OBJECTIVETo investigate the Raman spectral characteristics of oral squamous cell carcinoma, high-grade epithelial dysplasia and normal mucosa.

METHODSFifty- six fresh samples of oral carcinoma, 50 of high-grade epithelial dysplasia and 32 of normal mucosa were collected. The i-Raman spectrometer with an optical fiber tube was applied to acquire Raman spectrum. The diagnostic model established by principle component analysis (PCA) and discriminant function analysis (DFA) was used to analyze and classify the spectra of different samples.

RESULTSThere were significant differences among the Raman spectra of these samples. Compared with the spectra of normal mucosa, the spectra of oral carcinoma and dysplasia showed strong peaks which were contributed to nucleic acids, proteins and lipids. The diagnostic models established by PCA-DFA could successfully classify these Raman spectra of different samples with a high accuracy of 96.4% (133/138). The model was evaluated by 'Leave one out' cross-validation and reached a high accuracy of 92.8% (128/138).

CONCLUSIONSThe proliferation and metabolism of oral squamous cell carcinoma and epithelial high-grade dysplasia are more active than normal mucosa. The diagnostic models established by PCA-DFA can classify these Raman spectra of different samples with a high accuracy.