Determination of collagen and proteoglycan concentration in osteoarthritic and healthy articular cartilage by Fourier transform infrared imaging and partial least square

Abstract

Fourier transform infrared imaging (FTIRI) combined with chemometrics has potential to determine the molecular and chemical properties of biological tissues. In this study, FTIRI with partial least square (PLS) algorithm was used to quantitatively study the concentration distributions of two principal components (collagen and proteoglycan) in healthy and osteoarthritis (OA) articular cartilage. 10μm-thick sections of canine tibial cartilage were imaged at 6.25μm/pixel. The spectra extracted from the infrared images were imported into the PLS model to predict the concentrations of collagen and proteoglycan in the corresponding cartilage sections. Spectral pre-processes in this model included multiplicative scatter correction, normalization, and baseline correction, in order to reduce negative factors in the spectra. Leave-one-out cross validation was also performed in this model. The obtained coefficients of Pearson’s r (0.964) and root mean square error of calibration sample (RMSEC, 5.4%) suggest the excellent representativeness of the PLS model. The prediction results indicate that proteoglycan concentration is lower than that of collagen in both healthy and OA cartilage, and decreases in the OA cartilage, especially at superficial zone. The prediction with spatial resolution can help to understand the osteoarthritic development and demonstrate that FTIRI combined with PLS is a powerful analytical approach in biomedical research.