Early-stage cancer detection is challenging due to the lack of associated oral-tissue clinical features and absence of changes on conventional cellular-imaging, serological and histopathological exams. By using a molecular-sensitive optical technique such as Fourier-transform infrared (FT-IR) spectroscopy, disease-specific biochemical changes can be detected non-destructively, non-invasively and with small sample volumes. In this study, we have used FT-IR spectroscopy to analyze saliva samples of control, smoker, and occasional smoker groups in the fingerprint region (900cm-1 to 1800cm-1). Saliva-sample classification was performed with a neural network algorithm and leave-one-out validation. Correctly classified instances were 72.7% for the control group, 65.5% for occasional smokers and 75% for smokers.
|