Identification of Knee Prostheses from Lateral Radiographs Using Deep Learning Techniques

Abstract

The field of medical imaging has seen significant progress in recent years, particularly with the evolution of deep learning techniques. In the context of arthroplasty, the ability to accurately detect and identify specific implant models is crucial for proper patient care and revision surgery. However, manual identification of implants from radiographic images (X-Ray) takes a lot of time and is affected by manual human error. To solve this, this research proposes a deep learning-based novel approach to automate the identification of Knee arthroplasty implants from Lateral (LAT) view X-ray images of the implant. Pre trained convolutional neural networks are used for this purpose. Best results are obtained using VGG16 which produces a higher accuracy of 81.61% and stronger Area Under Curve (AUC) of 0.9547.