Detection of Knee Osteoarthritis Using the Xception Model and Voting Classifier

The work proposed here deals with the diagnosis of knee osteoarthritis by using X-ray images. Since they are the cheapest and most commonly used for evaluating the health of the knee joint, X-ray images are preferred more often. Currently available image processing techniques lack in terms of precision and sensitivity. For that purpose, a novel method has been developed, which is a custom implementation of Deep learning model coupled with a voting classifier to identify minute features. In this model, we have explored various models that's been customized specially for image classification. This would improve robustness in feature detection.The aim of the proposed model is the accurate detection and grading of the severity of knee osteoarthritis based on the Kellgren and Lawrence system. It is achieved using a combination of advanced classification models, such as Xception and InceptionV3, and efficient object detection frameworks, like YOLOv5 and YOLOv8. A user-friendly interface is developed using Flask to ensure smooth and reliable test conditions while interacting with the diagnostic system. Key concepts used in this study include deep learning, multi-scale features, machine learning, and human-computer interaction.