Identification Of Articular Cartilage And Osteoarthritis By Fourier Transform Infrared Imaging With Chemometrics

Articular cartilage(AC)is an important composition of joints,which covers the articular surface for reducing friction,bearing pressure and buffering vibration during joint motion and loading.The primary molecular components of AC matrix are type II collagen and proteoglycan.Age,obesity,injury and other factors could cause the degeneration or damage of AC,which might lead to the osteoarthritis(OA)gradually.In the early stage of OA,the major characteristics represent as the concentration reduction of the principal components in cartilage matrix.However,there is no macroscopic tissue injury in this stage,which causes a difficulty in clinical diagnosis.Fourier transform infrared spectroscopic imaging(FTIRI)technique can synchronously obtain the infrared spectrum and the image of the cartilage sample with fine spatial and spectral resolutions.The chemometric methods can extract the characteristic information from the spectra effectively,which could be used in quantitative and qualitative analysis.FTIRI technique combined with chemometric methods were applied to effectively identify the healthy and OA articular cartilage samples to explore a newly potential and effective method for distinguishing healthy and early OA cartilages.In this study,principal component analysis(PCA),Fisher discriminant analysis(FDA),partial least squares discriminant analysis(PLS-DA)and support vector machine discriminant analysis(SVM-DA)were applied to construct the discriminant model for classifying the healthy and OA cartilage samples.(1)Based on the preprocessed spectra,PLS-DA was used to distinguish the healthy and 2-year OA samples with the accuracy of 96.92%.(2)PCA combined with FDA were used to identify the spectra of AC samples from 8-week OA cartilage with the accuracy of 89.23% and 2-year OA cartilage with the accuracy of 92.31%,respectively,which is based on the origin spectra.(3)SVM-DA was used to achieve the multiple discriminant analysis of the AC,8-week OA and 2-year OA samples with the accuracy of 90.33% and the classification of the healthy and 2-year OA samples with the accuracy of 97.7%.The results indicated that the healthy and OA samples can be classified effectively by the above models.By comparing the results of three discriminant models,SVM-DA has the best ability to identify the healthy samples from OA and to achieve the multiple discriminant analysis.It might be a promising approach for the diagnosis of the early OA,and could provide theoretical basis and data support for the further research.