DEEP LEARNING-DRIVEN PREDICTION OF KNEE OSTEOARTHRITIS PROGRESSION USING X-RAY IMAGES: A VGG16-BASED APPROACH

Abstract— Osteoarthritis, or OA, is a widespread degenerative musculoskeletal condition affectingmillions of people worldwide, significantly lowering their quality of life, with most cases involvingthe knee joint. Characterized by loss of articular cartilage and its associated chronic pain with impairedknee functionality, knee OA becomes worsened by the process of degenerative changes occurring overtime. The mainstay of conventional diagnosis of clinical examination and plain radiography is oftencumbersome and not very reliable. To deal with these shortcomings, we propose a novel deep methodof learning that leverages the VGG16 model for predicting the development of knee OA from knee Xrayimages. Our approach tends to harness the immense image recognition potential of CNN to boostthe efficiency of diagnostics and increase its precision. A dataset of 3,300 knee X-ray images was usedto train the VGG16 model to classify different stages of OA severity with an average training andvalidation accuracy of 95% and 87%, respectively. This state-of-the-art approach speeds up thediagnosis, allows possibilities for early detection and planning for therapy, and may lead to improvethe results for patients. In this study, deep learning was used to design a complete framework forpredicting the course of knee OA. Training and testing sets were created from the dataset., composing80% of each for training and testing, respectively. VGG16 was selected based on its powerfularchitecture and its ability to capture complex features from images. The model was trained with thebatch size set to six, maintaining an accuracy and loss metric during the tracking of the training processitself. The results showed the excellent efficacy of our approach with average validation and trainingaccuracies of 87 and 95%, respectively. Ultimately, the trained model was saved for use in futureimplementations, thus making the system easy to use for clinical applications. Keywords— Knee Osteoarthritis, Deep Learning, VGG16, X-Ray Images, Disease Progression,Diagnostic Accuracy, Artificial Intelligence.