Knee Osteoarthritis Detection on X-rays: A Hybrid Approach Leveraging Deep Learning

Knee osteoarthritis (OA) is a common degenerative joint disease marked by pain, stiffness, and restricted mobility due to cartilage deterioration in the knee joint. This paper presents a comprehensive study on the application of deep learning techniques for the detection and classification of knee osteoarthritis using X-ray images. Leveraging a dataset of radiographic images, annotated with five distinct grades of OA severity (0 to 4), from healthy to severe conditions, the dataset is augmented to enhance diversity. Key preprocessing steps, including resizing, rescaling, Contrast Limited Adaptive Histogram Equalization (CLAHE), and data augmentation, are applied to optimize image quality and improve model robustness. The dataset is divided into a 70:30 ratio for training and testing, allowing for rigorous evaluation of multiple state-of-the-art deep learning architectures, such as CNN, VGG16, VGG19, ResNet50, ResNet152, EfficientNetB0, MobileNetV3 Large, and hybrid models combining ResNet50 with VGG16 and VGG19. The results demonstrate high classification accuracy, with models like MobileNetV3 Large achieving 98.80%, ResNet152 at 98.73%, and hybrid models ResNet50+VGG19 reaching 98.83%. This research highlights the efficacy of deep learning and hybrid models in delivering accurate, automated knee osteoarthritis detection, emphasizing their potential in enhancing clinical decision-making and patient care.