Diffuse reflectance spectroscopy and Raman spectroscopy for label-free molecular characterization and automated detection of human cartilage and subchondral bone

Abstract

Optical technologies hold potential for bio-medical research as well as the clinical routine, since they allow contact-free, non-destructive access to several clinically relevant features of tissues. In this study, diffuse reflectance spectroscopy and spontaneous Raman spectroscopy have been applied to human hip samples in order to characterize differences between cartilage and adjacent subchondral bone tissue. Mathematical decomposition of absorption and scattering properties is an important aspect in diffuse reflectance data analysis. Here, we present the novel application of Bayesian statistics to this analysis, which can extract probability density spectra of the absorption coefficient and the scattering coefficient. Using this concept of a Bayesian decomposition algorithm for the spectral decomposition, it was possible to find characteristic differences in the relative concentration of melanin and haemoglobin as well as in scattering properties. Furthermore, this allowed a reconstruction of probability density plots for scattering coefficient and absorption coefficient instead of singular values. Complemented by the results of Raman spectroscopy, it was further possible to detect significant differences in the ratio of minerals (such as hydroxyl apatite) to bio-molecules (such as proteins) when comparing bony tissue regions to the surrounding cartilage. Finally, it was possible to use these molecular contrast mechanisms for highly-accurate, automated tissue classification using machine learning algorithms on the level of individual patients.