allows certain recommendations to be suggested concerning terahertz reflection imaging of colon cancer samples.
|Number of pages||8|
|Journal||Expert Systems with Applications|
|Publication status||Published - May 2013|
- neural networks
- support vector machines
- decision tree
- colon cancer