Multivariate analysis of Brillouin imaging data by supervised and unsupervised learning

Brillouin-Raman micro-spectroscopy and machine learning techniques to classify osteoarthritic lesions in the human articular cartilage

In this study, Brillouin and Raman micro-Spectroscopy (BRamS) and Machine Learning were used to set-up a new diagnostic tool for Osteoarthritis (OA), potentially extendible to other musculoskeletal diseases. OA is a degenerative pathology, causing the onset of chronic pain due to cartilage disruption. Despite this, it is often diagnosed late and the radiological assessment during the routine examination may fail to recognize the threshold beyond which pharmacological treatment is no longer sufficient and prosthetic replacement is required. Here, femoral head resections of OA-affected patients were analyzed by BRamS, looking for distinctive mechanical and chemical markers of the progressive degeneration degree, and the result was compared to standard assignment via histological staining. The procedure was optimized for diagnostic prediction by using a machine learning algorithm and reducing the time required for measurements, paving the way for possible future in vivo characterization of the articular surface through endoscopic probes during arthroscopy.