FT-IR Spectra Analysis toward Cancer Detection

Computer-based research on cancer detection utilizes various types of data including MRI images, proteins and DNA. In this work, we investigate Fourier-transform infrared (FT-IR) spectra obtained from colon tissues as an alternative method to detect cancer. Fourier-transform infrared (FT-IR) spectra reflect the chemical composition of a source medium and hence provide utility for a wide range of applications including biology and medical science. They are characterized by high dimensionality and being prone to noise. In this paper, we present a study of machine learning algorithms to investigate the validity of FT-IR spectra for cancer detection. The Back-propagation neural network, the Decision tree, the Adaboost with decision tree, and the Support Vector Machine (SVM) algorithms were applied to 894 spectra in supervised learning. Of these algorithms, SVM held the most promise in binary classification. This result demonstrated the potential of FT-IR spectra in cancer research and the potential of SVM in spectra analysis.