Prediction of Knee Osteoarthritis Severity from X-Ray Images Using Ensemble Learning

This is the authors' abstract. We don't add key points for this paper.

Millions of people worldwide are affected by knee osteoarthritis (OA), a prevalent condition involving the deterioration of joint cartilage and underlying bone. Consequently, it causes discomfort, restricted movement, and diminished mobility. It is crucial to accurately and promptly assess the severity grading of the disease for effective clinical management. In this study, the important issue of predicting knee OA severity using X-ray images is addressed through the utilization of deep learning models, specifically convolutional neural networks (CNNs). An ensemble model that combines the strengths of different architectures was proposed to show the enhanced accuracy in Knee OA. The proposed work used the best features extracted using EfficientNet, DenseNet and ensemble models to attain an accuracy rate of 94.76%. The exceptional performance of the ensemble model underscores its reliability in classifying knee OA severity and provides a comprehensive and precise solution for health assessment. The proposed systematic approach demonstrates how deep learning can enhance diagnostic capabilities for knee osteoarthritis and improve clinical decision-making.