Interpretation of KOA by KL Grading System using Deep Learning

Knee osteoarthritis poses a significant global health challenge, often diagnosed using conventional radiographic grading systems such as the Kellgren-Lawrence scale, which primarily rely on X-ray images and may result in delayed diagnosis. Recent efforts to improve diagnostic accuracy and efficiency have turned to deep learning methods, particularly convolutional neural networks (CNN). In this study, we examined eight adaptive neural network models that utilize CNN for knee osteoarthritis detection. These models were trained and validated on a substantial dataset of knee $\mathbf{X}$-rays and underwent thorough evaluation to assess their capability in classifying knee osteoarthritis severity. Our results demonstrate that the top-performing model achieved an impressive accuracy of 98.73 %, outperforming its counterparts. This research highlights the potential of deep learning models, specifically CNN, in enhancing the precision and speed of knee osteoarthritis diagnosis.