Machine Learning Classification of Articular Cartilage Integrity Using Near Infrared Spectroscopy

Abstract

Introduction-Assessment of cartilage integrity during arthroscopy is limited by the subjective visual nature of the technique. To address this shortcoming in diagnostic evaluation of articular cartilage, near infrared spectroscopy (NIRS) has been proposed. In this study, we evaluated the capacity of NIRS, combined with machine learning techniques, to classify cartilage integrity. Methods-Rabbit (n = 14) knee joints with artificial injury, induced via unilateral anterior cruciate ligament transection (ACLT), and the corresponding contra-lateral (CL) joints, including joints from separate non-operated control (CNTRL) animals (n = 8), were used. After sacrifice, NIR spectra (1000-2500 nm) were acquired from different anatomical locations of the joints (n TOTAL = 313: n CNTRL = 111, n CL = 97, n ACLT = 105). Machine and deep learning methods (support vector machines-SVM, logistic regression-LR, and deep neural networks-DNN) were then used to develop models for classifying the samples based solely on their NIR spectra. Results-The results show that the model based on SVM is optimal of distinguishing between ACLT and CNTRL samples (ROC_AUC = 0.93, kappa = 0.86), LR is capable of distinguishing between CL and CNTRL samples (RO-C_AUC = 0.91, kappa = 0.81), while DNN is optimal for discriminating between the different classes (multi-class classification, kappa = 0.48). Conclusion-We show that NIR spectroscopy, when combined with machine learning techniques, is capable of holistic assessment of cartilage integrity, with potential for accurately distinguishing between healthy and diseased cartilage.

Machine learning augmented near-infrared spectroscopy: In vivo follow-up of cartilage defects

“…However, specimens, spectral regions, number of samples, and statistical methods vary substantially between the studies. Both McGoverin et al and Prakash et al reported similar validation accuracy for in vitro measurements of human tissue with R 2 ¼ 64% and NRMSE ¼ 15.3%, and r ¼ 0.83 and NRMSE ¼ 14%, respectively, at spectral regions comparable to this study 18,39 . In addition, Prakash et al 19 reported the performance (r ¼ 0.52 and NRMSE ¼ 25%) of the independent arthroscopic ex vivo test group.…”

Classification, Categorization, and Algorithms for Articular Cartilage Defects

“…56,57 Taken together, best current evidence suggests that arthroscopic and MRI-based scoring and grading systems for articular cartilage and whole-joint pathology are useful and important clinical tools, but can only provide a cross-sectional status, or "snapshot," and do not allow for determination of disease phenotype or progression. [2][3][4][5][6][7][8]58 Because clinical data are not routinely available until patients seek care for symptoms, most classification and staging algorithms are primarily informed by late-stage disease, which leads to ambiguity, overlap, and generalization, relegating physicians and patients to incomplete, possibly inaccurate, data for decision making regarding type and timing of treatment. [59][60][61][62][63][64][65][66][67][68] As such, more nuanced, earlier, longitudinal analytical methods, ideally including controls for relevant cohorts, which incorporate articular cartilage lesions features, whole-joint status, and whole-patient variables are needed to fill this unmet need in orthopaedic health care.…”

Applications of Vibrational Spectroscopy for Analysis of Connective Tissues

“…One study showed the application of PLS-R using NIR spectra as a correlate to the swelling characteristics of cartilage based on the PG content across healthy and mechanically degraded cartilage [ 122 ]. In another study, NIR spectroscopy combined with machine learning techniques successfully distinguished healthy and diseased cartilage [ 123 ]. Another study used PLS modelling techniques to predict the thickness of the healthy and injured cartilage [ 124 , 125 ], and biomechanical, histological and biochemical properties of articular cartilage [ 125 ].…”