Search

Mid-infrared spectroscopy (MIR), near-infrared spectroscopy (NIR), and Raman spectroscopy are all well-established analytical techniques in biomedical applications. Since they provide complementary chemical information, we aimed to determine whether combining them amplifies their strengths and mitigates their weaknesses. This study investigates the feasibility of the fusion of MIR, NIR, and Raman spectroscopic data for characterising articular cartilage integrity. Osteochondral specimens from bovine patellae were subjected to mechanical and enzymatic damage, and then MIR, NIR, and Raman data were acquired from the damaged and control specimens. We assessed the capacity of individual spectroscopic methods to classify the samples into damage or control groups using Partial Least Squares Discriminant Analysis (PLS-DA). Multi-block PLS-DA was carried out to assess the potential of data fusion by combining the dataset by applying two-block (MIR and NIR, MIR and Raman, NIR and Raman) and three-block approaches (MIR, NIR, and Raman). The results of the one-block models show a higher classification accuracy for NIR (93%) and MIR (92%) than for Raman (76%) spectroscopy. In contrast, we observed the highest classification efficiency of 94% and 93% for the two-block (MIR and NIR) and three-block models, respectively. The detailed correlative analysis of the spectral features contributing to the discrimination in the three-block models adds considerably more insight into the molecular origin of cartilage damage.

Preclassification of raw infrared spectra has often been neglected in scientific literature. Separating spectra of low spectral quality, due to low signal-to-noise ratio, presence of artifacts, and low analyte presence, is crucial for accurate model development. Furthermore, it is very important for sparse data, where it becomes challenging to visually inspect spectra of different natures. Hence, a preclassification approach to separate infrared spectra for sparse data is needed. In this study, we propose a preclassification approach based on Multiplicative Signal Correction (MSC). The MSC approach was applied on human and the bovine knee cartilage broadband Fourier Transform Infrared (FTIR) spectra and on a sparse data subset comprising of only seven wavelengths. The goal of the preclassification was to separate spectra with analyte-rich signals (i.e., cartilage) from spectra with analyte-poor (and high-matrix) signals (i.e., water). The human datasets 1 and 2 contained 814 and 815 spectra, while the bovine dataset contained 396 spectra. A pure water spectrum was used as a reference spectrum in the MSC approach. A threshold for the root mean square error (RMSE) was used to separate cartilage from water spectra for broadband and the sparse spectral data. Additionally, standard noise-to-ratio and principle component analysis were applied on broadband spectra. The fully automated MSC preclassification approach, using water as reference spectrum, performed as well as the manual visual inspection. Moreover, it enabled not only separation of cartilage from water spectra in broadband spectral datasets, but also in sparse datasets where manual visual inspection cannot be applied.

The aim of the study was to optimize preprocessing of sparse infrared spectral data. The sparse data were obtained by reducing broadband Fourier transform infrared attenuated total reflectance spectra of bovine and human cartilage, as well as of simulated spectral data, comprising several thousand spectral variables into datasets comprising only seven spectral variables. Different preprocessing approaches were compared, including simple baseline correction and normalization procedures, and model-based preprocessing, such as multiplicative signal correction (MSC). The optimal preprocessing was selected based on the quality of classification models established by partial least squares discriminant analysis for discriminating healthy and damaged cartilage samples. The best results for the sparse data were obtained by preprocessing using a baseline offset correction at 1800 cm⁻¹, followed by peak normalization at 850 cm⁻¹ and preprocessing by MSC.