Abstract

Osteoarthritis (OA), a degenerative joint disease presenting as loss of cartilage, is a leading cause of disability worldwide, increasingly with aging populations. Early detection is crucial for effective treatment since there is no definitive cure, yet, current assessment techniques fall short and rely on ionising radiation or invasive procedures. We report an application of multimodal "spectromics", low-level abstraction data fusion of non-destructive NIR Raman and NIR-SWIR absorption spectroscopy, providing an enhanced, interpretable "fingerprint" for diagnosis of OA in human cartilage. Under multivariate statistical analyses and supervised machine learning, cartilage was classified with high precision and disease state predicted accurately. Discriminatory features within the spectromics fingerprint elucidated clinically relevant tissue components (OA biomarkers). Further, we have developed an automated goniometric 3D hyperspectral mapping setup, and characterised OA cartilage on whole human femoral heads post hip arthroplasty for spatially-resolved spectromics. These results lay foundation for minimally invasive, deeply penetrating, label-free, chemometric diagnosis of the hip.