Feasibility of Near-Infrared Spectroscopy for Identification of L-Fucose and L-Proline—Towards Detecting Cancer Biomarkers from Saliva

Abstract

Near-infrared spectroscopy (NIRS) is a non-ionizing optical technique that can be used to quantify proteins, carbohydrates, fats, and other organic and biological substances. The aim of this study was to determine the ability of NIRS to identify different concentrations of L-fucose and L-proline solutions by utilizing different NIR spectral regions. NIR spectra of solid L-fucose and L-proline, their aqueous solutions in different concentrations, and the spectra of saliva samples collected from two patients with oral squamous cell carcinoma (OSCC) were studied. Differences in spectra of the pure solid reference samples and water were most noticeable in spectral regions 800–1250 nm and 1418–1867 nm. The saliva sample with an atypically high concentration of oral cancer biomarkers showed a similar spectral feature between 1530–1650 nm as the liquid samples with cancer biomarkers. In addition, a fine k-nearest neighbors (kNN) classifier was trained to differentiate the aqueous solutions and achieved 75.97% validation accuracy. The preliminary study presents that NIRS can be utilized to detect differences in spectra between the different biomarker concentrations in aqueous solutions. However, the qualitative measures may have resulted in limited sensitivity, which could be enhanced by additional samples and using a measurement probe dedicated to fluid measurements.

Identification of gallbladder cancer by direct near-infrared measurement of raw bile combined with two-trace two-dimensional correlation analysis

“…This previously reported NIR bile analysis scheme has potential for clinical on-site analysis due to the simple and rapid analysis that requires no pretreatment of raw bile. Other than the analysis of bile, NIR spectroscopy has been extensively adopted for biomedical diagnosis 2,3 using relevant biological specimens including blood, 4,5 serum, [6][7][8] saliva, 9 and tissue, 10,11 and diverse chemometric and deep learning methods such as partial least squares (PLS), [5][6][7][8]12 linear discriminant analysis (LDA), 3,11,14 support vector machine (SVM), 13 and artificial neural network (ANN) 13 are combined for multivariate analysis of acquired NIR spectra.…”