Optical monitoring of inflammatory response in joint explant cultures using near infrared spectroscopy.

OBJECTIVE: To evaluate near-infrared (NIR) spectroscopy combined with machine learning (ML) for detecting and estimating inflammatory and matrix-related biomarkers released into the culture medium of an explant model. MATERIALS AND METHODS: Bovine osteochondral explants and synovial tissues were co-cultured under inflammatory factor-treated (IF-treated) conditions that mimic osteoarthritis (OA) and non-IF-treated conditions. The co-cultures were maintained for seven days with the culture medium collected and replaced daily. The samples were subjected to NIR spectral measurements (900-1700 nm), followed by measurement of sulphated glycosaminoglycans (sGAG), nitric oxide (NO), interleukin (IL)-6, and IL-8 released into the culture. Acquired spectra were preprocessed and analyzed for classification and estimation of biomarkers using ML-based models. Feature importance analyses were performed to identify key wavelength features that contribute to classification and prediction of biomarkers. RESULTS: Biomarker release showed a substantial increase in the OA-like setup compared to the non-IF-treated setup. During classification, the IF-treated and non-IF-treated spectra showed distinct patterns, with noticeable differences in the 1440-1520 nm and 1600-1700 nm regions. Using these spectral variations, the best performing classifier achieved an accuracy of 95.3% in distinguishing OA-like from non-IF-treated samples. The regression analysis also showed good regression performance with R² values of 0.44 (sGAG), 0.48 (NO), 0.42 (IL-6), and 0.36 (IL-8). CONCLUSIONS: This study showed that NIR spectroscopy combined with ML can detect and quantify OA-related spectral signatures in explant media, thereby demonstrating its potential as a non-destructive and rapid method for monitoring inflammatory activity in ex vivo OA models.