Near infrared spectroscopy exhibits a tremendous potential for clinical chemistry and tissue pathology. Owing to its penetration depth into human skin, near infrared radiation can probe chemical and structural information non-invasively. Metabolic diseases such as diabetes mellitus increase nonenzymatic glycation with the effect of glucose molecules bonding chemically to proteins. In addition, glycation accumulates on tissue proteins with the clearest evidence found in extracellular skin collagen, affecting also covalent crosslinking between adjacent protein strands, which reduces their
flexibility, elasticity, and functionality. Non-enzymatically glycated proteins in human skin and following chemical and
structural skin changes were our spectroscopic target. We carried out measurements on 109 subjects using two different NIR-spectrometers equipped with diffuse reflection accessories. Spectra of different skin regions (finger and hand/forearm skin) were recorded for comparison with clinical blood analysis data and further patient information allowing classification into diabetics and non-diabetics. Multivariate analysis techniques for supervised classification such as linear discriminant analysis (LDA) were applied using broad spectral interval data or a number of optimally selected wavelengths. Based on fingertip skin spectra recorded by fiber-optics, it was possible to classify diabetics and non-diabetics with a maximum accuracy of 87.8 % using leave-5-out cross-validation (sensitivity of 87.5. %, specificity of 88.2 %). With the results of this study, it can be concluded that ageing and glycation at elevated levels cannot always be separated from each other.
|