Objectively assessing visual analogue scale of knee osteoarthritis pain using thermal imaging

Abstract

Knee osteoarthritis (KOA) is a common degenerative joint disorder that significantly deteriorates the quality of life for affected patients, primarily through the symptom of knee pain. In this study, we developed a machine learning methodology that integrates infrared thermographic technology with health data to objectively evaluate the Visual Analogue Scale (VAS) scores for knee pain in patients suffering from KOA. We preprocessed thermographic data from two healthcare centers by removing background noise and extracting Regions of Interest (ROI), which allowed us to capture image features. These were then merged with patient health data to build a comprehensive feature set. We employed various regression models to predict the VAS scores. The results indicate that the XGBoost model, using a 7:3 training-to-testing ratio, outperformed other models across several evaluation metrics. This study confirms the practicality and effectiveness of using thermographic imaging and machine learning for assessing knee pain, providing a new supportive tool for the management of pain in KOA and potentially increasing the objectivity of clinical assessments. The research is primarily focused on the middle-aged and elderly populations. In the future, we plan to extend the use of this technology to monitor risk factors in children’s knees, with the goal of improving their long-term quality of life and enhancing the overall well-being of the population.