We report here a study on human saliva tested as a diagnostic medium for oral cancer detection on three groups: oral squamous cell carcinoma (OSCC), oral sub mucous fibrosis (OSMF; precancer), and healthy controls (normal). Measurements have been conducted using fluorescence spectroscopy with 350 nm excitation and Stokes shift (SS) spectroscopy (SSS) with 120 nm shift from a total of 99 saliva samples. For classification, principal component analysis (PCA) and linear discriminant analysis (LDA) have been applied on the data sets. Linear discriminant (LD) scores of fluorescence spectra are able to differentiate OSCC to normal, OSMF to normal and OSCC to OSMF with sensitivities 91%, 92%, 91% and specificities 97%, 100%, 94% respectively, while LD scores of SS spectra differentiate respective groups with sensitivities 100%, 94%, 94% and specificities 97%, 100%, 94%. Cross-validation on the datasets of PC scores during LDA illustrates that sensitivity and specificity of SSS data are less affected than those of fluorescence data. Saliva is thus seen as a potential non-invasive and simple diagnostic medium, with SS spectroscopy as a better diagnostic tool for oral precancer.
Keywords: Fluorescence spectroscopy; Oral cancer; Principal component and linear discriminant analysis; Saliva; Stokes shift spectroscopy.