Document Type : Original article
Authors
Department of Mechanical Engineering, Orthopaedic-BioMechanic Research Group, Faculty of Engineering, University of Birjand, Birjand, Iran.
10.32598/SJRM.14.5.3386
Abstract
Background and Aims Knee osteoarthritis (OA) is a prevalent chronic condition in the elderly, primarily caused by degeneration of articular cartilage. It leads to pain and restricted mobility, and is often diagnosed at an advanced stage. Risk factors include obesity, genetic predisposition, and skeletal deformities such as valgus and varus knees. This study aimed to evaluate the potential of artificial intelligence (AI) and machine learning (ML) techniques in the statistical analysis and prediction of knee OA severity using biomechanical and clinical data.
Methods Based on a follow-up study, radiographic knee joint images were analyzed using AI-based methods to extract bone boundaries. Key biomechanical parameters, including joint space width (JSW), femoral-tibial angle (FTA), and joint line convergence angle (JLCA), as well as clinical variables, such as weight, height, body mass index (BMI), age, and gender, were measured. ML models, including K-nearest neighbors (KNN), artificial neural networks (ANN), binary and multiclass CatBoost, and random forest classifiers, were used in statistical analysis to determine the relationship between these variables and OA severity. Also, statistical regression and Pearson’s correlation coefficient were used to analyze the data.
Results The results showed that minimum JSW, FTA, JLCA, and weight significantly affected OA progression. The binary CatBoost model achieved 82% accuracy, and the multiclass CatBoost model demonstrated 60% accuracy with 73% sensitivity for identifying mild OA cases based on biomechanical parameters. Additionally, the random forest classification framework within CatBoost achieved an overall accuracy of 54% and showed better performance in analyzing clinical parameters compared to other models.
Conclusion ML models, particularly CatBoost and Random Forest, demonstrate promising performance in predicting and evaluating knee OA progression. These findings highlight the potential of ML techniques as supportive tools in clinical decision-making for the diagnosis and monitoring of OA.
Keywords
Main Subjects
)
