Tendinopathy is a musculoskeletal disorder characterized by pathological extracellular matrix remodeling, yet the associated biochemical alterations remain incompletely understood. In this study, SR-FTIR microspectroscopy was employed to investigate compositional changes in a mouse model of Achilles tendinopathy. A total of 1330 spectra were collected from normal and diseased tendon tissues across the wavenumber region 4000-600 cm-1. We used the relative content of each functional group as input features to construct eight classic machine learning models for tendinopathy prediction. Model evaluation results showed that Random Forest achieved the best classification performance. Further combined with SHAP analysis, we identified the CO of alcohol functional group as the feature contributing most significantly to the model classification. By integrating spectral analysis with tissue staining, we confirmed cholesterol as the molecular origin of this CO functional group and established cholesterol infiltration as a characteristic pathological feature of tendinopathy.
Keywords: Lipid infiltration; Machine learning; Model explanation; SR-FTIR; Tendinopathy.
Copyright © 2026. Published by Elsevier B.V.