A new method for classifying colon cancer patients and healthy people from FTIR signals using wavelet transform and machine learning techniques

Abstract

Fourier Transform Infrared (FTIR) spectroscopy is used in studies to determine cancer from blood samples due to its ability to detect chemical changes. The major challenge in studies distinguishing between the FTIR signal of patients and that of healthy people is the lack of clear spectral difference in the FTIR signals. In previous studies, blood samples were dried to overcome this difficulty and peak values or ratios were used in the signal obtained by FTIR measurement. In the proposed method, unlike the literature, plasma samples were measured in liquid form and the resulting FTIR signal then examined as a whole. The FTIR signal was decomposed into sub-bands using wavelet transform. Colon cancer patients and healthy subjects were classified by using the features extracted from the sub-bands. Artificial Neural Networks (ANN), Support Vector Machines (SVM) and k-Nearest Neighbors (k-NN) were used for classification. Colon cancer patients and healthy subjects were classified with an accuracy of 97.14% with SVM. Experimental results indicate that the proposed method may be useful in distinguishing between colon cancer patients and healthy individuals.