Published Paper Details:

Knee osteoarthritis (KOA) is a degenerative joint condition that affects millions globally, especially older adults. Timely and accurate diagnosis is essential to slow disease progression. This paper presents a deep learning-based system for automated KOA detection using X-ray images, graded according to the Kellgren and Lawrence (KL) scale. Four convolutional neural networks--ResNet-34, VGG-19, DenseNet-121, and DenseNet-161--are fine-tuned through transfer learning and combined using an ensemble strategy. To model the ordered nature of KOA severity, Conditional Ordinal Regression (CORN) is employed. The system integrates Explainable AI (XAI) using Eigen-CAM visualizations to highlight diagnostic regions in the X-ray images. Evaluation on the Osteoarthritis Initiative dataset shows state-of-the-art results, with 98% accuracy and a Quadratic Weighted Kappa (QWK) score of 0.99. The final model is deployed via a Streamlit web application, offering an accessible interface for real-time diagnosis. The approach provides a reliable and interpretable tool for assisting radiologists in KOA assessment.