A Lightweight CNN Model for Knee Osteoarthritis Grading: Development and Evaluation Using Radiographic X-rays

Knee Osteoarthritis (KOA) is a progressive degenerative joint disease-causing pain, stiffness, and functional disability, posing a major health concern, especially among the elderly. Conventional diagnosis relies on radiographs and the Kellgren–Lawrence (KL) grading system, which suffers from subjectivity, inter-observer variability, and inefficiency for large-scale screening. To overcome these challenges, this research proposes an AI-driven multi-stage KOA detection framework using deep learning for automated severity classification. The model employs a dual-input Convolutional Neural Network (CNN) trained on preprocessed grayscale knee X-rays. A symmetry-based preprocessing technique pairs each X-ray with its horizontally flipped version, emphasizing asymmetries in joint space narrowing, osteophytes, and bone deformities, key KOA indicators. This improves feature capture and interpretability. The CNN classifies knees into five KL stages with high accuracy and robustness. For clinical use, the model is integrated into a Flask-based web application enabling X-ray upload, real-time severity prediction, and personalized PDF report generation. The platform also provides educational resources including treatment guidance, diet suggestions, and hospital references. Experimental results demonstrate competitive performance against state-of-the-art methods while offering real-time usability. By reducing subjectivity and ensuring consistent grading, this framework serves as a reliable clinical decision-support tool, with future work exploring explainability, larger datasets, and cloud deployment. General Terms Machine Learning, Deep Learning, Medical Imaging, Web-based Applications.