A Deep Learning Approach for Development of Web Application for Automated Knee OA Classification

Knee Osteoarthritis (OA), often referred to as "Degenerative Joint disorder," is one of the most prevalent forms of musculoskeletal disease. Surgery for joint replacement persists to be the only option for treating severe knee OA. Hence, an early identification of a disease enhances the likelihood that a person will receive treatment at the initial onset. However, it has been averted due to a lack of adequate medical infrastructure in the country’s suburbs and rural areas. The automatic grading of knee OA severity using Computer Aided Diagnosis (CAD) from radiograph images has been garnering substantial attention. On account of this, we have suggested a deep learning-based automated lightweight architecture which can be used for the development of web application having the least computational burden. The tasks of feature extraction and classification have been automated by using transfer learning based VGG16-inspired 6-layer architecture (I-VGG16), thereby reducing the computational burden of the classification framework. For the suggested web-application, the user will have to upload a knee-radiograph image, and the trained CNN model at the back-end classifies an image accordingly into one of the five KL grades. Finally, the application generates a document identical to the medical report according to obtained severity grade. With an outstanding classification accuracy of 87.95%, our framework outperforms prior work in this area. As a consequence, the excellent performance results demonstrate the model’s ability to be deployed in routine usage for knee OA grade classification.