Investigation of infrared Fourier-Transform spectroscopy for oral cancer detection

Abstract

A system was developed using a Fourier Transform-n spectrometer to investigate spectral differences between malignant, benign and healthy oral tissue in the near-infrared range (2.0-2.5 microns). A hamster model for oral squamous cell carcinoma and one for benign lesions in soft oral tissue (i.e. inflammation) was used. After tissue transform-nation in the malignant and benign cases and when no transformation occurred (i.e. healthy), the animals were euthanized and the malignant, inflamed and healthy tissues were excised. Infrared absorption spectra of the buccal mucosa were then collected on all three models, in vitro. A total of 160 near-infrared (NIR) scans were taken, 70 on malignant tissue, 20 on benign, inflamed, tissue and 70 on healthy tissue. Multiplicative signal correction (MSC), used during preprocessing, together with principal component analysis (PCA) showed a 90% sensitivity, 87% specificity and a false negative rate of . 1 0 between malignant and healthy/benign tissue types across animals using this wavelength range. The eigenvectors from the PCA analysis indicate that differences were detectable in the 2.25-2.35 [mm range. Absorption spectra of different ratio mixtures of DNA/RNA and protein were studied to help identify which absorption bands were in this range. Small amounts of RNA were found to change the near-infrared absorption spectra significantly, when added to a mixture of DNA/protein. This leads to the conclusion that the concentration of RNA in the cell's nucleus may be a significant spectral marker in identifying malignant cells. Correlating the mixture's near-IR spectra with published information, the absorption bands in this range were found to be attributed to the N-H stretching, C=O stretching vibration, and C-H deformation vibrations. These molecular components can be found in the nucleotides of RNA and in the ribose sugar molecule.