Predicting Severity of Knee Osteoarthritis Using Bimodal Data and Machine Learning

Knee Osteoarthritis (KOA) is a prevalent degenerative disease, typically assessed via X-ray imaging. This study aims to explore and validate the potential of a bimodal dataset combining infrared thermographic imaging with patient health data for predicting the severity of KOA. Initially, the study involved preprocessing of the infrared thermographic data, including background elimination and extraction of regions of interest (ROI), from which key features such as temperature, texture, and structure were derived. These features were then integrated with patient health data. Considering the issue of class imbalance in the dataset, the Synthetic Minority Over-sampling Technique (SMOTE) was employed to augment the minority class samples. Subsequently, the efficacy of this method was validated using various machine learning classification models, among which the decision tree model demonstrated superior performance. The findings of this study not only confirm the effectiveness of infrared thermographic imaging in diagnosing the severity of KOA but also provide a novel auxiliary tool for the clinical diagnosis of knee osteoarthritis.