Abstract: Knee osteoarthritis stands out as a prevalent form of arthritis, characterized by joint space reduction, osteophyte emergence, sclerosis, and bone distortion, which are observable through radiographs. Radiography stands as the benchmark and the most accessible and cost-effective modality. X-ray images are assessed based on the Kellgren and Lawrence grading system, which ranks osteoarthritis severity from normal to severe. Early identification is crucial for prompt intervention and slowing down knee osteoarthritis progression. Unfortunately, many current methods either combine or exclude complex grades to enhance model performance. This study aims to automatically detect and categorize knee osteoarthritis in line with the KL grading system for radiographs. We propose an automated deep learning-based ordinal classification approach for early detection and grading of knee OA using a single posteroanterior standing knee x-ray image.

Keywords: Knee osteoarthritis, Radiography, Deep learning, Ordinal classification, Transfer learning, Ensemble model, Kellgren and Lawrence grading system

| DOI: 10.17148/IARJSET.2024.11489